Frontend: hamburger menu v hlavicke na mobile

Hlavicka GameListu sa na uzkych displejoch nezmestila do jedneho
riadku. Pod sm: breakpointom sa polozky zbalia do dropdownu.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

.dockerignore

@@ -10,3 +10,6 @@ frontend
 .env
 .env.*
 docker-compose*.yaml
+# trenovacie artefakty do image nepatria (produkcia cita len rl/weights/)
+rl/checkpoints
+rl/runs

.gitignore

@@ -10,3 +10,5 @@ frontend/.vite/
 .env.*
 !.env.example
 geoip/*.mmdb
+rl/runs/
+rl/checkpoints/

Dockerfile

@@ -11,6 +11,7 @@ RUN pip install --no-cache-dir -r requirements.txt
 COPY bridzik.py app.py ./
 COPY api ./api
 COPY db ./db
+COPY rl ./rl
 COPY tests ./tests
 
 RUN useradd --create-home --uid 1000 appuser \

TASKS.md

@@ -11,6 +11,10 @@ Hodinová routine v Claude Code číta túto sekciu, implementuje a presúva hot
 
 ## Hotovo
 
+- [x] Skús upraviť dizajn scrollbaru v zozname bodov v hre tak, aby zodpovedal celej hre. — 2026-07-07: pridaný `.velvet-scroll` v `index.css` (tenký priehľadný track + zlatý polopriehľadný thumb, webkit aj Firefox `scrollbar-color`) a aplikovaný na scrollovateľný zoznam kôl v `Standings.tsx` (herný sidebar aj mobilný panel). Commit 6c95c64.
+- [x] V hre keď sa hádžu karty, pri hodení poslednej karty všetky karty na chvíľu zmiznú a potom sa znovu objavia cez animáciu. Uprav to tak, aby sa posledná karta pridala k predchádzajúcim bez zmiznutia kôpky, a až následne kôpka zmizla smerom k hráčovi, ktorý ju zobral. — 2026-07-07: `GameTable.tsx` číta `previous_stash` synchrónne (namiesto `lingeredStash` nastavovaného v `useEffect`), takže kôpka pri 4. karte nezmizne — posledná karta len pribudne k trom. Po krátkej pauze (`SETTLE_MS`) sa celá kôpka animáciou (`collect-*` keyframes v `index.css`) odsunie k sedadlu víťaza; víťaz sa počíta cez nový `stashWinner` v `gameRules.ts` podľa pravidiel enginu. Commit fb90737.
+- [x] Urob samostatné scrollovanie v zozname bodov v hlavnej hre, lebo teraz tam nevidno všetky hry. Zároveň dorob, aby sa po jednotlivých sériách dali body collapsnúť/expandnúť a ostali by len celkové body za sériu. — 2026-07-07: `Standings.tsx` rozdelený na fixnú hlavičku, scrollovateľný zoznam kôl (`overflow-y-auto`, na mobile `max-h-[45vh]`) a fixné súčty; dokončená séria (Σ riadok) je teraz klikateľná a zbaľuje/rozbaľuje svoje kolá. Commit 80929e7.
+- [x] V hre aj v histórii zruš preciarkavanie pri nesprávnych tipoch. — 2026-07-07: odstránené `line-through` z `Standings.tsx` (herný pohľad) aj `History.tsx` (detail hry) pri neúspešnom tipe. Commit 80929e7.
 - [x] V lobby hre sa dá skopírovať ID hry. Zmeň to tak, aby sa namiesto ID kopírovala celá URL linka na hru. — 2026-07-03: tlačidlo "Kopírovať" v Lobby.tsx teraz kopíruje `${window.location.origin}/lobby/${gid}` namiesto holého gid. Commit 49ac0d3.
 
 <!-- Sem routine presunie dokončené úlohy s dátumom a krátkym popisom + commit hashom. -->

api/__init__.py

@@ -13,8 +13,10 @@ import socketio
 from bridzik import Bridzik, BridzikException, Card
 from db.db import init_db
 from api import auth as auth_module, history
+from api import bots as bots_module
 from api import stats as stats_module
 from api.auth import AuthError, RegistrationIncomplete
+from rl.encoding import index_card
 
 
 def _env_bool(name: str, default: bool) -> bool:
@@ -179,6 +181,9 @@ class Game:
         self.players: list["Player"] = []
         self.started = False
         self.bridzik_core: Bridzik | None = None
+        # Serializuje tahovu slucku botov -- dva sucasne _run_bot_turns tasky
+        # by inak mohli tahat za to iste sedadlo.
+        self.bot_lock = asyncio.Lock()
 
     def start(self):
         self.bridzik_core = Bridzik()
@@ -202,6 +207,10 @@ class Player:
         self.player_id = player_id  # persistent account id (db.models.Player.id)
         self.token = str(uuid.uuid4())  # secret token used for secure reconnect
         self.connected = True
+        # Bot = sedadlo bez socketu; `brain` je rozhodovaci objekt s rozhranim
+        # guess(rnd, seat) / play(rnd, seat) z rl/players.py.
+        self.is_bot = False
+        self.brain = None
 
 
 class CardStatusEncoder(JSONEncoder):
@@ -214,7 +223,13 @@ class CardStatusEncoder(JSONEncoder):
 
 
 def public_games() -> list:
-    """Public lobby view — no sids, no reconnect tokens."""
+    """Public lobby view — no sids, no reconnect tokens.
+
+    A game that finished naturally (all 4 series played out) stays in the
+    `games` dict for reconnect purposes (e.g. a reload while still on the
+    GameOver screen), but it has nothing left to offer the lobby — drop it
+    here rather than have it linger forever as a "started"/resumable entry.
+    """
     return [
         {
             "gid": g.gid,
@@ -226,11 +241,13 @@ def public_games() -> list:
                     "name": p.name,
                     "connected": p.connected,
                     "player_id": p.player_id,
+                    "is_bot": p.is_bot,
                 }
                 for p in g.players
             ],
         }
         for g in games.values()
+        if g.bridzik_core is None or not g.bridzik_core.is_completed()
     ]
 
 
@@ -255,7 +272,8 @@ async def send_game_status(gid: str):
             "completed": core.is_completed(),
             # Self-contained roster so the game view doesn't depend on the lobby snapshot.
             "players": [
-                {"order": p.order, "name": p.name, "connected": p.connected}
+                {"order": p.order, "name": p.name, "connected": p.connected,
+                 "is_bot": p.is_bot}
                 for p in sorted(game.players, key=lambda p: p.order)
             ],
             "series_number": core.series[-1].series_number,
@@ -292,11 +310,16 @@ async def _cleanup_abandoned_lobby(gid: str):
     task neskodny no-op (netreba nic rusit)."""
     await asyncio.sleep(LOBBY_ABANDON_GRACE_SECONDS)
     game = games.get(gid)
-    if game is not None and not game.started and not any(p.connected for p in game.players):
+    if game is not None and not game.started and not _any_human_connected(game):
         del games[gid]
         await broadcast_lobby()
 
 
+def _any_human_connected(game: "Game") -> bool:
+    """Boti su 'pripojeni' stale, pre opustenost lobby sa pocitaju len ludia."""
+    return any(p.connected for p in game.players if not p.is_bot)
+
+
 async def _mark_player_offline(game: "Game", player: "Player"):
     """Mark player disconnected. An unstarted game with nobody left gets a
     delayed cleanup (mobile sockets drop on screen lock, so an immediate
@@ -304,7 +327,7 @@ async def _mark_player_offline(game: "Game", player: "Player"):
     is kept in memory so it stays in the lobby and can be resumed (it's torn
     down only by end_game)."""
     player.connected = False
-    if not any(p.connected for p in game.players) and not game.started:
+    if not _any_human_connected(game) and not game.started:
         asyncio.create_task(_cleanup_abandoned_lobby(game.gid))
     await sio.emit(
         "player_connection",
@@ -345,11 +368,93 @@ def _load_game_into_memory(info: dict) -> "Game":
     for seat, (pid, uname) in enumerate(info["seats"]):
         player = Player(None, uname, seat, pid)
         player.connected = False
+        # Boti sa rozpoznaju konvenciou mena a dostanu novy mozog -- ozivi ich
+        # prvy _kick_bots (napr. ked sa clovek vrati cez rejoin_game).
+        if bots_module.is_bot_username(uname):
+            player.is_bot = True
+            player.brain = bots_module.make_brain(uname)
+            player.connected = True
         game.players.append(player)
     games[info["gid"]] = game
     return game
 
 
+# --- bot turns --------------------------------------------------------------
+
+# Pauza medzi tahmi bota, nech ludia stihaju sledovat hru (0 = okamzite).
+BOT_MOVE_DELAY_SECONDS = float(os.environ.get("BOT_MOVE_DELAY_SECONDS", "0.8"))
+# Kopka na stole sa po dohrati este chvilu zmieta smerom k vitazovi (SETTLE_MS +
+# COLLECT_MS vo frontend/src/pages/GameTable.tsx, spolu 1650ms) -- kym tato
+# animacia nedobehne vsetkym hracom, prvy bot na tahu nesmie zahodit kartu do
+# novej kopky, inak by mu karta "vyletela" uprostred zmetania predoslej.
+TRICK_SWEEP_SECONDS = 1.7
+
+
+def _kick_bots(gid: str) -> None:
+    """Ak je v hre bot, spusti (na pozadi) dohratie vsetkych botich tahov.
+    Vola sa po kazdej akcii, ktora mohla posunut tah na botie sedadlo."""
+    game = games.get(gid)
+    if game is not None and game.started and any(p.is_bot for p in game.players):
+        asyncio.create_task(_run_bot_turns(gid))
+
+
+async def _run_bot_turns(gid: str):
+    """Kym je na tahu botie sedadlo, vykonavaj jeho tahy tym istym internym
+    postupom ako handlery add_guess/play_card (engine validuje, historia sa
+    zapisuje, room dostava game_status). MC vypocet bezi v executori, aby
+    nedrzal event loop ostatnych hier."""
+    game = games.get(gid)
+    if game is None or not game.started or game.bridzik_core is None:
+        return
+    async with game.bot_lock:
+        core = game.bridzik_core
+        loop = asyncio.get_running_loop()
+        while not core.is_completed():
+            rnd = core.series[-1].get_last_round()
+            seat = rnd.get_active_player()
+            bot = game.player_by_order(seat)
+            if bot is None or not bot.is_bot:
+                return  # na tahu je clovek
+            delay = BOT_MOVE_DELAY_SECONDS
+            if rnd.is_guessing_completed() and not rnd.get_last_stash().get_cards() \
+                    and core.get_previous_stash() is not None:
+                # Bot vedie novu kopku a este bezi zmetanie tej predoslej.
+                delay = max(delay, TRICK_SWEEP_SECONDS)
+            if delay > 0:
+                await asyncio.sleep(delay)
+            if games.get(gid) is not game:
+                return  # hru medzitym niekto ukoncil (end_game)
+            played_card = False
+            try:
+                if not rnd.is_guessing_completed():
+                    guess = await loop.run_in_executor(None, bot.brain.guess, rnd, seat)
+                    core.add_player_guess(seat, guess)
+                else:
+                    action = await loop.run_in_executor(None, bot.brain.play, rnd, seat)
+                    core.play_card(seat, index_card(action))
+                    await history.record_completed_rounds(gid, core)
+                    played_card = True
+            except BridzikException as exc:
+                # Nemalo by nastat (bot hra len legalne tahy) -- nezacykli sa,
+                # slucku znovu spusti dalsia akcia cloveka.
+                await send_error_room(gid, str(exc))
+                return
+            # game_status musi ist PRED player_cards -- klient podla neho (novy
+            # round_number + previous_stash) pozna, ze prave zacalo nove kolo, a
+            # dovtedy si drzi starych karty na obrazovke (pozri "displayedHand" vo
+            # frontend/src/pages/GameTable.tsx), kym nedobehne animacia zmetenia
+            # poslednej kopky. Opacne poradie by novu ruku odhalilo predcasne.
+            await send_game_status(gid)
+            if played_card:
+                for player in game.players:
+                    if player.sid:
+                        await send_player_cards(gid, player.order, player.sid)
+
+
+async def send_error_room(gid: str, message: str):
+    await sio.emit("error", {"error": message}, room=gid)
+
+
 # --- connection lifecycle -------------------------------------------------
 
 @sio.event
@@ -481,6 +586,62 @@ async def register_player(sid, gid):
     await broadcast_lobby()
 
 
+@sio.on("add_bot")
+async def add_bot(sid, gid, kind=None):
+    """Hostitel prida bota na najnizsie volne sedadlo nezacatej hry."""
+    sess = sessions.get(sid)
+    if sess is None or sess["gid"] != gid:
+        return await send_error(sid, "Nie ste v tejto hre.")
+    if sess["order"] != 0:
+        return await send_error(sid, "Iba hostitel moze pridavat botov.")
+    game = games.get(gid)
+    if game is None:
+        return await send_error(sid, "Hra neexistuje.")
+    if game.started:
+        return await send_error(sid, "Hra uz zacala.")
+    if len(game.players) >= 4:
+        return await send_error(sid, "Prekroceny pocet hracov.")
+
+    if kind == "neural" and not bots_module.neural_available():
+        return await send_error(sid, "AI bot nie je na tomto serveri dostupny.")
+    if kind not in bots_module.BOT_KINDS:
+        kind = bots_module.DEFAULT_KIND
+    account = await bots_module.ensure_bot_account(
+        kind, {p.player_id for p in game.players}
+    )
+    used = {p.order for p in game.players}
+    order = next(o for o in range(4) if o not in used)
+    player = Player(None, account["username"], order, account["player_id"])
+    player.is_bot = True
+    player.brain = bots_module.make_brain(account["username"])
+    game.players.append(player)
+    await broadcast_lobby()
+
+
+@sio.on("remove_bot")
+async def remove_bot(sid, gid, order):
+    """Hostitel odoberie bota z nezacatej hry (sedadlo sa uvolni)."""
+    sess = sessions.get(sid)
+    if sess is None or sess["gid"] != gid:
+        return await send_error(sid, "Nie ste v tejto hre.")
+    if sess["order"] != 0:
+        return await send_error(sid, "Iba hostitel moze odoberat botov.")
+    game = games.get(gid)
+    if game is None:
+        return await send_error(sid, "Hra neexistuje.")
+    if game.started:
+        return await send_error(sid, "Hra uz zacala.")
+    try:
+        seat = int(order)
+    except (TypeError, ValueError):
+        return await send_error(sid, "Neplatne sedadlo.")
+    player = game.player_by_order(seat)
+    if player is None or not player.is_bot:
+        return await send_error(sid, "Na tomto sedadle nie je bot.")
+    game.players.remove(player)
+    await broadcast_lobby()
+
+
 @sio.on("leave_game")
 async def leave_game(sid):
     """Explicit exit (e.g. a 'Back to lobby' button). The socket stays
@@ -528,7 +689,11 @@ async def start_game(sid, gid):
     await broadcast_lobby()
     await send_game_status(gid)
     for player in game.players:
+        # sid None = bot alebo offline sedadlo -- emit s to=None by karty
+        # broadcastol VSETKYM klientom, preto sa preskakuje.
+        if player.sid:
             await send_player_cards(gid, player.order, player.sid)
+    _kick_bots(gid)
 
 
 @sio.on("end_game")
@@ -583,6 +748,7 @@ async def reconnect_to_game(sid, gid, token):
         await send_player_cards(gid, player.order, sid)
     await sio.emit("player_connection", {"order": player.order, "connected": True}, room=gid)
     await broadcast_lobby()
+    _kick_bots(gid)
 
 
 @sio.on("rejoin_game")
@@ -620,6 +786,7 @@ async def rejoin_game(sid, gid):
         await send_player_cards(gid, player.order, sid)
     await sio.emit("player_connection", {"order": player.order, "connected": True}, room=gid)
     await broadcast_lobby()
+    _kick_bots(gid)
 
 
 @sio.on("restore_game")
@@ -678,6 +845,7 @@ async def add_guess(sid, guess):
     except BridzikException as exc:
         return await send_error(sid, str(exc))
     await send_game_status(game.gid)
+    _kick_bots(game.gid)
 
 
 @sio.on("play_card")
@@ -703,7 +871,14 @@ async def play_card(sid, card_key):
     await history.record_completed_rounds(game.gid, core)
     await send_game_status(game.gid)
     for player in game.players:
+        if player.sid:  # None (bot/offline) by broadcastoval karty vsetkym
             await send_player_cards(game.gid, player.order, player.sid)
+    # A naturally-finished game (all 4 series played out) has nothing left to
+    # offer the lobby -- refresh the list so it drops out immediately instead
+    # of lingering as "started"/resumable until someone happens to leave it.
+    if core.is_completed():
+        await broadcast_lobby()
+    _kick_bots(game.gid)
 
 
 # --- history (read-only) --------------------------------------------------

api/bots.py

@@ -0,0 +1,120 @@
+"""In-process boti: DB ucty botov a ich rozhodovacie "mozgy" z rl/players.py.
+
+Bot je normalny hrac na sedadle -- ma riadok v tabulke `players` (aby
+historia, standings a restore fungovali bez zmeny), ale ziadny socket.
+Tahovu slucku botov ma api/__init__.py (_run_bot_turns); tu je len to,
+co potrebuje DB a rl vrstvu.
+
+Bezpecnost botich uctov: totp_secret je nahodny a nikde sa neuklada v
+citatelnej podobe, totp_last_step sa nastavi na aktualny krok -- ucet tym
+padom NIE JE "nedokoncena registracia" (viz auth._is_unconfirmed), takze
+register_account ho odmietne prepisat a login bez secretu neprejde.
+"""
+
+import os
+from random import Random
+
+import pyotp
+from sqlalchemy import select
+
+from api import auth
+from db import crypto
+from db.db import async_session
+from db.models import Player
+from rl.players import HeuristicPlayer, RandomPlayer
+from rl.pure_net import DEFAULT_WEIGHTS_PATH, PureNet, PureNeuralPlayer
+
+# Prefix je konvencia na rozpoznanie bota (aj po restarte servera, kedy sa
+# sedadla obnovuju z DB len ako (player_id, username)).
+BOT_PREFIX = "bot:"
+DEFAULT_KIND = "heuristic"
+
+# Natrenovana siet -- vahy (rl/weights/, export z rl/export.py) sa nacitaju
+# raz a zdielaju medzi botmi (PureNet je bezstavovy, len cita).
+_pure_net: PureNet | None = None
+
+
+def _neural_brain() -> PureNeuralPlayer:
+    global _pure_net
+    if _pure_net is None:
+        _pure_net = PureNet.load()
+    return PureNeuralPlayer(_pure_net)
+
+
+def neural_available() -> bool:
+    return os.path.exists(DEFAULT_WEIGHTS_PATH)
+
+
+_BRAINS = {
+    "heuristic": lambda: HeuristicPlayer(Random()),
+    "random": lambda: RandomPlayer(Random()),
+    "neural": _neural_brain,
+}
+BOT_KINDS = tuple(_BRAINS)
+
+
+def available_kinds() -> tuple:
+    """Druhy botov ponuknutelne na tomto serveri (neural len s vahami)."""
+    return tuple(k for k in _BRAINS if k != "neural" or neural_available())
+
+
+def is_bot_username(username: str) -> bool:
+    return bool(username) and username.startswith(BOT_PREFIX)
+
+
+def kind_of(username: str) -> str:
+    """'bot:heuristic-2' -> 'heuristic'; neznamy druh padne na DEFAULT_KIND."""
+    body = username[len(BOT_PREFIX):]
+    kind = body.rsplit("-", 1)[0]
+    return kind if kind in _BRAINS else DEFAULT_KIND
+
+
+def make_brain(username: str):
+    """Rozhodovaci objekt (guess/play rozhranie z rl/players.py) pre bota.
+
+    Neural bez suboru vah (napr. restore hry na serveri bez exportu) padne
+    na heuristiku -- sedadlo hra dalej, len inym mozgom.
+    """
+    kind = kind_of(username)
+    if kind == "neural" and not neural_available():
+        kind = DEFAULT_KIND
+    return _BRAINS[kind]()
+
+
+def _suffix_number(username: str) -> int:
+    try:
+        return int(username.rsplit("-", 1)[1])
+    except (IndexError, ValueError):
+        return 0
+
+
+async def ensure_bot_account(kind: str, exclude_ids: set) -> dict:
+    """Najde alebo zalozi boti ucet daneho druhu; vrati {player_id, username}.
+
+    `exclude_ids` su ucty uz obsadene v danej hre -- kazde sedadlo potrebuje
+    INY ucet (Game.playerN_id aj unikat v Guess predpokladaju 4 rozne ID).
+    Ucty sa cisluju bot:<kind>-1, -2, ... a recykluju sa medzi hrami.
+    """
+    prefix = f"{BOT_PREFIX}{kind}-"
+    async with async_session() as session:
+        rows = (
+            await session.scalars(
+                select(Player).where(Player.username.like(prefix + "%"))
+            )
+        ).all()
+        for player in sorted(rows, key=lambda p: _suffix_number(p.username)):
+            if player.id not in exclude_ids:
+                return {"player_id": player.id, "username": player.username}
+
+        number = 1 + max((_suffix_number(p.username) for p in rows), default=0)
+        username = f"{prefix}{number}"
+        player = Player(
+            username=username,
+            # nahodny secret, ktory sa zahodi -- nikto sa zan neprihlasi
+            totp_secret=crypto.encrypt(pyotp.random_base32()),
+            # nenulovy last_step = ucet sa netvari ako nedokoncena registracia
+            totp_last_step=auth._current_step(),
+        )
+        session.add(player)
+        await session.commit()
+        return {"player_id": player.id, "username": player.username}

bridzik.py

@@ -19,6 +19,10 @@ class Card_colors(Enum):
             return self.name == other.name
         return NotImplemented
 
+    # vlastne __eq__ rusi zdedeny __hash__ -- obnovit konzistentne s __eq__
+    def __hash__(self):
+        return hash(self.name)
+
 
 class Card_values(Enum):
     C7 = 1
@@ -51,6 +55,10 @@ class Card_values(Enum):
             return self.name == other.name
         return NotImplemented
 
+    # vlastne __eq__ rusi zdedeny __hash__ -- obnovit konzistentne s __eq__
+    def __hash__(self):
+        return hash(self.name)
+
 
 class Card():
     def __init__(self, color: Card_colors, value: Card_values):
@@ -63,6 +71,10 @@ class Card():
                 and self.value == other.value
         return NotImplemented
 
+    # vlastne __eq__ rusi zdedeny __hash__ -- obnovit konzistentne s __eq__
+    def __hash__(self):
+        return hash((self.color, self.value))
+
     def __str__(self):
         return '{}_{}'.format(self.color.name, self.value.name)
 

frontend/src/components/Hand.tsx

@@ -26,7 +26,6 @@ interface Props {
 
 export default function Hand({ hand, myTurn, isPlayPhase, playableKeys, desktop = false }: Props) {
   const groups = groupedByColor(hand);
-  if (groups.length === 0) return null;
 
   const canPlay = isPlayPhase && myTurn;
 
@@ -52,6 +51,10 @@ export default function Hand({ hand, myTurn, isPlayPhase, playableKeys, desktop
         <div className="h-px flex-1 max-w-[80px] bg-gradient-to-l from-transparent to-gold/20" />
       </div>
 
+      {/* Fixed height reserves the card row even when the hand is briefly
+          empty (last card of a round just played, next deal not in yet) --
+          otherwise this whole area collapses and the layout jumps. */}
+      <div className="flex items-end justify-center" style={{ minHeight: desktop ? 100 : 84 }}>
         {desktop ? (
           // Desktop has room — keep cards grouped by suit, wrap if needed.
           <div className="flex flex-wrap gap-3 justify-center items-end">
@@ -67,6 +70,7 @@ export default function Hand({ hand, myTurn, isPlayPhase, playableKeys, desktop
           <MobileHand groups={groups} cardProps={cardProps} />
         )}
       </div>
+    </div>
   );
 }
 

frontend/src/components/HeaderMenu.tsx

@@ -0,0 +1,100 @@
+import { useEffect, useRef, useState } from 'react';
+
+interface Props {
+  username?: string;
+  onHistory: () => void;
+  onDonate: () => void;
+  onLogout: () => void;
+}
+
+/** Header navigation: a row of links from `sm:` up, a hamburger dropdown below it. */
+export default function HeaderMenu({ username, onHistory, onDonate, onLogout }: Props) {
+  const [open, setOpen] = useState(false);
+  const ref = useRef<HTMLDivElement>(null);
+
+  useEffect(() => {
+    if (!open) return;
+    const onPointerDown = (e: PointerEvent) => {
+      if (!ref.current?.contains(e.target as Node)) setOpen(false);
+    };
+    const onKeyDown = (e: KeyboardEvent) => {
+      if (e.key === 'Escape') setOpen(false);
+    };
+    document.addEventListener('pointerdown', onPointerDown);
+    document.addEventListener('keydown', onKeyDown);
+    return () => {
+      document.removeEventListener('pointerdown', onPointerDown);
+      document.removeEventListener('keydown', onKeyDown);
+    };
+  }, [open]);
+
+  const pick = (action: () => void) => () => {
+    setOpen(false);
+    action();
+  };
+
+  return (
+    <>
+      <div className="hidden sm:flex items-center gap-3 text-sm">
+        <span className="text-green-dim">{username}</span>
+        <button onClick={onHistory} className="text-gold hover:text-gold-bright">
+          História
+        </button>
+        <button onClick={onDonate} className="text-gold hover:text-gold-bright">
+          Na kávu
+        </button>
+        <button onClick={onLogout} className="text-green-dim hover:text-gold">
+          Odhlásiť
+        </button>
+      </div>
+
+      <div ref={ref} className="relative sm:hidden">
+        <button
+          onClick={() => setOpen((o) => !o)}
+          aria-label="Menu"
+          aria-expanded={open}
+          aria-haspopup="menu"
+          className="flex flex-col justify-center gap-[5px] w-9 h-9 items-center rounded-lg border border-gold/20 text-gold hover:border-gold/50 transition-colors"
+        >
+          <span className="block w-4 h-px bg-current" />
+          <span className="block w-4 h-px bg-current" />
+          <span className="block w-4 h-px bg-current" />
+        </button>
+
+        {open && (
+          <div
+            role="menu"
+            className="absolute right-0 top-11 z-40 w-44 bg-header border border-[#142018] rounded-xl py-1 shadow-[0_20px_60px_rgba(0,0,0,.6)]"
+          >
+            {username && (
+              <p className="px-4 py-2 text-xs text-green-dim border-b border-[#142018] truncate">
+                {username}
+              </p>
+            )}
+            <button
+              role="menuitem"
+              onClick={pick(onHistory)}
+              className="block w-full text-left px-4 py-2.5 text-sm text-gold hover:bg-gold/10"
+            >
+              História
+            </button>
+            <button
+              role="menuitem"
+              onClick={pick(onDonate)}
+              className="block w-full text-left px-4 py-2.5 text-sm text-gold hover:bg-gold/10"
+            >
+              Na kávu
+            </button>
+            <button
+              role="menuitem"
+              onClick={pick(onLogout)}
+              className="block w-full text-left px-4 py-2.5 text-sm text-green-dim hover:bg-gold/10 hover:text-gold"
+            >
+              Odhlásiť
+            </button>
+          </div>
+        )}
+      </div>
+    </>
+  );
+}

frontend/src/components/PlayerCircle.tsx

@@ -12,25 +12,34 @@ interface Props {
 export default function PlayerCircle({ name, won, guess, active, size = 52 }: Props) {
   const nameFont = Math.max(9, Math.round(size * 0.17));
   const valueFont = Math.round(size * 0.32);
-  // Oval: width = size, height a touch shorter so it reads as an ellipse.
+  // Oval: height a touch shorter than size so it reads as an ellipse. Width
+  // starts at `size` (a circle for short names) but grows with the name via
+  // fit-content + padding, up to a cap beyond which the name is ellipsised
+  // rather than wrapping (wrapping would break the fixed height/oval shape).
   const height = Math.round(size * 0.78);
+  const hPad = Math.round(size * 0.16);
+  const maxWidth = Math.round(size * 2);
 
   return (
     <div
       // Only the active player is highlighted (gold ring + glow) — colors come
       // from the velvet-table palette tokens (tailwind.config.js), not literals.
-      className={`flex flex-col items-center justify-center rounded-[50%] ${
+      className={`flex flex-col items-center justify-center rounded-full ${
         active ? 'bg-circle-active border-2 border-gold' : 'bg-circle border-[1.5px] border-gold/20'
       }`}
       style={{
-        width: size,
+        width: 'fit-content',
+        minWidth: size,
+        maxWidth,
         height,
+        paddingLeft: hPad,
+        paddingRight: hPad,
         boxShadow: '0 2px 10px rgba(0,0,0,.45)',
         animation: active ? 'ar 2.2s ease-in-out infinite' : undefined,
       }}
     >
       <span
-        className={`uppercase leading-tight text-center ${active ? 'text-gold' : 'text-green-circle'}`}
+        className={`uppercase leading-tight text-center overflow-hidden text-ellipsis whitespace-nowrap max-w-full ${active ? 'text-gold' : 'text-green-circle'}`}
         style={{
           fontFamily: '"DM Sans",sans-serif',
           fontSize: nameFont,

frontend/src/components/Standings.tsx

@@ -1,6 +1,7 @@
 import { useState } from 'react';
 import type { PlayerInfo } from '../types';
 import { computeTotal } from '../lib/standings';
+import { displayName } from '../lib/names';
 
 interface Props {
   standings: number[][][];
@@ -14,6 +15,15 @@ interface Props {
 
 export default function Standings({ standings, guesses = [], players, myOrder, desktop = false }: Props) {
   const [open, setOpen] = useState(false);
+  // Finished series (8/8 rounds) collapsed to just their Σ row; toggled per series.
+  const [collapsedSeries, setCollapsedSeries] = useState<Set<number>>(new Set());
+  const toggleSeries = (si: number) =>
+    setCollapsedSeries((prev) => {
+      const next = new Set(prev);
+      if (next.has(si)) next.delete(si);
+      else next.add(si);
+      return next;
+    });
 
   // Player columns in seat order; the local player's column is highlighted.
   const cols = [...players].sort((a, b) => a.order - b.order);
@@ -39,13 +49,11 @@ export default function Standings({ standings, guesses = [], players, myOrder, d
     total: desktop ? 20 : 18,
   };
 
-  const table = (
-    <div className="flex-1 flex flex-col px-3 pt-3 pb-4">
-      {/* Column headers */}
-      <div
-        className="grid items-end mb-1"
-        style={{ gridTemplateColumns: `28px repeat(${cols.length}, 1fr)` }}
-      >
+  const gridCols = { gridTemplateColumns: `28px repeat(${cols.length}, 1fr)` };
+
+  const columnHeader = (
+    <div className="px-3 pt-3 flex-shrink-0">
+      <div className="grid items-end mb-1" style={gridCols}>
         <div />
         {cols.map((p) => (
           <div
@@ -55,32 +63,35 @@ export default function Standings({ standings, guesses = [], players, myOrder, d
             }`}
             style={{ fontSize: fz.head }}
           >
-            {p.name}
+            {displayName(p.name)}
           </div>
         ))}
       </div>
-      <div className="h-px bg-gold/10 mb-1" />
+      <div className="h-px bg-gold/10" />
+    </div>
+  );
 
-      {/* Completed rounds, grouped by series with a per-series summary row */}
-      {standings.flatMap((seriesRounds, si) => {
+  // Completed rounds, grouped by series with a per-series summary row. Finished
+  // series can be collapsed to just their Σ row so long games stay scannable.
+  const rows = standings.flatMap((seriesRounds, si) => {
     const priorRounds = seriesRoundOffsets[si];
-        const elems = seriesRounds.map((scores, lri) => (
-          <div
-            key={`r-${si}-${lri}`}
-            className="grid items-center py-1 border-b border-gold/[.05]"
-            style={{ gridTemplateColumns: `28px repeat(${cols.length}, 1fr)` }}
-          >
+    const isFinished = seriesRounds.length === ROUNDS_PER_SERIES;
+    const isCollapsed = isFinished && collapsedSeries.has(si);
+    const elems = isCollapsed
+      ? []
+      : seriesRounds.map((scores, lri) => (
+          <div key={`r-${si}-${lri}`} className="grid items-center py-1 border-b border-gold/[.05]" style={gridCols}>
             <div className="text-center text-[#7a7252]" style={{ fontSize: fz.idx }}>
               {priorRounds + lri + 1}
             </div>
             {cols.map((p) => {
               const points = scores[p.order] ?? 0;
-              // Failed tip (0 points) → show the struck-through tip instead of 0.
+              // Failed tip → show the tip in the dimmer "0 points" color, no strikethrough.
               if (points === 0) {
                 return (
                   <div
                     key={p.order}
-                    className="text-center font-serif leading-none line-through"
+                    className="text-center font-serif leading-none"
                     style={{ fontSize: fz.cell, color: '#7a6e4a' }}
                   >
                     {guesses[si]?.[lri]?.[p.order] ?? 0}
@@ -100,15 +111,21 @@ export default function Standings({ standings, guesses = [], players, myOrder, d
           </div>
         ));
 
-        // After a finished series, sum its points per player.
-        if (seriesRounds.length === ROUNDS_PER_SERIES) {
+    // After a finished series, sum its points per player. Clicking toggles
+    // whether that series' individual rounds are shown.
+    if (isFinished) {
       elems.push(
         <div
           key={`s-${si}`}
-              className="grid items-center py-1 my-0.5 rounded bg-gold/[.07]"
-              style={{ gridTemplateColumns: `28px repeat(${cols.length}, 1fr)` }}
+          role="button"
+          tabIndex={0}
+          onClick={() => toggleSeries(si)}
+          onKeyDown={(e) => (e.key === 'Enter' || e.key === ' ') && toggleSeries(si)}
+          className="grid items-center py-1 my-0.5 rounded bg-gold/[.07] cursor-pointer select-none"
+          style={gridCols}
         >
-              <div className="text-center font-serif text-gold" style={{ fontSize: fz.sigma }}>
+          <div className="text-center font-serif text-gold flex items-center justify-center gap-[2px]" style={{ fontSize: fz.sigma }}>
+            <span className="text-[8px] text-green-dim">{isCollapsed ? '▸' : '▾'}</span>
             Σ{si + 1}
           </div>
           {cols.map((p) => {
@@ -129,27 +146,26 @@ export default function Standings({ standings, guesses = [], players, myOrder, d
       );
     }
     return elems;
-      })}
+  });
+
+  const scrollableRounds = (
+    <div className={`velvet-scroll flex-1 min-h-0 overflow-y-auto px-3 ${desktop ? '' : 'max-h-[45vh]'}`}>
+      {rows}
 
       {/* Active round placeholder */}
-      <div
-        className="grid items-center py-1 rounded mt-0.5 bg-gold/[.04]"
-        style={{ gridTemplateColumns: `28px repeat(${cols.length}, 1fr)` }}
-      >
+      <div className="grid items-center py-1 rounded mt-0.5 bg-gold/[.04]" style={gridCols}>
         <div className="text-center font-medium text-gold" style={{ fontSize: fz.idx }}>{completedRounds + 1}</div>
         {cols.map((p) => (
           <div key={p.order} className="text-center text-[#7a7252]" style={{ fontSize: fz.dot }}>·</div>
         ))}
       </div>
+    </div>
+  );
 
-      <div className="flex-1 min-h-2" />
+  const totalsBlock = (
+    <div className="px-3 pb-4 pt-2 flex-shrink-0">
       <div className="h-px bg-gold/20 mb-2" />
-
-      {/* Totals */}
-      <div
-        className="grid items-center py-0.5"
-        style={{ gridTemplateColumns: `28px repeat(${cols.length}, 1fr)` }}
-      >
+      <div className="grid items-center py-0.5" style={gridCols}>
         <div className="text-center uppercase tracking-[.08em] text-green-dim" style={{ fontSize: fz.sigma }}>
           Σ
         </div>
@@ -168,13 +184,21 @@ export default function Standings({ standings, guesses = [], players, myOrder, d
     </div>
   );
 
+  const content = (
+    <div className="flex-1 min-h-0 flex flex-col">
+      {columnHeader}
+      {scrollableRounds}
+      {totalsBlock}
+    </div>
+  );
+
   if (desktop) {
     return (
       <aside className="w-[268px] flex-shrink-0 bg-header border-l border-[#142018] flex flex-col">
-        <div className="h-[58px] flex items-center gap-2 px-5 border-b border-[#14221a]">
+        <div className="h-[58px] flex items-center gap-2 px-5 border-b border-[#14221a] flex-shrink-0">
           <span className="font-serif uppercase tracking-[.12em] text-[13px] text-gold">Skóre</span>
         </div>
-        {table}
+        {content}
       </aside>
     );
   }
@@ -189,7 +213,7 @@ export default function Standings({ standings, guesses = [], players, myOrder, d
         <span>Skóre</span>
         <span className="text-green-dim">{open ? '▲' : '▼'}</span>
       </button>
-      {open && table}
+      {open && content}
     </div>
   );
 }

frontend/src/components/Trick.tsx

@@ -1,5 +1,6 @@
 import type { PlayerInfo, StashData } from '../types';
 import CardView from './CardView';
+import { displayName } from '../lib/names';
 
 interface Props {
   stash: StashData | null;
@@ -18,7 +19,7 @@ export default function Trick({ stash, players, myOrder }: Props) {
     : [];
 
   const nameFor = (order: number) =>
-    players.find((p) => p.order === order)?.name ?? '';
+    displayName(players.find((p) => p.order === order)?.name);
 
   const overlap = -16;
   const slotH = 80;

frontend/src/index.css

@@ -30,6 +30,27 @@
   }
 }
 
+/* Slim, on-theme scrollbar (velvet felt + gold accent) for opt-in scroll areas
+   such as the in-game score list, instead of the chunky OS default. */
+.velvet-scroll {
+  scrollbar-width: thin;
+  scrollbar-color: rgba(201, 168, 76, 0.35) transparent;
+}
+.velvet-scroll::-webkit-scrollbar {
+  width: 7px;
+  height: 7px;
+}
+.velvet-scroll::-webkit-scrollbar-track {
+  background: transparent;
+}
+.velvet-scroll::-webkit-scrollbar-thumb {
+  background: rgba(201, 168, 76, 0.28);
+  border-radius: 4px;
+}
+.velvet-scroll::-webkit-scrollbar-thumb:hover {
+  background: rgba(201, 168, 76, 0.5);
+}
+
 /* Velvet-table animations (design handoff). Declared as raw CSS so they work
    both via Tailwind's animate-* utilities and inline `animation:` strings. */
 @keyframes tp {
@@ -70,3 +91,21 @@
   from { opacity: 0; transform: translateX(110px) scale(0.82); }
   to { opacity: 1; transform: none; }
 }
+
+/* A completed trick is swept off the table towards the seat that won it. */
+@keyframes collect-bottom {
+  from { opacity: 1; transform: none; }
+  to { opacity: 0; transform: translateY(150px) scale(0.66); }
+}
+@keyframes collect-top {
+  from { opacity: 1; transform: none; }
+  to { opacity: 0; transform: translateY(-150px) scale(0.66); }
+}
+@keyframes collect-left {
+  from { opacity: 1; transform: none; }
+  to { opacity: 0; transform: translateX(-190px) scale(0.66); }
+}
+@keyframes collect-right {
+  from { opacity: 1; transform: none; }
+  to { opacity: 0; transform: translateX(190px) scale(0.66); }
+}

frontend/src/lib/gameRules.ts

@@ -1,4 +1,4 @@
-import type { CardColor, Hand } from '../types';
+import type { CardColor, CardValue, Hand, StashData } from '../types';
 
 export function computePlayable(hand: Hand, ledColor: CardColor | null): Set<string> {
   const keys = Object.keys(hand);
@@ -13,6 +13,35 @@ export function computePlayable(hand: Hand, ledColor: CardColor | null): Set<str
   return new Set(keys);
 }
 
+const VALUE_ORDER: CardValue[] = ['C7', 'C8', 'C9', 'C10', 'LOWER', 'UPPER', 'KING', 'ACE'];
+
+/** Seat that wins a completed 4-card trick — mirrors Stash.get_winner in the
+ *  engine: HEARTS (červeň) is the permanent trump, otherwise the highest card
+ *  of the led colour wins. Safe to call on a partial stash (returns the current
+ *  leader among the cards played so far). */
+export function stashWinner(stash: StashData): number {
+  const led = stash.cards[String(stash.first_player)];
+  if (!led) return stash.first_player;
+  let winner = stash.first_player;
+  let best = led;
+  for (let i = 0; i < 4; i++) {
+    const c = stash.cards[String(i)];
+    if (!c) continue;
+    if (c.color === led.color || c.color === 'HEARTS') {
+      if (c.color === best.color) {
+        if (VALUE_ORDER.indexOf(c.value) >= VALUE_ORDER.indexOf(best.value)) {
+          best = c;
+          winner = i;
+        }
+      } else if (c.color === 'HEARTS') {
+        best = c;
+        winner = i;
+      }
+    }
+  }
+  return winner;
+}
+
 /** The bid the last guesser may not make: the four bids must not sum to the
  *  number of tricks in the round (mirrors Round.add_player_guess in the engine).
  *  Returns null while earlier players are still guessing. */

frontend/src/lib/names.ts

@@ -0,0 +1,22 @@
+/** Display helpers for player names. Bot accounts follow the server-side
+ *  convention "bot:<kind>-<n>" (see api/bots.py) — render them as a short
+ *  friendly label instead of the raw username. */
+
+const BOT_PREFIX = 'bot:';
+
+export function isBotName(name?: string | null): boolean {
+  return !!name && name.startsWith(BOT_PREFIX);
+}
+
+/** "bot:heuristic-2" -> "Bot 2", "bot:neural-1" -> "AI bot 1"; other kinds
+ *  keep a suffix ("Bot 1 (random)"); non-bot names pass through unchanged. */
+export function displayName(name?: string | null): string {
+  if (!name || !isBotName(name)) return name ?? '';
+  const body = name.slice(BOT_PREFIX.length);
+  const dash = body.lastIndexOf('-');
+  const kind = dash > 0 ? body.slice(0, dash) : body;
+  const num = dash > 0 ? body.slice(dash + 1) : '';
+  if (kind === 'neural') return num ? `AI bot ${num}` : 'AI bot';
+  const label = num ? `Bot ${num}` : 'Bot';
+  return kind === 'heuristic' ? label : `${label} (${kind})`;
+}

frontend/src/lib/socket.ts

@@ -50,6 +50,10 @@ export const emit = {
   // Reopen a prematurely-ended game from history back into the lobby.
   restoreGame: (gid: string) => socket.emit('restore_game', gid),
   leaveGame: () => socket.emit('leave_game'),
+  // Bots: host-only, before the game starts (seat picked by the server).
+  // kind: 'heuristic' (default) | 'neural' (trained net) | 'random'
+  addBot: (gid: string, kind: string = 'heuristic') => socket.emit('add_bot', gid, kind),
+  removeBot: (gid: string, order: number) => socket.emit('remove_bot', gid, order),
   endGame: (gid: string) => socket.emit('end_game', gid),
   startGame: (gid: string) => socket.emit('start_game', gid),
   reconnectToGame: (gid: string, token: string) => socket.emit('reconnect_to_game', gid, token),

frontend/src/pages/GameList.tsx

@@ -3,6 +3,7 @@ import { useNavigate } from 'react-router-dom';
 import { useGameStore } from '../store/gameStore';
 import { emit, socket, setAuthToken } from '../lib/socket';
 import { trackEvent } from '../lib/track';
+import HeaderMenu from '../components/HeaderMenu';
 import NameModal from '../components/NameModal';
 import RulesModal from '../components/RulesModal';
 
@@ -33,18 +34,12 @@ export default function GameList() {
     <div className="max-w-md mx-auto p-4 pt-8 min-h-screen">
       <div className="flex items-center justify-between mb-6">
         <h1 className="font-serif text-2xl text-gold tracking-wide">Bridžik</h1>
-        <div className="flex items-center gap-3 text-sm">
-          <span className="text-green-dim">{account?.username}</span>
-          <button onClick={() => navigate('/history')} className="text-gold hover:text-gold-bright">
-            História
-          </button>
-          <button onClick={() => navigate('/donate')} className="text-gold hover:text-gold-bright">
-            Na kávu
-          </button>
-          <button onClick={handleLogout} className="text-green-dim hover:text-gold">
-            Odhlásiť
-          </button>
-        </div>
+        <HeaderMenu
+          username={account?.username}
+          onHistory={() => navigate('/history')}
+          onDonate={() => navigate('/donate')}
+          onLogout={handleLogout}
+        />
       </div>
 
       <div className="flex flex-col gap-3 mb-6">

frontend/src/pages/GameOver.tsx

@@ -2,6 +2,7 @@ import { useNavigate } from 'react-router-dom';
 import type { PlayerInfo } from '../types';
 import { computeTotal } from '../lib/standings';
 import { leaveGame } from '../lib/leaveGame';
+import { displayName } from '../lib/names';
 
 interface Props {
   players: PlayerInfo[];
@@ -30,7 +31,7 @@ export default function GameOver({ players, standings }: Props) {
           >
             <div className="flex items-center gap-3">
               <span className="text-2xl w-8">{medals[i]}</span>
-              <span className="font-serif text-green-score">{p.name}</span>
+              <span className="font-serif text-green-score">{displayName(p.name)}</span>
             </div>
             <span className={`font-serif text-xl ${i === 0 ? 'text-gold-bright' : 'text-gold'}`}>
               {p.total}

frontend/src/pages/GameTable.tsx

@@ -3,8 +3,9 @@ import { useNavigate } from 'react-router-dom';
 import { useGameStore } from '../store/gameStore';
 import { emit } from '../lib/socket';
 import { leaveGame } from '../lib/leaveGame';
-import { computePlayable } from '../lib/gameRules';
+import { computePlayable, stashWinner } from '../lib/gameRules';
 import { computeTotal } from '../lib/standings';
+import { displayName } from '../lib/names';
 import { useIsDesktop } from '../lib/useIsDesktop';
 import { useFitScale } from '../lib/useFitScale';
 import Hand from '../components/Hand';
@@ -14,9 +15,14 @@ import Standings from '../components/Standings';
 import PlayerCircle from '../components/PlayerCircle';
 import FaceDownCards from '../components/FaceDownCards';
 import GameOver from './GameOver';
-import type { PlayerInfo, StashData } from '../types';
+import type { Hand as HandCards, PlayerInfo, StashData } from '../types';
 
-const TRICK_LINGER_MS = 3000;
+// A completed trick stays face-up for SETTLE_MS (so the last card visibly joins
+// the pile), then is swept towards the winner over COLLECT_MS.
+const SETTLE_MS = 1100;
+const COLLECT_MS = 550;
+// Sweep direction by the winner's seat offset from me: 0=me(bottom) 1=left 2=top 3=right.
+const COLLECT_BY_OFFSET = ['collect-bottom', 'collect-left', 'collect-top', 'collect-right'];
 
 export default function GameTable() {
   const navigate = useNavigate();
@@ -28,28 +34,75 @@ export default function GameTable() {
   const gameStatus = useGameStore((s) => s.gameStatus);
   const hand = useGameStore((s) => s.hand);
 
-  // Hold the last completed trick visible for TRICK_LINGER_MS after it finishes.
-  const [lingeredStash, setLingeredStash] = useState<StashData | null>(null);
-  const lingerTimer = useRef<ReturnType<typeof setTimeout> | null>(null);
+  // Once a completed trick has been swept away, its key is remembered here so it
+  // is not shown again while we wait for the winner to lead the next trick.
+  const [dismissedKey, setDismissedKey] = useState<string | null>(null);
+  // Turns on for the collect (fly-to-winner) phase, after the settle pause.
+  const [collecting, setCollecting] = useState(false);
 
   const previousStash = gameStatus?.status.previous_stash ?? null;
   // Every game_status payload recreates the stash object, so identify the trick
-  // by content — the timer must restart only when a *different* trick completes.
+  // by content — the timers must restart only when a *different* trick completes.
   const previousStashKey = previousStash
     ? `${previousStash.first_player}:${JSON.stringify(previousStash.cards)}`
     : null;
 
+  // On first load of an already-running game (reconnect / restore-on-startup) a
+  // completed trick is already present; adopt it as "already swept" so we don't
+  // replay a stale sweep over the live board. A freshly started game has no
+  // completed trick at this point, so its very first trick still animates.
+  const booted = useRef(false);
   useEffect(() => {
-    if (!previousStash) return;
-    setLingeredStash(previousStash);
-    if (lingerTimer.current) clearTimeout(lingerTimer.current);
-    lingerTimer.current = setTimeout(() => setLingeredStash(null), TRICK_LINGER_MS);
+    if (booted.current || !gameStatus) return;
+    booted.current = true;
+    if (previousStashKey) setDismissedKey(previousStashKey);
+  }, [gameStatus, previousStashKey]);
+
+  useEffect(() => {
+    if (!previousStashKey) return;
+    setCollecting(false);
+    const settle = setTimeout(() => setCollecting(true), SETTLE_MS);
+    const done = setTimeout(() => {
+      setCollecting(false);
+      setDismissedKey(previousStashKey);
+    }, SETTLE_MS + COLLECT_MS);
     return () => {
-      if (lingerTimer.current) clearTimeout(lingerTimer.current);
+      clearTimeout(settle);
+      clearTimeout(done);
     };
-    // eslint-disable-next-line react-hooks/exhaustive-deps
   }, [previousStashKey]);
 
+  // A just-completed trick that hasn't been swept away yet always wins the centre
+  // — even once the winner has already led the next trick. The engine reveals that
+  // lead card (and, at a round boundary, the next bidding phase) the instant the
+  // 4th card lands, so without holding the pile here the sweep would be cut off
+  // after every trick. Reading `previousStash` synchronously (rather than a state
+  // set in an effect) also means the pile never blinks to empty on the frame the
+  // 4th card lands — the last card simply joins the three already there, then the
+  // whole pile is collected before the next trick takes over.
+  const finishing = previousStashKey !== null && previousStashKey !== dismissedKey;
+
+  // The new round's dealt hand arrives (via `player_cards`) the instant the last
+  // trick of the previous round is played, but the centre oval is still sweeping
+  // that trick away — hold an empty hand on screen (the previous round's last card
+  // really was just played, the engine just never broadcasts that "0 cards" beat on
+  // its own since it deals the new round in the same step) until the sweep finishes,
+  // so the new cards don't appear before the previous round has visibly wrapped up.
+  // Mid-round plays (round number unchanged) still update instantly, since that's
+  // just the player's own card leaving their hand, not a fresh deal.
+  const [displayedHand, setDisplayedHand] = useState<HandCards>(hand);
+  const lastAppliedRoundRef = useRef<number | null>(gameStatus?.round_number ?? null);
+  useEffect(() => {
+    if (!gameStatus) return;
+    const isNewRound = gameStatus.round_number !== lastAppliedRoundRef.current;
+    if (isNewRound && finishing) {
+      setDisplayedHand({}); // last card of the previous round is gone; new deal waits
+      return;
+    }
+    lastAppliedRoundRef.current = gameStatus.round_number;
+    setDisplayedHand(hand);
+  }, [hand, gameStatus, finishing]);
+
   if (!gameStatus || !myPlayer) {
     return <p className="text-center text-green-dim pt-20 font-serif italic">Načítava sa…</p>;
   }
@@ -73,14 +126,27 @@ export default function GameTable() {
   const myTurnToPlay = isPlayPhase && active_player === myOrder;
 
   const activeCards = active_stash ? Object.keys(active_stash.cards).length : 0;
-  const displayedStash: StashData | null =
-    activeCards > 0 && active_stash ? active_stash : lingeredStash ?? null;
+  const displayedStash: StashData | null = finishing
+    ? previousStash
+    : activeCards > 0 && active_stash
+      ? active_stash
+      : null;
 
-  const playableKeys = myTurnToPlay && active_stash
-    ? computePlayable(hand, active_stash.cards[String(active_stash.first_player)]?.color ?? null)
+  // During the collect phase, sweep the pile towards whoever won it.
+  const collectAnim =
+    collecting && finishing && previousStash
+      ? COLLECT_BY_OFFSET[(stashWinner(previousStash) - myOrder + 4) % 4]
+      : null;
+
+  // Block play until the previous trick's sweep animation has finished — otherwise
+  // the winner could lead the next card while the pile is still visibly clearing.
+  const canPlayNow = myTurnToPlay && !finishing;
+
+  const playableKeys = canPlayNow && active_stash
+    ? computePlayable(displayedHand, active_stash.cards[String(active_stash.first_player)]?.color ?? null)
     : undefined;
 
-  const activePlayerName = players.find((p) => p.order === active_player)?.name ?? '';
+  const activePlayerName = displayName(players.find((p) => p.order === active_player)?.name);
 
   // Seat mapping relative to "Ty": left / across / right.
   const seat = (offset: number): PlayerInfo | undefined =>
@@ -140,7 +206,7 @@ export default function GameTable() {
       {opponents.map((p) => (
         <div key={p.order} className="text-center">
           <div className="uppercase tracking-[.1em] text-green-dim mb-0.5" style={{ fontSize: 11 }}>
-            {p.name}
+            {displayName(p.name)}
           </div>
           <div className="font-serif text-green-score leading-none" style={{ fontSize: compact ? 16 : 20 }}>
             {computeTotal(standings, p.order)}
@@ -159,8 +225,14 @@ export default function GameTable() {
   );
 
   // Center of the oval: trick during play, guess controls during bidding.
-  const ovalContent = isPlayPhase ? (
+  // `finishing` also keeps the trick on screen while the round's *last* stash is
+  // swept away: the engine advances to the next round's bidding the instant the
+  // 4th card lands, so `isPlayPhase` flips to false immediately — without this,
+  // that final trick would vanish straight into the guess controls with no sweep.
+  const ovalContent = isPlayPhase || finishing ? (
+    <div style={collectAnim ? { animation: `${collectAnim} ${COLLECT_MS}ms ease-in both` } : undefined}>
       <Trick stash={displayedStash} players={players} myOrder={myOrder} />
+    </div>
   ) : (
     active_round_guesses !== undefined && active_player !== undefined ? (
       <GuessControls
@@ -175,14 +247,14 @@ export default function GameTable() {
 
   const topSeat = (
     <div className="flex flex-col items-center gap-1.5">
-      <PlayerCircle name={topP?.name ?? '—'} {...seatProps(topP?.order)} size={desktop ? 64 : 52} />
+      <PlayerCircle name={displayName(topP?.name) || '—'} {...seatProps(topP?.order)} size={desktop ? 64 : 52} />
       <FaceDownCards count={cardsInHandOf(topP?.order)} direction="row" desktop={desktop} />
     </div>
   );
 
   const sideSeat = (p?: PlayerInfo) => (
     <div className="flex flex-col items-center gap-1.5">
-      <PlayerCircle name={p?.name ?? '—'} {...seatProps(p?.order)} size={desktop ? 60 : 48} />
+      <PlayerCircle name={displayName(p?.name) || '—'} {...seatProps(p?.order)} size={desktop ? 60 : 48} />
       <FaceDownCards count={cardsInHandOf(p?.order)} direction="col" desktop={desktop} />
     </div>
   );
@@ -194,7 +266,7 @@ export default function GameTable() {
   );
 
   const handArea = (
-    <Hand hand={hand} myTurn={myTurnToPlay} isPlayPhase={isPlayPhase} playableKeys={playableKeys} desktop={desktop} />
+    <Hand hand={displayedHand} myTurn={canPlayNow} isPlayPhase={isPlayPhase} playableKeys={playableKeys} desktop={desktop} />
   );
 
   // ── DESKTOP LAYOUT ───────────────────────────────────────────────

frontend/src/pages/History.tsx

@@ -3,6 +3,7 @@ import { useNavigate } from 'react-router-dom';
 import { useGameStore } from '../store/gameStore';
 import { emit, socket } from '../lib/socket';
 import { useIsDesktop } from '../lib/useIsDesktop';
+import { displayName } from '../lib/names';
 import type { GameDetail, GameDetailRound } from '../types';
 
 function fmtDate(iso: string | null): string {
@@ -62,7 +63,7 @@ export default function History() {
               className="flex-1 min-w-0 text-left px-4 py-3 hover:bg-white/[.02] transition-colors"
             >
               <p className="font-serif text-green-score truncate">{g.name || 'Hra'}</p>
-              <p className="text-xs text-green-dim mt-1 truncate">{g.players.join(', ')}</p>
+              <p className="text-xs text-green-dim mt-1 truncate">{g.players.map((n) => displayName(n)).join(', ')}</p>
               <p className="text-xs text-[#7a7058] mt-0.5">
                 {fmtDate(g.created_at)} · {g.completed ? 'dohraná' : 'predčasne ukončená'}
               </p>
@@ -128,7 +129,7 @@ function GameDetailView({ detail, onBack }: { detail: GameDetail; onBack: () =>
     return r.won ? (
       <span className="font-serif" style={{ fontSize: 14, color: '#c8bb95' }}>{r.points}</span>
     ) : (
-      <span className="font-serif line-through" style={{ fontSize: 14, color: '#7a6e4a' }}>{r.guess}</span>
+      <span className="font-serif" style={{ fontSize: 14, color: '#7a6e4a' }}>{r.guess}</span>
     );
   };
   const seriesTotal = (s: number | undefined, seat: number) =>
@@ -158,7 +159,7 @@ function GameDetailView({ detail, onBack }: { detail: GameDetail; onBack: () =>
           }}
         >
           <span className="uppercase text-gold" style={{ letterSpacing: '.06em', fontSize: 11 }}>
-            {p.username}
+            {displayName(p.username)}
           </span>
           <span className="font-serif text-gold-dim" style={{ fontWeight: 700, fontSize: 16 }}>
             {totals[c]}

frontend/src/pages/Lobby.tsx

@@ -3,6 +3,7 @@ import { useNavigate, useParams } from 'react-router-dom';
 import { useGameStore } from '../store/gameStore';
 import { emit } from '../lib/socket';
 import { leaveGame } from '../lib/leaveGame';
+import { displayName } from '../lib/names';
 
 export default function Lobby() {
   const { gid } = useParams<{ gid: string }>();
@@ -65,11 +66,35 @@ export default function Lobby() {
           return (
             <div key={order} className="flex items-center gap-3">
               <span className={`text-lg ${p ? 'text-gold' : 'text-[#7a7058]'}`}>
-                {p ? '✦' : '○'}
+                {p ? (p.is_bot ? '⚙' : '✦') : '○'}
               </span>
               <span className={p ? 'font-serif text-green-score' : 'text-green-dim italic'}>
-                {p ? `${p.name}${myPlayer?.order === p.order ? ' (ty)' : ''}` : 'Čaká sa…'}
+                {p ? `${displayName(p.name)}${myPlayer?.order === p.order ? ' (ty)' : ''}` : 'Čaká sa…'}
               </span>
+              {isHost && p?.is_bot && (
+                <button
+                  onClick={() => gid && emit.removeBot(gid, order)}
+                  className="ml-auto text-xs text-green-dim hover:text-gold"
+                >
+                  Odobrať
+                </button>
+              )}
+              {isHost && !p && (
+                <span className="ml-auto flex gap-2">
+                  <button
+                    onClick={() => gid && emit.addBot(gid)}
+                    className="px-2 py-0.5 rounded-lg text-xs border border-gold/30 text-gold hover:bg-gold hover:text-table transition-colors"
+                  >
+                    + Bot
+                  </button>
+                  <button
+                    onClick={() => gid && emit.addBot(gid, 'neural')}
+                    className="px-2 py-0.5 rounded-lg text-xs border border-gold/30 text-gold hover:bg-gold hover:text-table transition-colors"
+                  >
+                    + AI bot
+                  </button>
+                </span>
+              )}
             </div>
           );
         })}

frontend/src/types.ts

@@ -11,6 +11,7 @@ export interface PlayerInfo {
   name: string;
   connected: boolean;
   player_id?: number;
+  is_bot?: boolean;
 }
 
 export interface MyPlayer {

requirements-rl.txt

@@ -0,0 +1,5 @@
+# RL trening (rl/model.py, rl/selfplay.py, rl/train.py) -- zamerne oddelene
+# od requirements.txt: server ani Docker image torch nepotrebuju, boti v hre
+# pouzivaju len cisto-Python rl/players.py (a neskor natrenovane vahy cez
+# torch az ked sa neuralny bot nasadi).
+torch>=2.4

rl/DESIGN.md

@@ -0,0 +1,181 @@
+# RL bot pre bridzik — navrh (2026-07-01)
+
+Ciel: naucit sa principy self-play reinforcement learningu (v duchu AlphaGo Zero)
+na praktickom priklade — natrenovat sietovy policy pre hranie bridziku. Zamerne
+zjednodusene oproti AlphaGo Zero: bez MCTS (skryta informacia neumoznuje priamy
+prehladavaci strom), cisty self-play policy gradient (PPO/REINFORCE) nad `bridzik.py`
+enginom.
+
+## Kluc: `Round` je nezavisla epizoda
+
+Bodovanie (`Round.get_points_summary`) je cisto lokalne pre jedno kolo — nezavisi
+od predoslych ani nasledujucich kol, len od tipu a poctu kopiek v danom kole.
+Netreba teda simulovat cely `Bridzik`/`Series` state machine na trening — staci
+instanciovat `Round(round_number, first_player, shuffler)` priamo, opakovane,
+s roznymi `round_number` (0-7, teda 8 az 1 karta v ruke). Kazdy `Round` je
+samostatna self-play epizoda.
+
+Vyhody:
+- jednoduchsi self-play loop (ziadne series/game bookkeeping)
+- vela nezavislych epizod, jednoduchá paralelizacia
+
+**Curriculum vs. uniformne samplovanie (bod 4).** Povodny napad "zacat na
+`round_number=7`" je zavadzajuci: pri 1 karte je tip masked na {0,1} a jedina
+karta je vynutena — nula card-play rozhodnuti, ziadne ucenie, len smoke-test.
+Realne ucenie je v kolach `round_number` 0-3 (6-8 kariet). Preto:
+- `round_number` je aj tak v observacii, takze **default = uniformne samplovat
+  `round_number` 0-7** a nechat siet zdielat vahy naprieč velkostami ruk.
+- Ak curriculum, tak **od tazkych (viac kariet) k lahsim**, nie naopak; alebo
+  aspon uniformne s miernym zvyhodnenim tazsich kol.
+- `round_number=7` drzat len ako sanity/smoke test pipeline, nie ako trening.
+
+## 1. State encoding (observation)
+
+Spolocne pre guess aj play fazu, budovane z `Round` objektu pre daneho hraca:
+
+- **vlastna ruka** — 32-dim multi-hot (4 farby x 8 hodnot)
+- **round_number** — one-hot (8) alebo normalizovane cislo (urcuje velkost ruk)
+- **tipy vsetkych 4 hracov** — 4x (flag "uz tipoval" + normalizovana hodnota)
+- **vlastny tip** (po tipnuti) — kriticke pre play fazu (viem, ci este potrebujem
+  vyhrat kopku, alebo sa jej mam vyhybat)
+- **kolko kopiek uz kazdy hrac vyhral v tomto kole** — 4 scalars, odvodene
+  z dokoncenych `Stash` objektov v `self.stashes`
+- **aktualna kopka v procese** — 4 sloty (karta alebo prazdne) + `first_player`
+  aktualnej kopky
+- **uz odohrane/videne karty v tomto kole** — 32-dim multi-hot (bod 1). KRITICKE:
+  bez toho observacia NIE JE Markovovska. Pri viac kartach (round_number 0-2)
+  su dve rovnake ruky s rovnakou aktualnou kopkou, ale roznou historiou uz
+  odohranych kariet, rozne stavy s roznym optimalnym tahom (vies, ci este visi
+  eso/cerven). Bez tejto zlozky sa siet nemoze naucit card-counting a strop hry
+  ostane nizky. Kodovat karty odohrane v predoslych dokoncenych `Stash`-och
+  (mimo tvojej ruky a mimo aktualnej rozohranej kopky).
+- **(volitelne, neskor — bod 8) znama neúčasť supperov vo farbe (voids)** —
+  ked supper neprizna vynasanu farbu, prezradi void → per-hrac x per-farba
+  flag. Silna informacia, ale nechat na neskorsie rozsirenie.
+
+**Egocentricka rotacia (bod 2) — povinny invariant.** Aby parameter sharing
+medzi 4 sedadlami fungoval, VSETKY 4-hracske vektory (tipy, pocty vyhranych
+kopiek, sloty aktualnej kopky, `first_player`) musia byt rotovane tak, ze
+"ja" = index 0 a ostatni relativne (+1, +2, +3 v smere hry). Toto zapisat do
+`encoding.py` ako tvrdy invariant + unit test — je to najpravdepodobnejsie
+miesto tichej chyby, ktora pokazi ucenie.
+
+Zamerne vynechane: priebezne skore/standings naprieč hrou — kedze odmena je
+per-round nezavisla, optimalne rozhodnutie v danom kole na standings nezavisi.
+
+**Velkost observacie (bod 5).** Povodny odhad ~80-100 floatov je podstrelený.
+Ak sa 4 sloty aktualnej kopky koduju one-hot (4x32=128) + ruka 32 + videne
+karty 32 + tipy/pocty/round_number/first_player, realny `input_dim` je skor
+~200. Nie je to problem, len podla toho nastavit vstupnu vrstvu siete.
+
+## 2. Akcny priestor + maskovanie
+
+- **Guess**: 9 kategorii (0-8), maskovane na `0..(8-round_number)`; pre 4.
+  (posledneho) tipujuceho naviac zamaskovat hodnotu, ktora by sposobila
+  `BridzikException` (sucet tipov = pocet kopiek) — vypocitatelne vopred
+  z `self.guesses`. Zakazana hodnota = `(8-round_number) - sum(3 tipov)`;
+  ak vyjde mimo `0..(8-round_number)`, je uz aj tak nelegalna a nemaskuje sa
+  nic navyse (osetrit rozsah).
+- **Play card**: 32 kategorii (rovnaka indexacia farba+hodnota ako hand-encoding),
+  maskovane na karty, ktore hrac realne ma A splnaju follow-suit pravidlo
+  (rovnaka logika ako v `Round.play_card`: farba prvej karty v kopke, inak
+  povinna cervena ak ju hrac ma).
+
+## 3. Sieť
+
+Zdielany "trup" (2 hidden layers, ~128-256 neuronov, ReLU) nad observation
+vektorom (~80-100 floatov), s troma vystupmi:
+
+- guess head (9 logitov)
+- play head (32 logitov)
+- value head (1 scalar) — odhad ocakavanej odmeny do konca kola (baseline
+  pre actor-critic)
+
+`input_dim` nastavit podla realnej velkosti observacie (~200, viz bod 5
+v sekcii 1), nie podla povodneho ~80-100.
+
+Fazovy flag v observacii + maskovanie urcuje, ktora hlava je pouzitelna
+v danom kroku (guess a play fazy sa nikdy neprelinaju).
+
+## 4. Odmena a trening
+
+- Odmena = 0 pocas kola; na konci kola kazdy hrac dostane
+  `points_summary[player]` (0 alebo `10+guess`) ako terminalnu odmenu za
+  VSETKY svoje rozhodnutia v danom kole (guess + vsetky `play_card` tahy).
+  Sparse terminal reward, ziadne discountovanie netreba — `gamma=1` (bod 9),
+  kolo ma max 9 rozhodnuti na hraca: round 0 = 1 tip + 8 kariet.
+- Algoritmus: self-play PPO (prip. najprv jednoduchsie REINFORCE + baseline),
+  jedna zdielana siet hra vsetkych 4 hracov v kazdom `Round` (parameter
+  sharing, rovnaky princip ako AlphaGo Zero).
+- **Normalizacia odmeny (bod 7).** `10+guess` je v rozsahu 10-18 a lisi sa
+  per kolo; pri miesanych `round_number` to zvysuje varianciu policy gradientu.
+  Standardizovat advantage per batch (odcitat priemer, delit std) — bezna
+  PPO praktika, tu je nutnejsia kvoli rozne velkym odmenam.
+- Paralelizacia: `Round` instancie su nezavisle bez shared state, self-play
+  generovanie sa da paralelizovat cez multiprocessing naprieč jadrami CPU
+  (pripadne batchovanim viacerych epizod naraz cez sietovy forward).
+
+**Caveat: hra nie je zero-sum (bod 3).** Je to 4-hracska general-sum hra —
+viacero hracov moze naraz trafit tip a vsetci skoruju, zaroven sa o kopky
+sutazi (`sum(kopky) = pocet kopiek`). Self-play so zdielanymi vahami preto
+NEMA konvergencne zaruky ako AlphaGo Zero (2-hracska zero-sum); skonverguje
+k *nejakemu* equilibriu, nie nutne k optimu, a moze oscilovat. Na ucebny
+ciel to staci, ale: (a) nepredavat si to ako "AlphaZero-grade optimalitu",
+(b) sledovat progres proti FIXNYM baseline-om (sekcia 5), nie len podla
+self-play reward, ktory sa hybe s protihracom.
+
+## 5. Vyhodnotenie
+
+- priemerne body/kolo oproti baseline (nahodny legalny hrac, Monte Carlo
+  heuristicky tipper — pozri sekciu nizsie)
+- presnost tipu (% kôl, kde sa tip presne trafil) — interpretovatelnejsia
+  metrika nez surove body
+
+## Alternativa/doplnok pre tipovaciu fazu: Monte Carlo namiesto siete
+
+Kedze tipovanie je v podstate odhad pravdepodobnosti pri neznamom rozdeleni
+zvysnych kariet, da sa riesit aj bez siete:
+
+1. **Naivna MC simulacia** — vygenerovat vela nahodnych rozdeleni zvysnych
+   kariet medzi ostatnych 3 hracov, odsimulovat kolo s jednoduchou heuristickou
+   hracou strategiou, spocitat rozdelenie poctu vlastnych kopiek. POZOR: kedze
+   bodujeme len presnu zhodu, spravny cieľ je **mod** rozdelenia, nie priemer.
+   POZOR 2 (bod 6 — "discard pile"): `deal_starting_cards` zahodí prvych
+   `4*round_number` kariet (`round_cards[4*round_number:]`), takze v kole NIE
+   su rozdane vsetky karty. MC teda z `32 - vlastna_ruka` kariet rozdá kazdemu
+   z 3 supperov len `(8-round_number)` kariet a **zvysok necha v neznamej kope
+   mimo hru** — nerozdavat vsetko medzi supperov, inak nadhodnotis, kolko
+   vysokych kariet/cervene supperi drzia.
+2. **Silnejsia verzia** — rovnaky MC rollout, ale simulovat zvysok kola
+   s uz natrenovanou card-play sietou namiesto naivnej heuristiky (analogia
+   MCTS + value network v AlphaZero namiesto ciste nahodnych rolloutov).
+
+Toto sa da pouzit ako rychly heuristicky baseline bez trenovania siete na
+tipovanie vobec, alebo ako silnejsi hybrid s uz existujucou play sietou.
+
+## Poradie implementacie
+
+1. `rl/encoding.py` — cistě funkcie observation + mask (nad `Round` objektom),
+   testovatelne izolovane. Uz tu zapracovat bod 1 (videne karty) a bod 2
+   (egocentricka rotacia). Unit/property testy: maska NIKDY nepovoli tah, ktory
+   `Round.play_card`/`add_player_guess` odmietne (fuzz-test proti enginu);
+   rotacia je konzistentna pre vsetky 4 sedadla.
+2. `rl/env.py` — step/reset wrapper okolo jedneho `Round` (nie celeho `Bridzik`)
+3. **baseline hraci + evaluacny harness UZ TU** (nahodny legalny hrac, MC
+   heuristicky tipper podla sekcie vyssie) — nech je metrika k dispozicii od
+   prvej trenovacej epochy a da sa sledovat progres (bod 3: proti fixnym
+   baseline-om, nie len self-play reward).
+4. sieť (PyTorch, trup + 3 hlavy) — `input_dim` podla realnej velkosti obs (~200)
+5. self-play generator (paralelne `Round` epizody)
+6. PPO update krok (advantage standardizovat per batch — bod 7)
+7. (neskor, volitelne) rozsirit na cely `Series`/`Bridzik` self-play, ak by
+   sa ukazalo, ze cross-round dynamika (rotacia first_player a pod.) predsa
+   len nieco mení — podla bodu 1 to nie je ocakavane
+
+## Technologie
+
+- **PyTorch** — samostatny `requirements-rl.txt`, oddeleny od zakladneho
+  `requirements.txt` projektu
+- vlastna mensia implementacia PPO (v duchu CleanRL) namiesto Stable-Baselines3/
+  RLlib — nas pripad (multi-agent self-play so zdielanymi vahami, maskovanie
+  akcii) sa bije s ich single-agent Gym abstrakciou viac, nez by pomohla

rl/encoding.py

@@ -0,0 +1,168 @@
+"""Observation a action-mask encoding nad `Round` objektom z `bridzik.py`.
+
+Ciste funkcie bez externych zavislosti (ziadny numpy/torch) -- vystupy su
+obycajne zoznamy floatov/boolov, konverzia na tenzory je vecou volajuceho.
+
+Tvrdy invariant (viz rl/DESIGN.md): VSETKY 4-hracske zlozky observacie su
+egocentricky rotovane -- "ja" (parameter `player`) je vzdy index 0, ostatni
+hraci +1/+2/+3 v smere hry. Bez tejto rotacie parameter sharing medzi
+sedadlami nefunguje.
+
+Masky zrkadlia pravidla enginu: maska NIKDY nesmie povolit akciu, ktoru by
+`Round.add_player_guess` / `Round.play_card` odmietli, a naopak. Tuto zhodu
+vynucuje fuzz-test v tests/test_encoding.py.
+"""
+
+from bridzik import Card, Card_colors, Card_values, ROUNDS_PER_SERIES
+
+# Kanonicke poradie farieb a hodnot = poradie deklaracie v enum-och.
+COLORS = list(Card_colors)   # HEARTS, LEAVES, ACORNS, BELLS
+VALUES = list(Card_values)   # C7, C8, C9, C10, LOWER, UPPER, KING, ACE
+
+N_CARDS = 32
+N_GUESS_ACTIONS = 9          # tipy 0..8
+N_PLAY_ACTIONS = N_CARDS     # jedna akcia = jedna karta
+
+_COLOR_INDEX = {color: i for i, color in enumerate(COLORS)}
+_VALUE_INDEX = {value: i for i, value in enumerate(VALUES)}
+
+# Layout observacie -- offsety su sucast verejneho kontraktu (testy aj siet
+# sa na ne odkazuju menom, nie magickym cislom).
+OFF_HAND = 0                          # 32 multi-hot: vlastna ruka
+OFF_SEEN = OFF_HAND + N_CARDS         # 32 multi-hot: karty z dokoncenych kopiek
+OFF_ROUND = OFF_SEEN + N_CARDS        # 8 one-hot: round_number
+OFF_PHASE = OFF_ROUND + ROUNDS_PER_SERIES   # 1 flag: 1.0 = tipovacia faza
+OFF_GUESSES = OFF_PHASE + 1           # 4 x (flag "uz tipoval", tip/8), rel. poradie
+OFF_TRICKS = OFF_GUESSES + 8          # 4 x (vyhrane kopky / 8), rel. poradie
+OFF_STASH = OFF_TRICKS + 4            # 4 x 32 one-hot: aktualna kopka, rel. poradie
+OFF_STASH_LEADER = OFF_STASH + 4 * N_CARDS  # 4 one-hot: rel. first_player kopky
+OFF_VOIDS = OFF_STASH_LEADER + 4      # 4 hraci x 4 farby: dedukovane voidy, rel. poradie
+OBS_DIM = OFF_VOIDS + 16              # = 233
+
+
+def card_index(card: Card) -> int:
+    """Index karty 0..31: farba (blok po 8) + hodnota."""
+    return _COLOR_INDEX[card.color] * 8 + _VALUE_INDEX[card.value]
+
+
+def index_card(index: int) -> Card:
+    """Inverzia card_index."""
+    return Card(COLORS[index // 8], VALUES[index % 8])
+
+
+def relative_seat(seat: int, player: int) -> int:
+    """Egocentricka rotacia: `player` -> 0, dalsi v smere hry -> 1, 2, 3."""
+    return (seat - player) % 4
+
+
+def deduce_voids(rnd) -> dict:
+    """Isto-dedukovane chybajuce farby hracov z priebehu kola.
+
+    Follow-suit pravidlo prezradza: kto nepriznal vynasanu farbu, uz ju nema;
+    kto pri tom nezahral ani cerven (tromf), nema ani tu. Vynasajuci hrac
+    neprezradza nic. Vracia dict seat -> set(Card_colors). Kedze karty pocas
+    kola len ubudaju, raz dedukovany void plati do konca kola.
+    """
+    voids = {seat: set() for seat in range(4)}
+    for stash in rnd.stashes:
+        first = stash.get_first_card()
+        if first is None:
+            continue
+        for seat, card in stash.get_cards().items():
+            if seat == stash.first_player:
+                continue
+            if card.color != first.color:
+                voids[seat].add(first.color)
+                if card.color != Card_colors['HEARTS']:
+                    voids[seat].add(Card_colors['HEARTS'])
+    return voids
+
+
+def encode_observation(rnd, player: int) -> list:
+    """Observacia kola z pohladu hraca `player` (OBS_DIM floatov v [0, 1]).
+
+    Funguje v oboch fazach (tipovanie aj hra) aj na terminalnom stave;
+    necita get_active_player(), takze sa da zavolat pre lubovolneho hraca
+    kedykolvek.
+    """
+    obs = [0.0] * OBS_DIM
+
+    for card in rnd.player_cards[player]:
+        obs[OFF_HAND + card_index(card)] = 1.0
+
+    # Dokoncene kopky -> "videne karty"; jedina pripadna nedokoncena kopka
+    # (posledna) je aktualna rozohrana a koduje sa do slotov nizsie.
+    current_stash = None
+    for stash in rnd.stashes:
+        if stash.is_completed():
+            for card in stash.get_cards().values():
+                obs[OFF_SEEN + card_index(card)] = 1.0
+        else:
+            current_stash = stash
+
+    obs[OFF_ROUND + rnd.round_number] = 1.0
+    obs[OFF_PHASE] = 0.0 if rnd.is_guessing_completed() else 1.0
+
+    for seat, guess in rnd.guesses.items():
+        rel = relative_seat(seat, player)
+        obs[OFF_GUESSES + 2 * rel] = 1.0
+        obs[OFF_GUESSES + 2 * rel + 1] = guess / 8
+
+    tricks = rnd.get_stashes_winner_summary()
+    for seat in range(4):
+        obs[OFF_TRICKS + relative_seat(seat, player)] = tricks[seat] / 8
+
+    if current_stash is not None:
+        for seat, card in current_stash.get_cards().items():
+            rel = relative_seat(seat, player)
+            obs[OFF_STASH + rel * N_CARDS + card_index(card)] = 1.0
+        obs[OFF_STASH_LEADER + relative_seat(current_stash.first_player, player)] = 1.0
+
+    for seat, banned in deduce_voids(rnd).items():
+        rel = relative_seat(seat, player)
+        for color in banned:
+            obs[OFF_VOIDS + rel * 4 + _COLOR_INDEX[color]] = 1.0
+
+    return obs
+
+
+def guess_mask(rnd) -> list:
+    """Maska legalnych tipov (N_GUESS_ACTIONS boolov) pre aktivneho tipujuceho.
+
+    Zrkadli Round.add_player_guess: tip 0..(8 - round_number); poslednemu
+    (stvrtemu) tipujucemu je navyse zakazana hodnota, pri ktorej by sucet
+    tipov vysiel presne na pocet kopiek v kole.
+    """
+    max_guess = 8 - rnd.round_number
+    mask = [g <= max_guess for g in range(N_GUESS_ACTIONS)]
+    if len(rnd.guesses) == 3:
+        forbidden = max_guess - sum(rnd.guesses.values())
+        if 0 <= forbidden <= max_guess:
+            mask[forbidden] = False
+    return mask
+
+
+def legal_cards(hand: list, first_card) -> list:
+    """Karty z `hand`, ktore smie hrac zahrat do kopky vynasanej `first_card`.
+
+    Zrkadli follow-suit logiku Round.play_card: povinna farba prvej karty
+    kopky; ak ju hrac nema, povinna cervena (tromf); ak nema ani tu,
+    lubovolna karta. Pri vynasani (`first_card is None`) lubovolna karta.
+    """
+    if first_card is None:
+        return list(hand)
+    same_color = [c for c in hand if c.color == first_card.color]
+    if same_color:
+        return same_color
+    hearts = [c for c in hand if c.color == Card_colors['HEARTS']]
+    return hearts if hearts else list(hand)
+
+
+def play_mask(rnd, player: int) -> list:
+    """Maska legalnych kariet (N_PLAY_ACTIONS boolov) pre hraca `player`."""
+    stash = rnd.get_last_stash()
+    first_card = stash.get_first_card() if stash is not None else None
+    mask = [False] * N_PLAY_ACTIONS
+    for card in legal_cards(rnd.player_cards[player], first_card):
+        mask[card_index(card)] = True
+    return mask

rl/env.py

@@ -0,0 +1,70 @@
+"""Self-play prostredie nad jednym `Round`-om (viz rl/DESIGN.md).
+
+Jedno kolo = jedna epizoda. Prostredie je multi-agentne a tahove: v kazdom
+kroku je na tahu prave jeden hrac (`Decision.player`), akciu zan doda
+volajuci (zdielana siet, heuristika, ...). Odmena je sparse a terminalna --
+`Round.get_points_summary()` pre vsetkych 4 hracov naraz na konci kola.
+
+Akcie: v tipovacej faze index tipu 0..8, v hracej faze index karty 0..31
+(kanonicke cislovanie z rl/encoding.py). Legalne akcie urcuje
+`Decision.mask`; nelegalna akcia prebuble ako BridzikException z enginu.
+"""
+
+from random import Random
+from typing import NamedTuple
+
+from bridzik import Round, ROUNDS_PER_SERIES
+from rl.encoding import encode_observation, guess_mask, index_card, play_mask
+
+PHASE_GUESS = 'guess'
+PHASE_PLAY = 'play'
+
+
+class Decision(NamedTuple):
+    """Jeden rozhodovaci bod: kto je na tahu, v akej faze, co vidi a co smie."""
+    player: int
+    phase: str
+    obs: list
+    mask: list
+
+
+class RoundEnv:
+    def __init__(self, rng: Random = None):
+        self.rng = rng if rng is not None else Random()
+        self.round = None
+
+    def reset(self, round_number: int = None, first_player: int = None) -> Decision:
+        """Zacne novu epizodu; nezadane parametre sa sampluju uniformne."""
+        if round_number is None:
+            round_number = self.rng.randrange(ROUNDS_PER_SERIES)
+        if first_player is None:
+            first_player = self.rng.randrange(4)
+        self.round = Round(round_number, first_player, shuffler=self.rng.shuffle)
+        return self._decision()
+
+    def step(self, action: int):
+        """Vykona akciu hraca na tahu.
+
+        Vracia (decision, rewards, done): pocas kola (Decision, None, False),
+        na konci kola (None, [body 4 hracov], True).
+        """
+        if self.round is None or self.round.is_completed():
+            raise RuntimeError('Epizoda nebezi -- najprv zavolaj reset().')
+        player = self.round.get_active_player()
+        if not self.round.is_guessing_completed():
+            self.round.add_player_guess(player, action)
+        else:
+            self.round.play_card(player, index_card(action))
+        if self.round.is_completed():
+            return None, self.round.get_points_summary(), True
+        return self._decision(), None, False
+
+    def _decision(self) -> Decision:
+        player = self.round.get_active_player()
+        if not self.round.is_guessing_completed():
+            return Decision(player, PHASE_GUESS,
+                            encode_observation(self.round, player),
+                            guess_mask(self.round))
+        return Decision(player, PHASE_PLAY,
+                        encode_observation(self.round, player),
+                        play_mask(self.round, player))

rl/evaluate.py

@@ -0,0 +1,89 @@
+"""Evaluacny harness: odohra N kol medzi 4 hracmi a spocita metriky.
+
+Metriky per hrac (viz rl/DESIGN.md, sekcia 5): priemerne body na kolo
+a presnost tipu (% kol s presne trafenym tipom). Sedadla sa medzi kolami
+rotuju, aby ziadny hrac nebol systematicky zvyhodneny poradim tipovania.
+
+Spustenie ako skript porovna baseline botov:
+    py -m rl.evaluate --rounds 500 --seed 7
+"""
+
+import argparse
+from random import Random
+
+from bridzik import ROUNDS_PER_SERIES
+from rl.env import PHASE_GUESS, RoundEnv
+
+
+def play_round(players: list, env: RoundEnv, round_number: int = None,
+               first_player: int = None) -> list:
+    """Odohra jedno kolo; `players[seat]` rozhoduje za sedadlo `seat`.
+
+    Vrati body 4 sedadiel (`Round.get_points_summary()`).
+    """
+    decision = env.reset(round_number, first_player)
+    while True:
+        seat = decision.player
+        if decision.phase == PHASE_GUESS:
+            action = players[seat].guess(env.round, seat)
+        else:
+            action = players[seat].play(env.round, seat)
+        decision, rewards, done = env.step(action)
+        if done:
+            return rewards
+
+
+def evaluate(players: list, n_rounds: int, rng: Random = None,
+             round_numbers: list = None) -> list:
+    """Odohra `n_rounds` kol s rotaciou sedadiel; vrati stats per hrac.
+
+    Vystup: zoznam dictov v poradi `players` --
+    {'avg_points': float, 'hit_rate': float, 'rounds': int}.
+    """
+    rng = rng if rng is not None else Random()
+    env = RoundEnv(rng)
+    points = [0] * 4
+    hits = [0] * 4
+    for i in range(n_rounds):
+        round_number = rng.choice(round_numbers) if round_numbers \
+            else rng.randrange(ROUNDS_PER_SERIES)
+        # rotacia: sedadlo s obsadzuje players[(s + i) % 4]
+        seating = [players[(s + i) % 4] for s in range(4)]
+        rewards = play_round(seating, env, round_number)
+        for seat in range(4):
+            player_index = (seat + i) % 4
+            points[player_index] += rewards[seat]
+            hits[player_index] += rewards[seat] > 0
+    return [{'avg_points': points[p] / n_rounds,
+             'hit_rate': hits[p] / n_rounds,
+             'rounds': n_rounds} for p in range(len(players))]
+
+
+def main():
+    from rl.players import HeuristicPlayer, RandomPlayer
+
+    parser = argparse.ArgumentParser(description='Evaluacia baseline botov')
+    parser.add_argument('--rounds', type=int, default=500)
+    parser.add_argument('--seed', type=int, default=7)
+    parser.add_argument('--mc-samples', type=int, default=100)
+    args = parser.parse_args()
+
+    rng = Random(args.seed)
+    lineups = [
+        ('4x random', [RandomPlayer(rng) for _ in range(4)]),
+        ('1x heuristika + 3x random',
+         [HeuristicPlayer(rng, n_samples=args.mc_samples)]
+         + [RandomPlayer(rng) for _ in range(3)]),
+        ('4x heuristika',
+         [HeuristicPlayer(rng, n_samples=args.mc_samples) for _ in range(4)]),
+    ]
+    for label, players in lineups:
+        stats = evaluate(players, args.rounds, rng)
+        print(f'\n{label} ({args.rounds} kol):')
+        for i, s in enumerate(stats):
+            print(f'  hrac {i}: {s["avg_points"]:6.2f} bodov/kolo, '
+                  f'tip trafeny {100 * s["hit_rate"]:5.1f} %')
+
+
+if __name__ == '__main__':
+    main()

rl/export.py

@@ -0,0 +1,69 @@
+"""Export vah natrenovanej siete do formatu pre cisto-Python inferenciu.
+
+Torch je len trenovacia zavislost (host); produkcia hra cez rl/pure_net.py,
+ktory cita tento subor bez torch/numpy. Vahy sa uladaju ako base64 float32
+little-endian (bit-exact kopia checkpointu, ziadna strata presnosti).
+
+Pouzitie:
+    py -m rl.export rl/checkpoints/latest.pt rl/weights/neural-bot.json
+"""
+
+import argparse
+import base64
+import json
+import os
+
+from rl.encoding import OBS_DIM
+from rl.train import load_checkpoint
+
+DEFAULT_WEIGHTS_PATH = os.path.join('rl', 'weights', 'neural-bot.json')
+
+
+def _pack(tensor) -> dict:
+    """Tensor -> {shape, base64 float32 LE}. Cez struct, bez numpy -- tolist()
+    vracia presne hodnoty float32, takze zapis je bit-exact."""
+    import struct
+    data = tensor.detach().to('cpu').contiguous().float()
+    flat = data.reshape(-1).tolist()
+    return {
+        'shape': list(data.shape),
+        'data': base64.b64encode(struct.pack(f'<{len(flat)}f', *flat)).decode('ascii'),
+    }
+
+
+def export(checkpoint_path: str, out_path: str) -> dict:
+    net = load_checkpoint(checkpoint_path)
+    hidden = net.trunk[0].out_features
+    payload = {
+        'obs_dim': OBS_DIM,
+        'hidden': hidden,
+        'weights': {
+            'trunk0_w': _pack(net.trunk[0].weight),
+            'trunk0_b': _pack(net.trunk[0].bias),
+            'trunk2_w': _pack(net.trunk[2].weight),
+            'trunk2_b': _pack(net.trunk[2].bias),
+            'guess_w': _pack(net.guess_head.weight),
+            'guess_b': _pack(net.guess_head.bias),
+            'play_w': _pack(net.play_head.weight),
+            'play_b': _pack(net.play_head.bias),
+        },
+    }
+    os.makedirs(os.path.dirname(out_path), exist_ok=True)
+    with open(out_path, 'w') as f:
+        json.dump(payload, f)
+    return payload
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Export vah pre pure-Python inferenciu')
+    parser.add_argument('checkpoint', nargs='?', default='rl/checkpoints/latest.pt')
+    parser.add_argument('out', nargs='?', default=DEFAULT_WEIGHTS_PATH)
+    args = parser.parse_args()
+    payload = export(args.checkpoint, args.out)
+    size = os.path.getsize(args.out)
+    print(f'Exportovane: {args.checkpoint} (hidden={payload["hidden"]}) '
+          f'-> {args.out} ({size / 1024:.0f} kB)')
+
+
+if __name__ == '__main__':
+    main()

rl/model.py

@@ -0,0 +1,49 @@
+"""Siet pre self-play PPO (viz rl/DESIGN.md, sekcia 3).
+
+Zdielany trup nad observaciou z rl/encoding.py a tri hlavy:
+guess (9 logitov), play (32 logitov), value (1 skalar -- baseline pre
+actor-critic). Ktora policy hlava plati, urcuje faza rozhodnutia; nelegalne
+akcie sa odrezavaju maskou (logit -inf), takze distribucia nikdy nenavzorkuje
+tah, ktory by engine odmietol.
+"""
+
+import torch
+import torch.nn as nn
+
+from rl.encoding import N_GUESS_ACTIONS, N_PLAY_ACTIONS, OBS_DIM
+
+
+class BridzikNet(nn.Module):
+    def __init__(self, hidden: int = 256):
+        super().__init__()
+        self.trunk = nn.Sequential(
+            nn.Linear(OBS_DIM, hidden), nn.ReLU(),
+            nn.Linear(hidden, hidden), nn.ReLU(),
+        )
+        self.guess_head = nn.Linear(hidden, N_GUESS_ACTIONS)
+        self.play_head = nn.Linear(hidden, N_PLAY_ACTIONS)
+        self.value_head = nn.Linear(hidden, 1)
+
+    def forward(self, obs: torch.Tensor):
+        """obs (B, OBS_DIM) -> (guess_logits (B,9), play_logits (B,32), value (B,))."""
+        h = self.trunk(obs)
+        return self.guess_head(h), self.play_head(h), self.value_head(h).squeeze(-1)
+
+
+def masked_categorical(logits: torch.Tensor, mask: torch.Tensor) -> torch.distributions.Categorical:
+    """Kategoricka distribucia s nelegalnymi akciami odrezanymi na -inf.
+
+    `mask` je bool tensor rovnakeho tvaru ako `logits`; kazdy riadok musi mat
+    aspon jednu povolenu akciu (garantuju masky z rl/encoding.py).
+    """
+    return torch.distributions.Categorical(
+        logits=logits.masked_fill(~mask, float('-inf'))
+    )
+
+
+def obs_tensor(obs: list) -> torch.Tensor:
+    return torch.tensor(obs, dtype=torch.float32)
+
+
+def mask_tensor(mask: list) -> torch.Tensor:
+    return torch.tensor(mask, dtype=torch.bool)

rl/players.py

@@ -0,0 +1,264 @@
+"""Baseline hraci pre evaluaciu a neskorsi warm-start siete (viz rl/DESIGN.md).
+
+Spolocne rozhranie: `guess(rnd, seat) -> int` (tip 0..8) a
+`play(rnd, seat) -> int` (index karty 0..31). Hrac vidi len to, co by videl
+pri stole -- vlastnu ruku, tipy, dokoncene kopky a rozohranu kopku; do cudzich
+ruk nesiaha.
+"""
+
+from collections import Counter
+from random import Random
+
+from bridzik import cards, Card_colors, Stash
+from rl.encoding import (
+    N_GUESS_ACTIONS, N_PLAY_ACTIONS,
+    card_index, deduce_voids, guess_mask, legal_cards, play_mask,
+)
+
+
+class RandomPlayer:
+    """Uniformne nahodny legalny tah -- najslabsi mozny baseline."""
+
+    def __init__(self, rng: Random = None):
+        self.rng = rng if rng is not None else Random()
+
+    def guess(self, rnd, seat: int) -> int:
+        mask = guess_mask(rnd)
+        return self.rng.choice([g for g in range(N_GUESS_ACTIONS) if mask[g]])
+
+    def play(self, rnd, seat: int) -> int:
+        mask = play_mask(rnd, seat)
+        return self.rng.choice([i for i in range(N_PLAY_ACTIONS) if mask[i]])
+
+
+def _strength(card) -> tuple:
+    """Absolutna sila karty: kazda cervena (tromf) bije kazdu necervenu."""
+    return (card.color == Card_colors['HEARTS'], card.value.value)
+
+
+def _current_best(stash):
+    """Zatial vitazna karta rozohranej kopky (None ak sa este nevynieslo)."""
+    first = stash.get_first_card() if stash is not None else None
+    if first is None:
+        return None
+    best = first
+    for card in stash.get_cards().values():
+        if _beats(card, best):
+            best = card
+    return best
+
+
+def _beats(card, best) -> bool:
+    """Ci `card` prebije `best` (karta drziaca kopku; jej farba je smerodajna)."""
+    if card.color == best.color:
+        return card.value > best.value
+    return card.color == Card_colors['HEARTS']
+
+
+def simulate_tricks(hands: dict, leader: int, rng: Random) -> list:
+    """Dohra kopky s nahodnou legalnou strategiou; vrati pocty vyhier hracov.
+
+    `hands` je dict seat -> zoznam kariet (rovnako velke ruky); zoznamy sa
+    spotrebuju. Vitaza kopky urcuje enginovy Stash.get_winner() -- pravidla
+    sa tu neduplikuju.
+    """
+    tricks = [0] * 4
+    for _ in range(len(hands[leader])):
+        stash = Stash(leader)
+        for i in range(4):
+            seat = (leader + i) % 4
+            card = rng.choice(legal_cards(hands[seat], stash.get_first_card()))
+            hands[seat].remove(card)
+            stash.add_card(seat, card)
+        leader = stash.get_winner()
+        tricks[leader] += 1
+    return tricks
+
+
+def deal_consistent(unknown: list, hand_sizes: dict, voids: dict,
+                    rng: Random, max_tries: int = 20) -> dict:
+    """Nahodne rozdanie neznamych kariet superom respektujuce voidy.
+
+    Greedy priradenie po zamiesani (najviac obmedzeni hraci prvi); ak sa
+    konzistentne rozdanie nepodari za `max_tries`, padne na rozdanie bez
+    voidov (zriedkave, radsej mierne skreslena vzorka nez ziadna).
+    Zvysok kariet ostava v odlozenej kope mimo hry.
+    """
+    seats = sorted(hand_sizes, key=lambda s: len(voids.get(s, ())), reverse=True)
+    pool = list(unknown)
+    for _ in range(max_tries):
+        rng.shuffle(pool)
+        remaining = list(pool)
+        hands = {}
+        for seat in seats:
+            hand, rest, banned = [], [], voids.get(seat, set())
+            for card in remaining:
+                if len(hand) < hand_sizes[seat] and card.color not in banned:
+                    hand.append(card)
+                else:
+                    rest.append(card)
+            if len(hand) < hand_sizes[seat]:
+                break
+            hands[seat] = hand
+            remaining = rest
+        else:
+            return hands
+    rng.shuffle(pool)
+    idx = 0
+    hands = {}
+    for seat in seats:
+        hands[seat] = pool[idx:idx + hand_sizes[seat]]
+        idx += hand_sizes[seat]
+    return hands
+
+
+def mc_guess_distribution(rnd, seat: int, n_samples: int, rng: Random) -> Counter:
+    """Monte Carlo odhad rozdelenia poctu vlastnych kopiek v kole.
+
+    Nezname karty sa v kazdej vzorke nahodne rozdelia ostatnym trom hracom
+    -- kazdemu len (8 - round_number) kariet, zvysok ostava v odlozenej kope
+    mimo hry (pozri DESIGN.md, pasca "discard pile"). Leader prvej kopky je
+    v case tipovania neznamy (najvyssi tip), sampluje sa uniformne.
+    """
+    hand = rnd.player_cards[seat]
+    hand_size = 8 - rnd.round_number
+    unknown = [c for c in cards if c not in hand]
+    counts = Counter()
+    for _ in range(n_samples):
+        rng.shuffle(unknown)
+        sim_hands = {seat: list(hand)}
+        others = [s for s in range(4) if s != seat]
+        for i, other in enumerate(others):
+            sim_hands[other] = unknown[i * hand_size:(i + 1) * hand_size]
+        tricks = simulate_tricks(sim_hands, rng.randrange(4), rng)
+        counts[tricks[seat]] += 1
+    return counts
+
+
+def finish_round(hands: dict, current_cards: dict, first_player: int,
+                 me: int, my_card, tricks: list, rng: Random) -> list:
+    """Dohra kolo od mojho tahu: dokonci rozohranu kopku (ja hram `my_card`,
+    dalsi nahodne legalne) a zvysne kopky dohra nahodnou legalnou strategiou.
+    `tricks` su uz vyhrane kopky (mutuje sa kopia volajuceho); vrati final."""
+    stash = Stash(first_player)
+    for seat, card in current_cards.items():
+        stash.add_card(seat, card)
+    stash.add_card(me, my_card)
+    while not stash.is_completed():
+        seat = stash.get_active_player()
+        card = rng.choice(legal_cards(hands[seat], stash.get_first_card()))
+        hands[seat].remove(card)
+        stash.add_card(seat, card)
+    leader = stash.get_winner()
+    tricks[leader] += 1
+    while hands[leader]:
+        stash = Stash(leader)
+        for i in range(4):
+            seat = (leader + i) % 4
+            card = rng.choice(legal_cards(hands[seat], stash.get_first_card()))
+            hands[seat].remove(card)
+            stash.add_card(seat, card)
+        leader = stash.get_winner()
+        tricks[leader] += 1
+    return tricks
+
+
+class HeuristicPlayer:
+    """MC tipper + jednoducha hracia heuristika riadena vlastnym tipom."""
+
+    def __init__(self, rng: Random = None, n_samples: int = 100):
+        self.rng = rng if rng is not None else Random()
+        self.n_samples = n_samples
+
+    def guess(self, rnd, seat: int) -> int:
+        counts = mc_guess_distribution(rnd, seat, self.n_samples, self.rng)
+        mask = guess_mask(rnd)
+        # najcastejsi LEGALNY pocet kopiek (mod rozdelenia, nie priemer --
+        # boduje sa len presna zhoda); pri nule vzoriek pre legalny tip
+        # rozhodne blizkost k celkovemu modu
+        mode = counts.most_common(1)[0][0]
+        legal = [g for g in range(N_GUESS_ACTIONS) if mask[g]]
+        return max(legal, key=lambda g: (counts[g], -abs(g - mode)))
+
+    def play(self, rnd, seat: int) -> int:
+        hand = rnd.player_cards[seat]
+        stash = rnd.get_last_stash()
+        allowed = legal_cards(hand, stash.get_first_card() if stash else None)
+        need = rnd.guesses[seat] - rnd.get_stashes_winner_summary()[seat]
+        best = _current_best(stash)
+
+        if best is None:
+            # vynasam: chcem kopku -> najsilnejsia karta; nechcem -> najslabsia
+            chosen = max(allowed, key=_strength) if need > 0 else min(allowed, key=_strength)
+        else:
+            winning = [c for c in allowed if _beats(c, best)]
+            if need > 0 and winning:
+                # ber kopku co najlacnejsie
+                chosen = min(winning, key=_strength)
+            elif need <= 0 and len(winning) < len(allowed):
+                # kopku nechcem: zbav sa najsilnejsej neberucej karty
+                chosen = max((c for c in allowed if not _beats(c, best)), key=_strength)
+            elif need <= 0:
+                # vsetko berie -> ber co najlacnejsie (setri silne karty netreba,
+                # ale nizka karta drzi sancu, ze ma este niekto prebije)
+                chosen = min(allowed, key=_strength)
+            else:
+                # kopku chcem, ale nic neberie -> odhod najslabsiu
+                chosen = min(allowed, key=_strength)
+        return card_index(chosen)
+
+
+class McPlayer(HeuristicPlayer):
+    """Heuristika s MC hracou fazou: kazdy kandidatsky tah sa ohodnoti
+    simulaciami zvysku kola nad rozdaniami neznamych kariet konzistentnymi
+    s dedukovanymi voidmi (`use_voids=False` = ablacia bez dedukcie).
+    Tipovanie ostava MC tipper z HeuristicPlayer (pred prvou kartou niet
+    z coho voidy dedukovat)."""
+
+    def __init__(self, rng: Random = None, n_samples: int = 100,
+                 play_samples: int = 24, use_voids: bool = True):
+        super().__init__(rng, n_samples)
+        self.play_samples = play_samples
+        self.use_voids = use_voids
+
+    def play(self, rnd, seat: int) -> int:
+        hand = rnd.player_cards[seat]
+        stash = rnd.get_last_stash()
+        first_card = stash.get_first_card() if stash else None
+        candidates = legal_cards(hand, first_card)
+        if len(candidates) == 1:
+            return card_index(candidates[0])
+
+        target = rnd.guesses[seat]
+        base_tricks = rnd.get_stashes_winner_summary()
+        voids = deduce_voids(rnd) if self.use_voids else {}
+        seen = set()
+        played_count = Counter()
+        for st in rnd.stashes:
+            for other, card in st.get_cards().items():
+                seen.add(card)
+                played_count[other] += 1
+        unknown = [c for c in cards if c not in seen and c not in hand]
+        hand_size0 = 8 - rnd.round_number
+        hand_sizes = {s: hand_size0 - played_count[s] for s in range(4) if s != seat}
+        current_cards = stash.get_cards()
+
+        # spolocne rozdanie pre vsetkych kandidatov (common random numbers --
+        # porovnavame tahy na tych istych svetoch, mensia variancia)
+        scores = {card_index(c): 0 for c in candidates}
+        for _ in range(self.play_samples):
+            world = deal_consistent(unknown, hand_sizes, voids, self.rng)
+            for candidate in candidates:
+                sim_hands = {s: list(h) for s, h in world.items()}
+                sim_hands[seat] = [c for c in hand if c != candidate]
+                tricks = finish_round(sim_hands, dict(current_cards),
+                                      stash.first_player, seat, candidate,
+                                      list(base_tricks), self.rng)
+                if tricks[seat] == target:
+                    scores[card_index(candidate)] += 1
+
+        best_index = max(scores, key=scores.get)
+        if scores[best_index] == 0:
+            # tip uz je (takmer) nedosiahnutelny -> aspon rozumny pravidlovy tah
+            return super().play(rnd, seat)
+        return best_index

rl/policy_player.py

@@ -0,0 +1,42 @@
+"""Natrenovana siet ako hrac so standardnym rozhranim guess/play.
+
+Rovnake rozhranie ako rl/players.py, takze funguje v rl/evaluate.py aj ako
+boti "mozog" v api/bots.py. Hrac vidi len observaciu + masku z rl/encoding.py
+-- z principu nemoze podvadzat (do cudzich ruk sa nedostane).
+"""
+
+import torch
+
+from rl.encoding import encode_observation, guess_mask, play_mask
+from rl.model import BridzikNet, mask_tensor, masked_categorical, obs_tensor
+
+
+class NeuralPlayer:
+    def __init__(self, net: BridzikNet, greedy: bool = True):
+        self.net = net
+        self.greedy = greedy  # argmax pri evaluacii; sampling pre pestrost
+
+    def _act(self, rnd, seat: int, use_play_head: bool) -> int:
+        obs = obs_tensor(encode_observation(rnd, seat)).unsqueeze(0)
+        mask = mask_tensor(
+            play_mask(rnd, seat) if use_play_head else guess_mask(rnd)
+        ).unsqueeze(0)
+        self.net.eval()
+        with torch.no_grad():
+            guess_logits, play_logits, _ = self.net(obs)
+            logits = play_logits if use_play_head else guess_logits
+            logits = logits.masked_fill(~mask, float('-inf'))
+            if self.greedy:
+                return int(logits.argmax(dim=-1).item())
+            return int(masked_categorical(logits, mask).sample().item())
+
+    def guess(self, rnd, seat: int) -> int:
+        return self._act(rnd, seat, use_play_head=False)
+
+    def play(self, rnd, seat: int) -> int:
+        return self._act(rnd, seat, use_play_head=True)
+
+
+def load_player(checkpoint_path: str, greedy: bool = True) -> NeuralPlayer:
+    from rl.train import load_checkpoint
+    return NeuralPlayer(load_checkpoint(checkpoint_path), greedy=greedy)

rl/pure_net.py

@@ -0,0 +1,114 @@
+"""Cisto-Python inferencia natrenovanej siete (stdlib only, bez torch/numpy).
+
+Nacita vahy z exportu rl/export.py a implementuje forward pass MLP
+(trunk 2x ReLU + guess/play hlavy). Sluzi produkcnym botom v api/bots.py --
+torch ostava len trenovacia zavislost na hoste. Presnost overuje
+tests/test_pure_net.py porovnanim s torch vystupmi na zivych observaciach.
+
+Vykon: ~250k nasobeni na tah (~desiatky ms) -- pri pauze medzi tahmi bota
+(BOT_MOVE_DELAY_SECONDS) nepostrehnutelne.
+"""
+
+import base64
+import json
+import os
+import struct
+
+from rl.encoding import (
+    N_GUESS_ACTIONS, N_PLAY_ACTIONS, OBS_DIM,
+    encode_observation, guess_mask, play_mask,
+)
+
+DEFAULT_WEIGHTS_PATH = os.path.join(
+    os.path.dirname(__file__), 'weights', 'neural-bot.json'
+)
+
+
+def _unpack(entry: dict):
+    """{shape, base64 f32 LE} -> matica (list riadkov) alebo vektor."""
+    flat = list(struct.unpack(
+        f'<{_numel(entry["shape"])}f', base64.b64decode(entry['data'])
+    ))
+    shape = entry['shape']
+    if len(shape) == 1:
+        return flat
+    rows, cols = shape
+    return [flat[r * cols:(r + 1) * cols] for r in range(rows)]
+
+
+def _numel(shape: list) -> int:
+    n = 1
+    for dim in shape:
+        n *= dim
+    return n
+
+
+def _linear(weight, bias, x):
+    """weight (out x in) @ x + bias -- radove poradie ako torch.nn.Linear."""
+    return [sum(w * v for w, v in zip(row, x)) + b
+            for row, b in zip(weight, bias)]
+
+
+def _relu(x):
+    return [v if v > 0.0 else 0.0 for v in x]
+
+
+class PureNet:
+    def __init__(self, payload: dict):
+        if payload['obs_dim'] != OBS_DIM:
+            raise ValueError(
+                f'Vahy su pre obs_dim={payload["obs_dim"]}, kod ma {OBS_DIM} '
+                '-- treba re-export z aktualneho checkpointu.'
+            )
+        w = payload['weights']
+        self.trunk0_w = _unpack(w['trunk0_w'])
+        self.trunk0_b = _unpack(w['trunk0_b'])
+        self.trunk2_w = _unpack(w['trunk2_w'])
+        self.trunk2_b = _unpack(w['trunk2_b'])
+        self.guess_w = _unpack(w['guess_w'])
+        self.guess_b = _unpack(w['guess_b'])
+        self.play_w = _unpack(w['play_w'])
+        self.play_b = _unpack(w['play_b'])
+
+    @classmethod
+    def load(cls, path: str = DEFAULT_WEIGHTS_PATH) -> 'PureNet':
+        with open(path) as f:
+            return cls(json.load(f))
+
+    def _trunk(self, obs):
+        h = _relu(_linear(self.trunk0_w, self.trunk0_b, obs))
+        return _relu(_linear(self.trunk2_w, self.trunk2_b, h))
+
+    def guess_logits(self, obs) -> list:
+        return _linear(self.guess_w, self.guess_b, self._trunk(obs))
+
+    def play_logits(self, obs) -> list:
+        return _linear(self.play_w, self.play_b, self._trunk(obs))
+
+
+def _masked_argmax(logits: list, mask: list) -> int:
+    best, best_value = None, None
+    for i, allowed in enumerate(mask):
+        if allowed and (best is None or logits[i] > best_value):
+            best, best_value = i, logits[i]
+    return best
+
+
+class PureNeuralPlayer:
+    """Greedy hrac nad PureNet -- rovnake rozhranie a rovnake vstupy
+    (observacia + maska) ako rl.policy_player.NeuralPlayer(greedy=True)."""
+
+    def __init__(self, net: PureNet):
+        self.net = net
+
+    @classmethod
+    def load(cls, path: str = DEFAULT_WEIGHTS_PATH) -> 'PureNeuralPlayer':
+        return cls(PureNet.load(path))
+
+    def guess(self, rnd, seat: int) -> int:
+        obs = encode_observation(rnd, seat)
+        return _masked_argmax(self.net.guess_logits(obs), guess_mask(rnd))
+
+    def play(self, rnd, seat: int) -> int:
+        obs = encode_observation(rnd, seat)
+        return _masked_argmax(self.net.play_logits(obs), play_mask(rnd, seat))

rl/selfplay.py

@@ -0,0 +1,131 @@
+"""Self-play generator: jedna zdielana siet hra vsetkych 4 hracov v Round
+epizodach a zbiera trajektorie pre PPO (viz rl/DESIGN.md, sekcie 4-5).
+
+Odmena je sparse a terminalna: kazde rozhodnutie hraca v kole (tip aj vsetky
+karty) dostane ako return jeho `points_summary` z konca kola, gamma = 1.
+
+Masky sa ukladaju oddelene pre obe fazy (rozne velkosti akcneho priestoru);
+`phase_play` hovori, ktora hlava/maska pre dany krok plati.
+"""
+
+from random import Random
+
+import torch
+
+from rl.encoding import N_GUESS_ACTIONS, N_PLAY_ACTIONS
+from rl.env import PHASE_PLAY, RoundEnv
+from rl.model import BridzikNet, mask_tensor, masked_categorical, obs_tensor
+from rl.players import HeuristicPlayer, RandomPlayer
+
+# Returny sa skaluju do [0, 1] (max odmena je 10+8). Bez skalovania ma value
+# loss (MSE na 0-18) radovo vacsi gradient nez policy loss a cez zdielany
+# trup policy ucenie prevalcuje.
+REWARD_SCALE = 18.0
+
+
+def _assign_seats(rng: Random, mix_random: float, mix_heuristic: float,
+                  random_player, heuristic_player) -> dict:
+    """Obsadenie sedadiel pre jednu epizodu: None = siet, inak skriptovany
+    supper. Aspon jedno sedadlo musi hrat siet (inak niet co zbierat)."""
+    seats = {}
+    for seat in range(4):
+        roll = rng.random()
+        if roll < mix_random:
+            seats[seat] = random_player
+        elif roll < mix_random + mix_heuristic:
+            seats[seat] = heuristic_player
+        else:
+            seats[seat] = None
+    if not any(p is None for p in seats.values()):
+        seats[rng.randrange(4)] = None
+    return seats
+
+
+def collect_episodes(net: BridzikNet, n_episodes: int, rng: Random,
+                     round_numbers: list = None, mix_random: float = 0.0,
+                     mix_heuristic: float = 0.0,
+                     heuristic_samples: int = 40) -> dict:
+    """Odohra `n_episodes` self-play kol a vrati batch tenzorov:
+
+    obs (N, OBS_DIM), phase_play (N,) bool, action (N,), logp (N,), value (N,),
+    ret (N,), guess_mask (N, 9), play_mask (N, 32) -- maska nepatriacej fazy je
+    pre dany krok cela False a pri update sa nepouzije.
+    Navyse 'mean_points': priemerne body na sietove sedadlo a kolo.
+
+    Opponent mixing (robustnost na nie-self-play superov): s pravdepodobnostou
+    `mix_random` / `mix_heuristic` hra sedadlo RandomPlayer / HeuristicPlayer
+    namiesto siete. Tahy skriptovanych superov sa do batchu NEZAZNAMENAVAJU
+    (nie su z trenovanej policy) -- superi len obsadzuju stol.
+    """
+    env = RoundEnv(rng)
+    random_player = RandomPlayer(rng)
+    heuristic_player = HeuristicPlayer(rng, n_samples=heuristic_samples)
+    mixing = mix_random > 0 or mix_heuristic > 0
+    obs_l, phase_l, action_l, logp_l, value_l, ret_l = [], [], [], [], [], []
+    gmask_l, pmask_l = [], []
+    total_points = 0.0
+    net_seat_rounds = 0
+
+    net.eval()
+    with torch.no_grad():
+        for _ in range(n_episodes):
+            round_number = rng.choice(round_numbers) if round_numbers else None
+            decision = env.reset(round_number)
+            seats = _assign_seats(rng, mix_random, mix_heuristic,
+                                  random_player, heuristic_player) if mixing \
+                else {seat: None for seat in range(4)}
+            net_seat_rounds += sum(1 for p in seats.values() if p is None)
+            # indexy krokov sietovych sedadiel -- na priradenie returnu
+            player_steps = {p: [] for p in range(4) if seats[p] is None}
+            while True:
+                opponent = seats[decision.player]
+                if opponent is not None:
+                    # skriptovany supper: vykonaj tah, nic nezaznamenavaj
+                    if decision.phase == PHASE_PLAY:
+                        action_i = opponent.play(env.round, decision.player)
+                    else:
+                        action_i = opponent.guess(env.round, decision.player)
+                else:
+                    obs = obs_tensor(decision.obs).unsqueeze(0)
+                    mask = mask_tensor(decision.mask).unsqueeze(0)
+                    guess_logits, play_logits, value = net(obs)
+                    is_play = decision.phase == PHASE_PLAY
+                    dist = masked_categorical(
+                        play_logits if is_play else guess_logits, mask
+                    )
+                    action = dist.sample()
+                    action_i = action.item()
+
+                    player_steps[decision.player].append(len(obs_l))
+                    obs_l.append(decision.obs)
+                    phase_l.append(is_play)
+                    action_l.append(action_i)
+                    logp_l.append(dist.log_prob(action).item())
+                    value_l.append(value.item())
+                    ret_l.append(0.0)  # doplni sa na konci kola
+                    if is_play:
+                        gmask_l.append([False] * N_GUESS_ACTIONS)
+                        pmask_l.append(decision.mask)
+                    else:
+                        gmask_l.append(decision.mask)
+                        pmask_l.append([False] * N_PLAY_ACTIONS)
+
+                decision, rewards, done = env.step(action_i)
+                if done:
+                    for player, steps in player_steps.items():
+                        for i in steps:
+                            ret_l[i] = rewards[player] / REWARD_SCALE
+                        total_points += rewards[player]
+                    break
+
+    return {
+        'obs': torch.tensor(obs_l, dtype=torch.float32),
+        'phase_play': torch.tensor(phase_l, dtype=torch.bool),
+        'action': torch.tensor(action_l, dtype=torch.long),
+        'logp': torch.tensor(logp_l, dtype=torch.float32),
+        'value': torch.tensor(value_l, dtype=torch.float32),
+        'ret': torch.tensor(ret_l, dtype=torch.float32),
+        'guess_mask': torch.tensor(gmask_l, dtype=torch.bool),
+        'play_mask': torch.tensor(pmask_l, dtype=torch.bool),
+        'mean_points': total_points / max(net_seat_rounds, 1),
+    }

rl/train.py

@@ -0,0 +1,230 @@
+"""Self-play PPO trening (viz rl/DESIGN.md, sekcie 4-6).
+
+Slucka: nazbieraj self-play epizody -> PPO update -> kazdych par iteracii
+evaluacia GREEDY policy proti fixnym baseline-om (nahodny hrac, MC heuristika)
+z rl/players.py -- self-play reward sam o sebe nie je smerodajny (hra nie je
+zero-sum, protihrac sa hybe spolu so sietou).
+
+Spustenie:
+    py -m rl.train --iterations 200 --episodes 512
+    py -m rl.train --resume rl/checkpoints/latest.pt   # pokracovanie
+
+Checkpointy: rl/checkpoints/latest.pt (kazdu iteraciu) + best.pt (najlepsi
+priemer bodov proti heuristikam). Metriky sa pripisuju do rl/runs/train_log.csv.
+"""
+
+import argparse
+import csv
+import os
+import time
+from random import Random
+
+import torch
+
+from rl.evaluate import evaluate
+from rl.model import BridzikNet, masked_categorical
+from rl.players import HeuristicPlayer, RandomPlayer
+from rl.policy_player import NeuralPlayer
+from rl.selfplay import collect_episodes
+
+CHECKPOINT_DIR = os.path.join('rl', 'checkpoints')
+RUNS_DIR = os.path.join('rl', 'runs')
+
+
+def ppo_update(net: BridzikNet, optimizer: torch.optim.Optimizer, batch: dict,
+               clip: float = 0.2, epochs: int = 4, minibatch: int = 1024,
+               vf_coef: float = 1.0, ent_coef: float = 0.01,
+               max_grad_norm: float = 0.5) -> dict:
+    """Standardny clipped-PPO krok nad batchom zo self-play.
+
+    Advantage sa standardizuje per batch (bod 7 v DESIGN.md -- odmeny 10-18 sa
+    lisia medzi kolami a zvysovali by varianciu gradientu). Guess a play kroky
+    zdielaju trup aj value hlavu, policy loss ide vzdy cez hlavu svojej fazy.
+    """
+    n = batch['obs'].shape[0]
+    adv = batch['ret'] - batch['value']
+    adv = (adv - adv.mean()) / (adv.std() + 1e-8)
+
+    net.train()
+    stats = {'policy_loss': 0.0, 'value_loss': 0.0, 'entropy': 0.0, 'updates': 0}
+    for _ in range(epochs):
+        perm = torch.randperm(n)
+        for start in range(0, n, minibatch):
+            idx = perm[start:start + minibatch]
+            obs = batch['obs'][idx]
+            guess_logits, play_logits, value = net(obs)
+
+            is_play = batch['phase_play'][idx]
+            logp_new = torch.empty_like(batch['logp'][idx])
+            entropy = torch.empty_like(logp_new)
+            for phase_sel, logits, mask_key in (
+                (~is_play, guess_logits, 'guess_mask'),
+                (is_play, play_logits, 'play_mask'),
+            ):
+                if not bool(phase_sel.any()):
+                    continue
+                dist = masked_categorical(
+                    logits[phase_sel], batch[mask_key][idx][phase_sel]
+                )
+                logp_new[phase_sel] = dist.log_prob(batch['action'][idx][phase_sel])
+                entropy[phase_sel] = dist.entropy()
+
+            ratio = torch.exp(logp_new - batch['logp'][idx])
+            mb_adv = adv[idx]
+            policy_loss = -torch.min(
+                ratio * mb_adv,
+                torch.clamp(ratio, 1 - clip, 1 + clip) * mb_adv,
+            ).mean()
+            value_loss = (value - batch['ret'][idx]).pow(2).mean()
+            loss = policy_loss + vf_coef * value_loss - ent_coef * entropy.mean()
+
+            optimizer.zero_grad()
+            loss.backward()
+            torch.nn.utils.clip_grad_norm_(net.parameters(), max_grad_norm)
+            optimizer.step()
+
+            stats['policy_loss'] += policy_loss.item()
+            stats['value_loss'] += value_loss.item()
+            stats['entropy'] += entropy.mean().item()
+            stats['updates'] += 1
+
+    for key in ('policy_loss', 'value_loss', 'entropy'):
+        stats[key] /= max(stats['updates'], 1)
+    return stats
+
+
+def evaluate_against_baselines(net: BridzikNet, n_rounds: int, rng: Random,
+                               mc_samples: int = 60) -> dict:
+    """Greedy siet na sedadle 0 vs 3x random a vs 3x heuristika."""
+    neural = NeuralPlayer(net, greedy=True)
+    vs_random = evaluate(
+        [neural] + [RandomPlayer(rng) for _ in range(3)], n_rounds, rng
+    )[0]
+    vs_heuristic = evaluate(
+        [neural] + [HeuristicPlayer(rng, n_samples=mc_samples) for _ in range(3)],
+        n_rounds, rng,
+    )[0]
+    return {
+        'vs_random_points': vs_random['avg_points'],
+        'vs_random_hit': vs_random['hit_rate'],
+        'vs_heuristic_points': vs_heuristic['avg_points'],
+        'vs_heuristic_hit': vs_heuristic['hit_rate'],
+    }
+
+
+def save_checkpoint(net: BridzikNet, hidden: int, path: str) -> None:
+    torch.save({'hidden': hidden, 'state_dict': net.state_dict()}, path)
+
+
+def load_checkpoint(path: str) -> BridzikNet:
+    """Nacita checkpoint; podporuje aj stary format (bare state_dict)."""
+    payload = torch.load(path, map_location='cpu')
+    if isinstance(payload, dict) and 'state_dict' in payload:
+        net = BridzikNet(hidden=payload['hidden'])
+        net.load_state_dict(payload['state_dict'])
+    else:
+        net = BridzikNet()
+        net.load_state_dict(payload)
+    return net
+
+
+def train(iterations: int, episodes: int, lr: float, seed: int,
+          eval_every: int, eval_rounds: int, resume: str = None,
+          hidden: int = 384, ent_coef_start: float = 0.01,
+          ent_coef_final: float = 0.001, lr_final_frac: float = 0.1,
+          mix_random: float = 0.0, mix_heuristic: float = 0.0):
+    os.makedirs(CHECKPOINT_DIR, exist_ok=True)
+    os.makedirs(RUNS_DIR, exist_ok=True)
+    log_path = os.path.join(RUNS_DIR, 'train_log.csv')
+    log_exists = os.path.exists(log_path)
+
+    torch.manual_seed(seed)
+    rng = Random(seed)
+    if resume:
+        net = load_checkpoint(resume)
+        hidden = net.trunk[0].out_features
+        print(f'Pokracujem z checkpointu {resume} (hidden={hidden})')
+    else:
+        net = BridzikNet(hidden=hidden)
+    optimizer = torch.optim.Adam(net.parameters(), lr=lr)
+
+    best_vs_heuristic = float('-inf')
+    with open(log_path, 'a', newline='') as log_file:
+        log = csv.writer(log_file)
+        if not log_exists:
+            log.writerow(['iteration', 'selfplay_points', 'policy_loss',
+                          'value_loss', 'entropy', 'vs_random_points',
+                          'vs_random_hit', 'vs_heuristic_points',
+                          'vs_heuristic_hit', 'seconds'])
+
+        for iteration in range(1, iterations + 1):
+            started = time.time()
+            # linearny decay: lr klesa k lr*lr_final_frac, entropny bonus
+            # k ent_coef_final -- policy sa ku koncu behu moze doostrit
+            frac = 1 - (iteration - 1) / max(iterations - 1, 1)
+            for group in optimizer.param_groups:
+                group['lr'] = lr * (lr_final_frac + (1 - lr_final_frac) * frac)
+            ent_coef = ent_coef_final + (ent_coef_start - ent_coef_final) * frac
+
+            batch = collect_episodes(net, episodes, rng,
+                                     mix_random=mix_random,
+                                     mix_heuristic=mix_heuristic)
+            stats = ppo_update(net, optimizer, batch, ent_coef=ent_coef)
+            save_checkpoint(net, hidden, os.path.join(CHECKPOINT_DIR, 'latest.pt'))
+
+            row = [iteration, f'{batch["mean_points"]:.3f}',
+                   f'{stats["policy_loss"]:.4f}', f'{stats["value_loss"]:.2f}',
+                   f'{stats["entropy"]:.3f}']
+            line = (f'it {iteration:4d} | self-play {batch["mean_points"]:5.2f} '
+                    f'b/kolo | pi {stats["policy_loss"]:+.4f} '
+                    f'| V {stats["value_loss"]:7.2f} | H {stats["entropy"]:.3f}')
+
+            if iteration % eval_every == 0 or iteration == iterations:
+                ev = evaluate_against_baselines(net, eval_rounds, rng)
+                row += [f'{ev["vs_random_points"]:.3f}', f'{ev["vs_random_hit"]:.3f}',
+                        f'{ev["vs_heuristic_points"]:.3f}', f'{ev["vs_heuristic_hit"]:.3f}']
+                line += (f' | vs random {ev["vs_random_points"]:5.2f} '
+                         f'({100 * ev["vs_random_hit"]:.0f} %)'
+                         f' | vs heur {ev["vs_heuristic_points"]:5.2f} '
+                         f'({100 * ev["vs_heuristic_hit"]:.0f} %)')
+                if ev['vs_heuristic_points'] > best_vs_heuristic:
+                    best_vs_heuristic = ev['vs_heuristic_points']
+                    save_checkpoint(net, hidden,
+                                    os.path.join(CHECKPOINT_DIR, 'best.pt'))
+                    line += ' *best*'
+            else:
+                row += ['', '', '', '']
+
+            row.append(f'{time.time() - started:.1f}')
+            log.writerow(row)
+            log_file.flush()
+            print(line)
+
+    return net
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Self-play PPO trening bridzik siete')
+    parser.add_argument('--iterations', type=int, default=200)
+    parser.add_argument('--episodes', type=int, default=512,
+                        help='self-play kol na iteraciu')
+    parser.add_argument('--lr', type=float, default=3e-4)
+    parser.add_argument('--seed', type=int, default=1)
+    parser.add_argument('--eval-every', type=int, default=10)
+    parser.add_argument('--eval-rounds', type=int, default=400)
+    parser.add_argument('--hidden', type=int, default=384,
+                        help='sirka skrytych vrstiev trupu')
+    parser.add_argument('--resume', type=str, default=None,
+                        help='cesta k checkpointu (.pt) na pokracovanie')
+    parser.add_argument('--mix-random', type=float, default=0.0,
+                        help='pravdepodobnost RandomPlayer sedadla v epizode')
+    parser.add_argument('--mix-heuristic', type=float, default=0.0,
+                        help='pravdepodobnost HeuristicPlayer sedadla v epizode')
+    args = parser.parse_args()
+    train(args.iterations, args.episodes, args.lr, args.seed,
+          args.eval_every, args.eval_rounds, args.resume, hidden=args.hidden,
+          mix_random=args.mix_random, mix_heuristic=args.mix_heuristic)
+
+
+if __name__ == '__main__':
+    main()

tests/test_bots.py

@@ -0,0 +1,218 @@
+"""Testy in-process botov (api/bots.py + tahova slucka v api/__init__.py).
+
+Rovnaky setup ako tests/test_history.py: docasny SQLite subor, env pred
+importom. Socket.IO emity idu do prazdnych roomov (ziadny klient), takze
+handlery a slucka sa daju volat priamo bez klienta.
+"""
+
+import asyncio
+import os
+import tempfile
+import unittest
+import uuid
+
+# Nastav DB/ENCRYPTION_KEY PRED importom db/api modulov.
+_DB_FILE = os.path.join(tempfile.gettempdir(), f"bridzik_test_{uuid.uuid4().hex}.db")
+os.environ["DATABASE_URL"] = "sqlite+aiosqlite:///" + _DB_FILE.replace("\\", "/")
+
+from cryptography.fernet import Fernet  # noqa: E402
+
+os.environ.setdefault("ENCRYPTION_KEY", Fernet.generate_key().decode())
+
+from random import Random  # noqa: E402
+
+import api  # noqa: E402
+from api import auth, bots, history  # noqa: E402
+from db.db import init_db  # noqa: E402
+from rl.players import HeuristicPlayer, RandomPlayer  # noqa: E402
+
+
+def run(coro):
+    return asyncio.run(coro)
+
+
+def _make_bot_accounts(n):
+    exclude = set()
+    accounts = []
+    for _ in range(n):
+        acc = run(bots.ensure_bot_account("random", exclude))
+        exclude.add(acc["player_id"])
+        accounts.append(acc)
+    return accounts
+
+
+def _make_game(seat_accounts, brains):
+    """Postavi zacatu in-memory hru + Game riadok v DB."""
+    gid = str(uuid.uuid4())
+    game = api.Game(gid, "test")
+    for seat, acc in enumerate(seat_accounts):
+        player = api.Player(None, acc["username"], seat, acc["player_id"])
+        if brains[seat] is not None:
+            player.is_bot = True
+            player.brain = brains[seat]
+            player.connected = True
+        else:
+            player.connected = False
+        game.players.append(player)
+    game.start()
+    api.games[gid] = game
+    run(history.record_game_started(
+        gid, "test", [acc["player_id"] for acc in seat_accounts]
+    ))
+    return game
+
+
+class BotAccountCase(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        run(init_db())
+        api.BOT_MOVE_DELAY_SECONDS = 0
+        api.TRICK_SWEEP_SECONDS = 0
+
+    def test_username_conventions(self):
+        self.assertTrue(bots.is_bot_username("bot:heuristic-1"))
+        self.assertFalse(bots.is_bot_username("alice"))
+        self.assertEqual(bots.kind_of("bot:heuristic-2"), "heuristic")
+        self.assertEqual(bots.kind_of("bot:random-1"), "random")
+        self.assertEqual(bots.kind_of("bot:neural-1"), "neural")
+        self.assertEqual(bots.kind_of("bot:nezmysel-9"), bots.DEFAULT_KIND)
+        self.assertIsInstance(bots.make_brain("bot:heuristic-1"), HeuristicPlayer)
+        self.assertIsInstance(bots.make_brain("bot:random-3"), RandomPlayer)
+
+    @unittest.skipUnless(bots.neural_available(),
+                         'chyba export vah (py -m rl.export)')
+    def test_neural_kind(self):
+        from rl.pure_net import PureNeuralPlayer
+        self.assertIn("neural", bots.available_kinds())
+        brain = bots.make_brain("bot:neural-1")
+        self.assertIsInstance(brain, PureNeuralPlayer)
+        # zdielana instancia PureNet (vahy sa nacitavaju len raz)
+        self.assertIs(brain.net, bots.make_brain("bot:neural-2").net)
+
+    def test_ensure_bot_account_reuse_and_exclude(self):
+        first = run(bots.ensure_bot_account("heuristic", set()))
+        self.assertTrue(first["username"].startswith("bot:heuristic-"))
+        # bez vylucenia sa ucet recykluje
+        again = run(bots.ensure_bot_account("heuristic", set()))
+        self.assertEqual(first["player_id"], again["player_id"])
+        # s vylucenim vznikne dalsi ucet s inym ID
+        second = run(bots.ensure_bot_account("heuristic", {first["player_id"]}))
+        self.assertNotEqual(first["player_id"], second["player_id"])
+        self.assertNotEqual(first["username"], second["username"])
+
+    def test_bot_account_cannot_be_hijacked(self):
+        acc = run(bots.ensure_bot_account("heuristic", set()))
+        # registracia mena zlyha -- ucet sa netvari ako nedokoncena registracia
+        with self.assertRaises(auth.AuthError):
+            run(auth.register_account(acc["username"]))
+        # login zlyha na kode (secret nikto nepozna), NIE na RegistrationIncomplete
+        with self.assertRaises(auth.AuthError) as ctx:
+            run(auth.login(acc["username"], "000000"))
+        self.assertNotIsInstance(ctx.exception, auth.RegistrationIncomplete)
+
+
+class BotTurnLoopCase(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        run(init_db())
+        api.BOT_MOVE_DELAY_SECONDS = 0
+        api.TRICK_SWEEP_SECONDS = 0
+
+    def setUp(self):
+        api.games.clear()
+        api.sessions.clear()
+        api.accounts.clear()
+
+    def test_four_bots_play_whole_game(self):
+        accounts = _make_bot_accounts(4)
+        brains = [RandomPlayer(Random(seat)) for seat in range(4)]
+        game = _make_game(accounts, brains)
+
+        run(api._run_bot_turns(game.gid))
+
+        self.assertTrue(game.bridzik_core.is_completed())
+        # cela hra je zapisana: 4 serie x 8 kol, ended_at nastaveny
+        detail = run(history.get_game_detail(game.gid))
+        self.assertIsNotNone(detail["ended_at"])
+        self.assertEqual(len(detail["rounds"]), history.FULL_GAME_ROUNDS * 4)
+
+    def test_bots_stop_at_human_turn(self):
+        accounts = _make_bot_accounts(3)
+        # sedadlo 0 = clovek; identitu v DB mu robi dalsi (nepouzity) boti
+        # ucet -- pre historiu je to len player_id, wrapper bez mozgu = clovek
+        human_acc = run(bots.ensure_bot_account(
+            "random", {a["player_id"] for a in accounts}
+        ))
+        game = _make_game(
+            [human_acc] + accounts,
+            [None, HeuristicPlayer(Random(1), n_samples=20),
+             RandomPlayer(Random(2)), RandomPlayer(Random(3))],
+        )
+        core = game.bridzik_core
+        rnd = core.series[-1].get_last_round()
+
+        async def scenario():
+            # na tahu je clovek (first_player serie 0 je sedadlo 0) -> boti nic
+            await api._run_bot_turns(game.gid)
+            self.assertEqual(len(rnd.guesses), 0)
+            # clovek tipne -> boti dotipuju a hraju az po dalsi tah cloveka
+            core.add_player_guess(0, 1)
+            await api._run_bot_turns(game.gid)
+
+        run(scenario())
+        self.assertTrue(rnd.is_guessing_completed())
+        self.assertEqual(rnd.get_active_player(), 0)
+
+    def test_add_and_remove_bot_handlers(self):
+        async def scenario():
+            gid = str(uuid.uuid4())
+            api.games[gid] = api.Game(gid, "lobby-test")
+            host = api.Player("sid-host", "hostiteľ", 0, 999_100)
+            api.games[gid].players.append(host)
+            api.sessions["sid-host"] = {"gid": gid, "order": 0}
+            api.sessions["sid-guest"] = {"gid": gid, "order": 1}
+
+            # nehostitel nesmie pridat bota
+            await api.add_bot("sid-guest", gid)
+            self.assertEqual(len(api.games[gid].players), 1)
+
+            # hostitel prida dvoch botov -> rozne ucty, najnizsie volne sedadla
+            await api.add_bot("sid-host", gid)
+            await api.add_bot("sid-host", gid, "random")
+            players = api.games[gid].players
+            self.assertEqual(len(players), 3)
+            bots_added = [p for p in players if p.is_bot]
+            self.assertEqual(len(bots_added), 2)
+            self.assertEqual({p.order for p in bots_added}, {1, 2})
+            self.assertNotEqual(bots_added[0].player_id, bots_added[1].player_id)
+            self.assertIsNotNone(bots_added[0].brain)
+
+            # remove_bot: odmietne cloveka, odoberie bota
+            await api.remove_bot("sid-host", gid, 0)
+            self.assertEqual(len(api.games[gid].players), 3)
+            await api.remove_bot("sid-host", gid, 1)
+            self.assertEqual(len(api.games[gid].players), 2)
+            self.assertIsNone(api.games[gid].player_by_order(1))
+
+        run(scenario())
+
+    def test_restore_marks_bots(self):
+        from bridzik import Bridzik
+        info = {
+            "gid": str(uuid.uuid4()),
+            "name": "obnova",
+            "seats": [(1, "alice"), (2, "bot:heuristic-1"),
+                      (3, "bot:random-1"), (4, "bob")],
+            "core": Bridzik(),
+        }
+        game = api._load_game_into_memory(info)
+        self.assertFalse(game.players[0].is_bot)
+        self.assertFalse(game.players[0].connected)
+        self.assertTrue(game.players[1].is_bot)
+        self.assertTrue(game.players[1].connected)
+        self.assertIsInstance(game.players[1].brain, HeuristicPlayer)
+        self.assertIsInstance(game.players[2].brain, RandomPlayer)
+
+
+if __name__ == '__main__':
+    unittest.main(verbosity=2)

tests/test_encoding.py

@@ -0,0 +1,296 @@
+import copy
+import random
+import unittest
+
+from bridzik import cards, Card, Card_colors, Card_values, BridzikException, Round
+from rl.encoding import (
+    COLORS, VALUES, N_CARDS, N_GUESS_ACTIONS, N_PLAY_ACTIONS,
+    OFF_HAND, OFF_SEEN, OFF_ROUND, OFF_PHASE, OFF_GUESSES, OFF_TRICKS,
+    OFF_STASH, OFF_STASH_LEADER, OFF_VOIDS, OBS_DIM,
+    card_index, deduce_voids, index_card, relative_seat, encode_observation,
+    guess_mask, play_mask,
+)
+
+
+class CardIndexCase(unittest.TestCase):
+    def test_roundtrip_and_uniqueness(self):
+        indexes = set()
+        for card in cards:
+            idx = card_index(card)
+            self.assertIn(idx, range(N_CARDS))
+            self.assertEqual(index_card(idx), card)
+            indexes.add(idx)
+        self.assertEqual(len(indexes), N_CARDS)
+
+    def test_layout(self):
+        # farba = blok po 8, hodnota = pozicia v bloku
+        self.assertEqual(card_index(Card(Card_colors['HEARTS'], Card_values['C7'])), 0)
+        self.assertEqual(card_index(Card(Card_colors['HEARTS'], Card_values['ACE'])), 7)
+        self.assertEqual(card_index(Card(COLORS[3], Card_values['ACE'])), 31)
+
+
+class RotationCase(unittest.TestCase):
+    def test_relative_seat(self):
+        for player in range(4):
+            self.assertEqual(relative_seat(player, player), 0)
+            # smer hry = rastuce cislo sedadla mod 4
+            self.assertEqual(relative_seat((player + 1) % 4, player), 1)
+            self.assertEqual(relative_seat((player + 3) % 4, player), 3)
+
+    def test_guesses_rotated_for_all_seats(self):
+        r = Round(0, 2)
+        guesses = {2: 5, 3: 0, 0: 1, 1: 1}
+        for seat in [2, 3, 0, 1]:
+            r.add_player_guess(seat, guesses[seat])
+        for player in range(4):
+            obs = encode_observation(r, player)
+            for seat in range(4):
+                rel = relative_seat(seat, player)
+                self.assertEqual(obs[OFF_GUESSES + 2 * rel], 1.0)
+                self.assertEqual(obs[OFF_GUESSES + 2 * rel + 1], guesses[seat] / 8)
+
+    def test_partial_guesses_flags(self):
+        r = Round(3, 1)
+        r.add_player_guess(1, 2)
+        for player in range(4):
+            obs = encode_observation(r, player)
+            rel = relative_seat(1, player)
+            self.assertEqual(obs[OFF_GUESSES + 2 * rel], 1.0)
+            self.assertEqual(obs[OFF_GUESSES + 2 * rel + 1], 2 / 8)
+            for seat in [0, 2, 3]:
+                rel = relative_seat(seat, player)
+                self.assertEqual(obs[OFF_GUESSES + 2 * rel], 0.0)
+                self.assertEqual(obs[OFF_GUESSES + 2 * rel + 1], 0.0)
+
+
+class ObservationCase(unittest.TestCase):
+    @staticmethod
+    def _deterministic_round():
+        # rovnaka konstrukcia ako v test_engine.RoundCase.test_play_card
+        shuffler = lambda l: None
+        c0 = [Card(Card_colors['BELLS'], Card_values['UPPER']),
+              Card(Card_colors['HEARTS'], Card_values['UPPER'])]
+        c1 = [Card(Card_colors['BELLS'], Card_values['C7']),
+              Card(Card_colors['HEARTS'], Card_values['C10'])]
+        c2 = [Card(Card_colors['BELLS'], Card_values['ACE']),
+              Card(Card_colors['BELLS'], Card_values['C8'])]
+        c3 = [Card(Card_colors['LEAVES'], Card_values['C7']),
+              Card(Card_colors['BELLS'], Card_values['LOWER'])]
+        c = ['dummy'] * 24 + c0 + c1 + c2 + c3
+        r = Round(6, 1, c, shuffler)
+        return r, [c0, c1, c2, c3]
+
+    def test_hand_multi_hot(self):
+        r, hands = self._deterministic_round()
+        for player in range(4):
+            obs = encode_observation(r, player)
+            hand_indexes = {card_index(c) for c in hands[player]}
+            for i in range(N_CARDS):
+                self.assertEqual(obs[OFF_HAND + i], 1.0 if i in hand_indexes else 0.0)
+
+    def test_round_number_and_phase(self):
+        r, _ = self._deterministic_round()
+        obs = encode_observation(r, 0)
+        for i in range(8):
+            self.assertEqual(obs[OFF_ROUND + i], 1.0 if i == 6 else 0.0)
+        self.assertEqual(obs[OFF_PHASE], 1.0)  # tipovacia faza
+
+        for seat, guess in [(1, 0), (2, 0), (3, 1), (0, 2)]:
+            r.add_player_guess(seat, guess)
+        obs = encode_observation(r, 0)
+        self.assertEqual(obs[OFF_PHASE], 0.0)  # hracia faza
+
+    def test_current_stash_slots_and_seen(self):
+        r, hands = self._deterministic_round()
+        for seat, guess in [(1, 0), (2, 0), (3, 1), (0, 2)]:
+            r.add_player_guess(seat, guess)
+
+        # rozohrana kopka: hraju 0 a 1
+        r.play_card(0, hands[0][0])
+        r.play_card(1, hands[1][0])
+        for player in range(4):
+            obs = encode_observation(r, player)
+            slot0 = relative_seat(0, player)
+            slot1 = relative_seat(1, player)
+            self.assertEqual(obs[OFF_STASH + slot0 * N_CARDS + card_index(hands[0][0])], 1.0)
+            self.assertEqual(obs[OFF_STASH + slot1 * N_CARDS + card_index(hands[1][0])], 1.0)
+            self.assertEqual(sum(obs[OFF_STASH:OFF_STASH + 4 * N_CARDS]), 2.0)
+            # leader kopky je hrac 0 (najvyssi tip)
+            self.assertEqual(obs[OFF_STASH_LEADER + relative_seat(0, player)], 1.0)
+            # nic este nie je "videne" -- prva kopka nie je dokoncena
+            self.assertEqual(sum(obs[OFF_SEEN:OFF_SEEN + N_CARDS]), 0.0)
+
+        # dokoncena kopka -> karty sa presunu do SEEN, sloty sa vyprazdnia
+        r.play_card(2, hands[2][0])
+        r.play_card(3, hands[3][1])
+        obs = encode_observation(r, 0)
+        played = [hands[0][0], hands[1][0], hands[2][0], hands[3][1]]
+        for card in played:
+            self.assertEqual(obs[OFF_SEEN + card_index(card)], 1.0)
+        self.assertEqual(sum(obs[OFF_SEEN:OFF_SEEN + N_CARDS]), 4.0)
+        self.assertEqual(sum(obs[OFF_STASH:OFF_STASH + 4 * N_CARDS]), 0.0)
+        # novu kopku vynasa vitaz (hrac 2, BELLS ACE)
+        self.assertEqual(obs[OFF_STASH_LEADER + relative_seat(2, 0)], 1.0)
+        # pocty vyhranych kopiek rotovane
+        for player in range(4):
+            obs = encode_observation(r, player)
+            self.assertEqual(obs[OFF_TRICKS + relative_seat(2, player)], 1 / 8)
+
+    def test_terminal_state_encodable(self):
+        r = Round(7, 0)
+        for seat, guess in [(0, 0), (1, 0), (2, 0), (3, 0)]:
+            try:
+                r.add_player_guess(seat, guess)
+            except BridzikException:
+                r.add_player_guess(seat, 1)
+        while not r.is_completed():
+            player = r.get_active_player()
+            mask = play_mask(r, player)
+            r.play_card(player, index_card(mask.index(True)))
+        obs = encode_observation(r, 0)
+        self.assertEqual(len(obs), OBS_DIM)
+        self.assertEqual(sum(obs[OFF_SEEN:OFF_SEEN + N_CARDS]), 4.0)
+
+
+class VoidsInObservationCase(unittest.TestCase):
+    def test_voids_encoded_and_rotated(self):
+        # hrac 0 vynasa zelen; 2 tromfne (void zelen), 3 hodi gulu (void
+        # zelen aj cerven) -- viz deduce_voids
+        hand0 = [Card(Card_colors['LEAVES'], Card_values['C7']),
+                 Card(Card_colors['LEAVES'], Card_values['C8'])]
+        hand1 = [Card(Card_colors['LEAVES'], Card_values['C9']),
+                 Card(Card_colors['LEAVES'], Card_values['C10'])]
+        hand2 = [Card(Card_colors['HEARTS'], Card_values['C7']),
+                 Card(Card_colors['ACORNS'], Card_values['C7'])]
+        hand3 = [Card(Card_colors['BELLS'], Card_values['C7']),
+                 Card(Card_colors['BELLS'], Card_values['C8'])]
+        rest = [c for c in cards if c not in hand0 + hand1 + hand2 + hand3]
+        deck = rest[:24] + hand0 + hand1 + hand2 + hand3
+        r = Round(6, 0, deck, shuffler=lambda l: None)
+        r.add_player_guess(0, 2)
+        r.add_player_guess(1, 0)
+        r.add_player_guess(2, 0)
+        r.add_player_guess(3, 1)
+
+        obs = encode_observation(r, 0)
+        self.assertEqual(sum(obs[OFF_VOIDS:OFF_VOIDS + 16]), 0.0)
+
+        for seat, card in [(0, hand0[0]), (1, hand1[0]),
+                           (2, hand2[0]), (3, hand3[0])]:
+            r.play_card(seat, card)
+
+        leaves_i = COLORS.index(Card_colors['LEAVES'])
+        hearts_i = COLORS.index(Card_colors['HEARTS'])
+        for player in range(4):
+            obs = encode_observation(r, player)
+            block = lambda seat: obs[OFF_VOIDS + relative_seat(seat, player) * 4:
+                                     OFF_VOIDS + relative_seat(seat, player) * 4 + 4]
+            self.assertEqual(sum(block(0)), 0.0)  # vynasajuci neprezradza nic
+            self.assertEqual(sum(block(1)), 0.0)  # priznal farbu
+            self.assertEqual(block(2)[leaves_i], 1.0)
+            self.assertEqual(sum(block(2)), 1.0)
+            self.assertEqual(block(3)[leaves_i], 1.0)
+            self.assertEqual(block(3)[hearts_i], 1.0)
+            self.assertEqual(sum(block(3)), 2.0)
+
+
+class GuessMaskCase(unittest.TestCase):
+    def test_range_by_round_number(self):
+        for round_number in range(8):
+            r = Round(round_number, 0)
+            mask = guess_mask(r)
+            for g in range(N_GUESS_ACTIONS):
+                self.assertEqual(mask[g], g <= 8 - round_number)
+
+    def test_last_guesser_forbidden_value(self):
+        r = Round(0, 0)
+        r.add_player_guess(0, 2)
+        r.add_player_guess(1, 1)
+        r.add_player_guess(2, 3)
+        mask = guess_mask(r)
+        self.assertFalse(mask[2])  # 2+1+3+2 == 8 kopiek -> zakazane
+        for g in [0, 1, 3, 4, 5, 6, 7, 8]:
+            self.assertTrue(mask[g])
+
+    def test_forbidden_value_out_of_range(self):
+        # sucet tipov > pocet kopiek -> zakazana hodnota by bola zaporna,
+        # ziadne dodatocne maskovanie
+        r = Round(0, 0)
+        r.add_player_guess(0, 8)
+        r.add_player_guess(1, 5)
+        r.add_player_guess(2, 0)
+        mask = guess_mask(r)
+        self.assertEqual(mask, [True] * 9)
+
+
+class MaskEngineConsistencyCase(unittest.TestCase):
+    """Fuzz: maska presne zrkadli engine -- povolena akcia NIKDY nezlyha,
+    zakazana akcia VZDY vyhodi BridzikException."""
+
+    def _check_guess_mask(self, rnd, player, mask):
+        for g in range(N_GUESS_ACTIONS):
+            if mask[g]:
+                copy.deepcopy(rnd).add_player_guess(player, g)
+            else:
+                with self.assertRaises(BridzikException):
+                    rnd.add_player_guess(player, g)
+
+    def _check_play_mask(self, rnd, player, mask):
+        self.assertIn(True, mask)  # aktivny hrac ma vzdy legalny tah
+        for i in range(N_PLAY_ACTIONS):
+            if mask[i]:
+                copy.deepcopy(rnd).play_card(player, index_card(i))
+            else:
+                with self.assertRaises(BridzikException):
+                    rnd.play_card(player, index_card(i))
+
+    def _check_observation(self, rnd, player):
+        obs = encode_observation(rnd, player)
+        self.assertEqual(len(obs), OBS_DIM)
+        for v in obs:
+            self.assertGreaterEqual(v, 0.0)
+            self.assertLessEqual(v, 1.0)
+        hand = {card_index(c) for c in rnd.player_cards[player]}
+        for i in range(N_CARDS):
+            self.assertEqual(obs[OFF_HAND + i], 1.0 if i in hand else 0.0)
+            if i in hand:  # ruka a videne karty su disjunktne
+                self.assertEqual(obs[OFF_SEEN + i], 0.0)
+        # zakodovany void nikdy neprotireci realnej ruke hraca
+        for seat in range(4):
+            rel = relative_seat(seat, player)
+            held = {c.color for c in rnd.player_cards[seat]}
+            for ci, color in enumerate(COLORS):
+                if obs[OFF_VOIDS + rel * 4 + ci] == 1.0:
+                    self.assertNotIn(color, held)
+
+    def _fuzz_round(self, rng, round_number, first_player):
+        rnd = Round(round_number, first_player)
+        for _ in range(4):
+            player = rnd.get_active_player()
+            self._check_observation(rnd, player)
+            mask = guess_mask(rnd)
+            self._check_guess_mask(rnd, player, mask)
+            rnd.add_player_guess(
+                player, rng.choice([g for g in range(N_GUESS_ACTIONS) if mask[g]])
+            )
+        while not rnd.is_completed():
+            player = rnd.get_active_player()
+            self._check_observation(rnd, player)
+            mask = play_mask(rnd, player)
+            self._check_play_mask(rnd, player, mask)
+            rnd.play_card(
+                player, index_card(rng.choice([i for i in range(N_PLAY_ACTIONS) if mask[i]]))
+            )
+        # kolo dohrane do konca cisto cez masky -> bodovanie funguje
+        self.assertEqual(len(rnd.get_points_summary()), 4)
+
+    def test_fuzz_all_round_numbers_and_seats(self):
+        rng = random.Random(1337)
+        for round_number in range(8):
+            for first_player in range(4):
+                for _ in range(3):
+                    self._fuzz_round(rng, round_number, first_player)
+
+
+if __name__ == '__main__':
+    unittest.main(verbosity=2)

tests/test_engine.py

@@ -121,6 +121,16 @@ class StashCase(unittest.TestCase):
         self.assertRaises(BridzikException, s.get_active_player)
 
 
+class CardCase(unittest.TestCase):
+    def test_hashable_consistent_with_eq(self):
+        # vlastne __eq__ nesmie zrusit hashovatelnost (dict/set pouzitie)
+        heart_7 = Card(Card_colors['HEARTS'], Card_values['C7'])
+        self.assertEqual(hash(heart_7), hash(Card(Card_colors['HEARTS'], Card_values['C7'])))
+        self.assertEqual(len(set(cards)), 32)
+        self.assertEqual({Card_colors['HEARTS']: 1}[Card_colors['HEARTS']], 1)
+        self.assertEqual({Card_values['ACE']: 1}[Card_values['ACE']], 1)
+
+
 class RoundCase(unittest.TestCase):
     def test_round_constructor(self):
         self.assertRaises(BridzikException, Round, round_number=8, first_player=0)

tests/test_pure_net.py

@@ -0,0 +1,155 @@
+"""Testy presnosti cisto-Python inferencie (rl/pure_net.py) voci torch.
+
+Jadro suity: na zivych observaciach z nahodne rozohranych kol sa porovnavaju
+logity a zvolene akcie pure-Python siete s torch sietou nacitanou z toho
+isteho checkpointu. Case bez torch (cisty beh, legalnost, determinizmus)
+bezia vzdy; porovnavacie case sa preskocia, ak torch nie je nainstalovany.
+"""
+
+import copy
+import os
+import unittest
+from random import Random
+
+from bridzik import Round
+from rl.encoding import encode_observation, guess_mask, index_card, play_mask
+from rl.env import PHASE_GUESS, RoundEnv
+from rl.evaluate import play_round
+from rl.players import RandomPlayer
+from rl.pure_net import DEFAULT_WEIGHTS_PATH, PureNet, PureNeuralPlayer
+
+WEIGHTS_AVAILABLE = os.path.exists(DEFAULT_WEIGHTS_PATH)
+
+try:
+    import torch
+    from rl.policy_player import NeuralPlayer
+    from rl.train import load_checkpoint
+    TORCH_AVAILABLE = True
+except ImportError:  # pragma: no cover
+    TORCH_AVAILABLE = False
+
+CHECKPOINT = os.path.join('rl', 'checkpoints', 'latest.pt')
+
+
+def _random_decision_points(rng, n_rounds=12):
+    """Vygeneruje zive rozhodovacie body (rnd, seat, faza) nahodnou hrou."""
+    env = RoundEnv(rng)
+    points = []
+    for i in range(n_rounds):
+        decision = env.reset(round_number=i % 8)
+        while True:
+            # snapshot -- env.round sa dalsou hrou mutuje
+            points.append((copy.deepcopy(env.round), decision.player, decision.phase))
+            action = rng.choice([a for a, ok in enumerate(decision.mask) if ok])
+            decision, rewards, done = env.step(action)
+            if done:
+                break
+    return points
+
+
+@unittest.skipUnless(WEIGHTS_AVAILABLE, 'chyba export vah (py -m rl.export)')
+class PureOnlyCase(unittest.TestCase):
+    """Bezi aj bez torch -- presne to, co pobezi v produkcii."""
+
+    @classmethod
+    def setUpClass(cls):
+        cls.player = PureNeuralPlayer.load()
+
+    def test_plays_legal_full_rounds(self):
+        env = RoundEnv(Random(1))
+        players = [self.player, self.player,
+                   RandomPlayer(Random(2)), RandomPlayer(Random(3))]
+        for round_number in range(8):
+            rewards = play_round(players, env, round_number)
+            self.assertEqual(len(rewards), 4)
+
+    def test_deterministic(self):
+        r = Round(2, 0)
+        self.assertEqual(self.player.guess(r, 0), self.player.guess(r, 0))
+
+    def test_respects_masks(self):
+        rng = Random(4)
+        for rnd, seat, phase in _random_decision_points(rng, n_rounds=8):
+            if phase == PHASE_GUESS:
+                self.assertTrue(guess_mask(rnd)[self.player.guess(rnd, seat)])
+            else:
+                self.assertTrue(play_mask(rnd, seat)[self.player.play(rnd, seat)])
+
+
+@unittest.skipUnless(WEIGHTS_AVAILABLE and TORCH_AVAILABLE
+                     and os.path.exists(CHECKPOINT),
+                     'treba torch + checkpoint + export vah')
+class TorchParityCase(unittest.TestCase):
+    """Zhoda pure-Python inferencie s torch na tom istom checkpointe."""
+
+    @classmethod
+    def setUpClass(cls):
+        cls.pure = PureNet.load()
+        cls.torch_net = load_checkpoint(CHECKPOINT)
+        cls.torch_net.eval()
+        cls.points = _random_decision_points(Random(7), n_rounds=16)
+
+    def _torch_logits(self, obs, phase):
+        with torch.no_grad():
+            guess_logits, play_logits, _ = self.torch_net(
+                torch.tensor(obs, dtype=torch.float32).unsqueeze(0)
+            )
+        t = guess_logits if phase == PHASE_GUESS else play_logits
+        return t.squeeze(0).tolist()
+
+    def test_logits_match(self):
+        """Logity sa zhoduju na ~1e-4 (rozdiel = len poradie scitovania
+        float32 vs float64, ziadna strata z exportu -- vahy su bit-exact)."""
+        worst = 0.0
+        for rnd, seat, phase in self.points:
+            obs = encode_observation(rnd, seat)
+            pure = self.pure.guess_logits(obs) if phase == PHASE_GUESS \
+                else self.pure.play_logits(obs)
+            ref = self._torch_logits(obs, phase)
+            for a, b in zip(pure, ref):
+                worst = max(worst, abs(a - b))
+        self.assertLess(worst, 1e-3, f'najvacsi rozdiel logitov: {worst}')
+
+    def test_actions_match(self):
+        """Zvolena akcia je identicka vzdy, ked nejde o numericku remizu
+        (top-2 logity blizsie nez 1e-3 -- prakticky nenastava)."""
+        player = PureNeuralPlayer(self.pure)
+        torch_player = NeuralPlayer(self.torch_net, greedy=True)
+        compared = ties = 0
+        for rnd, seat, phase in self.points:
+            obs = encode_observation(rnd, seat)
+            if phase == PHASE_GUESS:
+                a, b = player.guess(rnd, seat), torch_player.guess(rnd, seat)
+                mask = guess_mask(rnd)
+                logits = self.pure.guess_logits(obs)
+            else:
+                a, b = player.play(rnd, seat), torch_player.play(rnd, seat)
+                mask = play_mask(rnd, seat)
+                logits = self.pure.play_logits(obs)
+            allowed = sorted((logits[i] for i in range(len(mask)) if mask[i]),
+                             reverse=True)
+            if len(allowed) > 1 and allowed[0] - allowed[1] < 1e-3:
+                ties += 1  # numericka remiza -- volba je legitimne lubovolna
+                continue
+            compared += 1
+            self.assertEqual(a, b, f'akcie sa lisia mimo remizy ({phase})')
+        self.assertGreater(compared, 50)  # test realne porovnaval
+
+    def test_full_rounds_identical_trajectories(self):
+        """Dve identicke partie: pure aj torch hrac na vsetkych 4 sedadlach
+        s rovnakym rozdanim musia zahrat uplne rovnake kolo."""
+        pure_player = PureNeuralPlayer(self.pure)
+        torch_player = NeuralPlayer(self.torch_net, greedy=True)
+        for round_number in range(8):
+            results = []
+            for player in (pure_player, torch_player):
+                env = RoundEnv(Random(100 + round_number))
+                rewards = play_round([player] * 4, env, round_number)
+                results.append((rewards,
+                                sorted(str(s.get_cards())
+                                       for s in env.round.stashes)))
+            self.assertEqual(results[0], results[1])
+
+
+if __name__ == '__main__':
+    unittest.main(verbosity=2)

tests/test_rl_players.py

@@ -0,0 +1,316 @@
+import unittest
+from random import Random
+
+from bridzik import cards, Card, Card_colors, Card_values, Round
+from rl.encoding import card_index, index_card, legal_cards
+from rl.env import Decision, PHASE_GUESS, PHASE_PLAY, RoundEnv
+from rl.evaluate import evaluate, play_round
+from rl.players import (
+    HeuristicPlayer, McPlayer, RandomPlayer,
+    _beats, _current_best, deal_consistent, deduce_voids,
+    mc_guess_distribution, simulate_tricks,
+)
+
+
+class RoundEnvCase(unittest.TestCase):
+    def test_episode_structure(self):
+        env = RoundEnv(Random(42))
+        decision = env.reset(round_number=6, first_player=1)
+        rng = Random(0)
+
+        # prve 4 rozhodnutia su tipy, v poradi od first_player
+        expected_guessers = [1, 2, 3, 0]
+        for expected in expected_guessers:
+            self.assertIsInstance(decision, Decision)
+            self.assertEqual(decision.phase, PHASE_GUESS)
+            self.assertEqual(decision.player, expected)
+            action = rng.choice([g for g in range(9) if decision.mask[g]])
+            decision, rewards, done = env.step(action)
+            self.assertIsNone(rewards)
+            self.assertFalse(done)
+
+        # potom hracie rozhodnutia az po terminal: 2 karty x 4 hraci
+        steps = 0
+        while True:
+            self.assertEqual(decision.phase, PHASE_PLAY)
+            self.assertEqual(decision.player, env.round.get_active_player())
+            action = rng.choice([i for i in range(32) if decision.mask[i]])
+            decision, rewards, done = env.step(action)
+            steps += 1
+            if done:
+                break
+        self.assertEqual(steps, 8)
+        self.assertIsNone(decision)
+        self.assertEqual(rewards, env.round.get_points_summary())
+        self.assertEqual(len(rewards), 4)
+
+        # po done sa step neda volat, reset zacne novu epizodu
+        self.assertRaises(RuntimeError, env.step, 0)
+        self.assertIsInstance(env.reset(), Decision)
+
+    def test_reset_samples_round_and_seat(self):
+        env = RoundEnv(Random(7))
+        seen_rounds, seen_seats = set(), set()
+        for _ in range(100):
+            env.reset()
+            seen_rounds.add(env.round.round_number)
+            seen_seats.add(env.round.first_player)
+        self.assertEqual(seen_rounds, set(range(8)))
+        self.assertEqual(seen_seats, set(range(4)))
+
+    def test_deterministic_with_seed(self):
+        rewards = []
+        for _ in range(2):
+            env = RoundEnv(Random(123))
+            players = [RandomPlayer(Random(5)) for _ in range(4)]
+            rewards.append(play_round(players, env, round_number=0))
+        self.assertEqual(rewards[0], rewards[1])
+
+
+class SimulationHelpersCase(unittest.TestCase):
+    def test_beats(self):
+        heart_7 = Card(Card_colors['HEARTS'], Card_values['C7'])
+        heart_8 = Card(Card_colors['HEARTS'], Card_values['C8'])
+        leaves_ace = Card(Card_colors['LEAVES'], Card_values['ACE'])
+        leaves_king = Card(Card_colors['LEAVES'], Card_values['KING'])
+        bells_ace = Card(Card_colors['BELLS'], Card_values['ACE'])
+
+        self.assertTrue(_beats(leaves_ace, leaves_king))   # vyssia vo farbe
+        self.assertFalse(_beats(leaves_king, leaves_ace))
+        self.assertTrue(_beats(heart_7, leaves_ace))       # tromf bije farbu
+        self.assertFalse(_beats(bells_ace, leaves_king))   # cudzia farba neberie
+        self.assertTrue(_beats(heart_8, heart_7))          # tromfy medzi sebou
+        self.assertFalse(_beats(leaves_ace, heart_7))      # farba nebije tromf
+
+    def test_current_best_tracks_stash(self):
+        from bridzik import Stash
+        leaves_7 = Card(Card_colors['LEAVES'], Card_values['C7'])
+        leaves_ace = Card(Card_colors['LEAVES'], Card_values['ACE'])
+        heart_7 = Card(Card_colors['HEARTS'], Card_values['C7'])
+
+        self.assertIsNone(_current_best(None))
+        s = Stash(0)
+        self.assertIsNone(_current_best(s))
+        s.add_card(0, leaves_7)
+        self.assertEqual(_current_best(s), leaves_7)
+        s.add_card(1, leaves_ace)
+        self.assertEqual(_current_best(s), leaves_ace)
+        s.add_card(2, heart_7)
+        self.assertEqual(_current_best(s), heart_7)
+
+    def test_simulate_tricks_consumes_hands(self):
+        rng = Random(3)
+        deck = list(cards)
+        rng.shuffle(deck)
+        hands = {seat: deck[seat * 8:(seat + 1) * 8] for seat in range(4)}
+        tricks = simulate_tricks(hands, leader=2, rng=rng)
+        self.assertEqual(sum(tricks), 8)
+        for seat in range(4):
+            self.assertEqual(hands[seat], [])
+
+
+class HeuristicPlayerCase(unittest.TestCase):
+    @staticmethod
+    def _round_with_hand(player0_hand):
+        # deterministicke rozdanie: player0_hand ide hracovi 0, zvysok dalej;
+        # deal_starting_cards najprv zahodi 4*round_number kariet, preto
+        # treba ruku umiestnit az ZA odkladaciu kopu
+        round_number = 8 - len(player0_hand)
+        rest = [c for c in cards if c not in player0_hand]
+        skip = 4 * round_number
+        deck = rest[:skip] + list(player0_hand) + rest[skip:]
+        return Round(round_number, 0, deck, shuffler=lambda l: None)
+
+    def test_mc_guess_all_hearts_is_certain(self):
+        # 8 cerveni = tromfy beru kazdu kopku bez ohladu na rozdanie a hru
+        all_hearts = [Card(Card_colors['HEARTS'], v) for v in Card_values]
+        r = self._round_with_hand(all_hearts)
+        counts = mc_guess_distribution(r, 0, n_samples=30, rng=Random(1))
+        self.assertEqual(counts, {8: 30})
+        self.assertEqual(HeuristicPlayer(Random(1), n_samples=30).guess(r, 0), 8)
+
+    def test_mc_guess_weak_hand_low(self):
+        # dve najnizsie necervene karty -> tip 0 s prehladom
+        weak = [Card(Card_colors['LEAVES'], Card_values['C7']),
+                Card(Card_colors['BELLS'], Card_values['C7'])]
+        r = self._round_with_hand(weak)
+        self.assertEqual(HeuristicPlayer(Random(2), n_samples=60).guess(r, 0), 0)
+
+    def test_mc_guess_respects_mask(self):
+        # posledny tipujuci: zakazana hodnota nesmie byt vratena, ani ked
+        # je modom rozdelenia
+        all_hearts = [Card(Card_colors['HEARTS'], v) for v in Card_values]
+        r = self._round_with_hand(all_hearts)
+        r.add_player_guess(0, 0)
+        r.add_player_guess(1, 0)
+        r.add_player_guess(2, 0)
+        # zakazany tip pre hraca 3 je 8; jeho ruka je nahodna, ale nech by
+        # simulacia vratila cokolvek, vysledok musi byt legalny
+        guess = HeuristicPlayer(Random(3), n_samples=20).guess(r, 3)
+        self.assertNotEqual(guess, 8)
+        self.assertIn(guess, range(8))
+
+    def test_play_takes_trick_when_needed(self):
+        hand = [Card(Card_colors['LEAVES'], Card_values['ACE']),
+                Card(Card_colors['LEAVES'], Card_values['C7']),
+                Card(Card_colors['BELLS'], Card_values['C7'])]
+        r = self._round_with_hand(hand)
+        r.add_player_guess(0, 3)   # najvyssi tip -> hrac 0 vynasa
+        r.add_player_guess(1, 0)
+        r.add_player_guess(2, 0)
+        r.add_player_guess(3, 1)
+        # hrac 0 potrebuje kopky -> vynasa najsilnejsiu kartu (LEAVES ACE)
+        action = HeuristicPlayer(Random(4)).play(r, 0)
+        self.assertEqual(index_card(action), hand[0])
+
+    def test_play_ducks_when_satisfied(self):
+        hand = [Card(Card_colors['LEAVES'], Card_values['ACE']),
+                Card(Card_colors['LEAVES'], Card_values['C7']),
+                Card(Card_colors['BELLS'], Card_values['C7'])]
+        r = self._round_with_hand(hand)
+        r.add_player_guess(0, 0)   # hrac 0 nechce ziadnu kopku
+        r.add_player_guess(1, 2)   # najvyssi tip -> vynasa hrac 1
+        r.add_player_guess(2, 0)
+        r.add_player_guess(3, 0)
+        first_card = legal_cards(r.player_cards[1], None)[0]
+        r.play_card(1, first_card)
+        action = HeuristicPlayer(Random(5)).play(r, 2)
+        # legalnost staci overit enginom; strategiu netestujeme natvrdo,
+        # lebo zavisi od nahodnej ruky hraca 2
+        r.play_card(2, index_card(action))
+
+    def test_play_duck_scenario_deterministic(self):
+        # hrac 0 tipol 0, ma na ruke LEAVES ACE aj C7; kopku vedie LEAVES C8
+        # -> musi priznat farbu a spravne je podliezt (C7), nie zobrat esom
+        hand0 = [Card(Card_colors['LEAVES'], Card_values['ACE']),
+                 Card(Card_colors['LEAVES'], Card_values['C7'])]
+        hand1 = [Card(Card_colors['LEAVES'], Card_values['C8']),
+                 Card(Card_colors['LEAVES'], Card_values['C9'])]
+        rest = [c for c in cards if c not in hand0 + hand1]
+        deck = rest[:24] + hand0 + hand1 + rest[24:]   # 24 = odkladacia kopa
+        r = Round(6, 0, deck, shuffler=lambda l: None)
+        r.add_player_guess(0, 0)
+        r.add_player_guess(1, 2)   # vynasa hrac 1
+        r.add_player_guess(2, 0)
+        r.add_player_guess(3, 1)   # 0+2+0+0 by bol zakazany sucet (2 kopky)
+        r.play_card(1, hand1[0])
+        action = HeuristicPlayer(Random(6)).play(r, 0)
+        self.assertEqual(index_card(action), hand0[1])
+
+
+class VoidDeductionCase(unittest.TestCase):
+    def test_deduce_voids_from_stash(self):
+        # hrac 0 vynasa zelen; 1 prizna farbu (nic), 2 tromfne cervenou
+        # (void zelen), 3 hodi gulu (void zelen AJ cerven)
+        hand0 = [Card(Card_colors['LEAVES'], Card_values['C7']),
+                 Card(Card_colors['LEAVES'], Card_values['C8'])]
+        hand1 = [Card(Card_colors['LEAVES'], Card_values['C9']),
+                 Card(Card_colors['LEAVES'], Card_values['C10'])]
+        hand2 = [Card(Card_colors['HEARTS'], Card_values['C7']),
+                 Card(Card_colors['ACORNS'], Card_values['C7'])]
+        hand3 = [Card(Card_colors['BELLS'], Card_values['C7']),
+                 Card(Card_colors['BELLS'], Card_values['C8'])]
+        rest = [c for c in cards if c not in hand0 + hand1 + hand2 + hand3]
+        deck = rest[:24] + hand0 + hand1 + hand2 + hand3
+        r = Round(6, 0, deck, shuffler=lambda l: None)
+        r.add_player_guess(0, 2)   # najvyssi tip -> vynasa 0
+        r.add_player_guess(1, 0)
+        r.add_player_guess(2, 0)
+        r.add_player_guess(3, 1)
+
+        self.assertEqual(deduce_voids(r), {0: set(), 1: set(), 2: set(), 3: set()})
+        r.play_card(0, hand0[0])
+        r.play_card(1, hand1[0])   # priznal farbu -> nic
+        r.play_card(2, hand2[0])   # cerven -> void zelen
+        r.play_card(3, hand3[0])   # gula -> void zelen aj cerven
+        voids = deduce_voids(r)
+        self.assertEqual(voids[0], set())    # vynasajuci neprezradza nic
+        self.assertEqual(voids[1], set())
+        self.assertEqual(voids[2], {Card_colors['LEAVES']})
+        self.assertEqual(voids[3], {Card_colors['LEAVES'], Card_colors['HEARTS']})
+
+    def test_deduced_voids_never_contradict_hands(self):
+        # fuzz: dedukovany void NIKDY neprotireci realnej ruke hraca
+        rng = Random(21)
+        for _ in range(30):
+            r = Round(rng.randrange(4), rng.randrange(4))
+            players = [RandomPlayer(Random(rng.random())) for _ in range(4)]
+            for _ in range(4):
+                seat = r.get_active_player()
+                r.add_player_guess(seat, players[seat].guess(r, seat))
+            while not r.is_completed():
+                seat = r.get_active_player()
+                r.play_card(seat, index_card(players[seat].play(r, seat)))
+                for other, banned in deduce_voids(r).items():
+                    held = {c.color for c in r.player_cards[other]}
+                    self.assertFalse(held & banned,
+                                     f'void {banned} vs ruka {held}')
+
+    def test_deal_consistent_respects_voids(self):
+        rng = Random(22)
+        unknown = [c for c in cards][:20]
+        voids = {1: {Card_colors['HEARTS']}, 2: set(),
+                 3: {Card_colors['LEAVES'], Card_colors['BELLS']}}
+        for _ in range(20):
+            hands = deal_consistent(unknown, {1: 4, 2: 4, 3: 4}, voids, rng)
+            self.assertTrue(all(len(h) == 4 for h in hands.values()))
+            for seat, banned in voids.items():
+                self.assertFalse({c.color for c in hands[seat]} & banned)
+
+
+class McPlayerCase(unittest.TestCase):
+    def test_plays_legal_full_rounds(self):
+        rng = Random(23)
+        env = RoundEnv(rng)
+        players = [McPlayer(Random(24), n_samples=20, play_samples=8),
+                   McPlayer(Random(25), n_samples=20, play_samples=8,
+                            use_voids=False)] \
+            + [RandomPlayer(Random(s)) for s in (26, 27)]
+        for round_number in range(8):
+            rewards = play_round(players, env, round_number)
+            self.assertEqual(len(rewards), 4)
+
+    def test_duck_scenario(self):
+        # tip 0, kopku vedie sused LEAVES C8 a ja som HNED na tahu (MC hrac
+        # stavia kopku poctivo, takze na rozdiel od pravidlovej heuristiky
+        # vyzaduje konzistentne poradie): mam ACE aj C7 -> podlezt sedmickou
+        hand0 = [Card(Card_colors['LEAVES'], Card_values['ACE']),
+                 Card(Card_colors['LEAVES'], Card_values['C7'])]
+        hand3 = [Card(Card_colors['LEAVES'], Card_values['C8']),
+                 Card(Card_colors['LEAVES'], Card_values['C9'])]
+        rest = [c for c in cards if c not in hand0 + hand3]
+        deck = rest[:24] + hand0 + rest[24:26] + rest[26:28] + hand3
+        r = Round(6, 0, deck, shuffler=lambda l: None)
+        r.add_player_guess(0, 0)
+        r.add_player_guess(1, 0)
+        r.add_player_guess(2, 0)
+        r.add_player_guess(3, 1)   # najvyssi tip -> vynasa hrac 3, po nom ja
+        r.play_card(3, hand3[0])
+        self.assertEqual(r.get_active_player(), 0)
+        action = McPlayer(Random(28), play_samples=30).play(r, 0)
+        self.assertEqual(index_card(action), hand0[1])
+
+
+class EvaluateCase(unittest.TestCase):
+    def test_full_random_matchup_runs(self):
+        rng = Random(11)
+        stats = evaluate([RandomPlayer(rng) for _ in range(4)], 40, rng)
+        for s in stats:
+            self.assertEqual(s['rounds'], 40)
+            self.assertGreaterEqual(s['avg_points'], 0)
+            self.assertLessEqual(s['hit_rate'], 1)
+
+    def test_heuristic_beats_random(self):
+        rng = Random(13)
+        players = [HeuristicPlayer(rng, n_samples=40)] \
+            + [RandomPlayer(rng) for _ in range(3)]
+        stats = evaluate(players, 120, rng)
+        heuristic, randoms = stats[0], stats[1:]
+        best_random = max(s['avg_points'] for s in randoms)
+        self.assertGreater(heuristic['avg_points'], best_random)
+        self.assertGreater(heuristic['hit_rate'],
+                           max(s['hit_rate'] for s in randoms))
+
+
+if __name__ == '__main__':
+    unittest.main(verbosity=2)

tests/test_rl_train.py

@@ -0,0 +1,200 @@
+"""Testy PPO pipeline (rl/model, rl/selfplay, rl/policy_player, rl/train).
+
+Vyzaduju torch (requirements-rl.txt); bez neho sa cely modul preskoci --
+ostatne suity (engine, encoding, boti) na torchi nezavisia.
+"""
+
+import unittest
+from random import Random
+
+try:
+    import torch
+except ImportError:  # pragma: no cover
+    raise unittest.SkipTest('torch nie je nainstalovany (requirements-rl.txt)')
+
+from bridzik import Round
+from rl.encoding import (
+    N_GUESS_ACTIONS, N_PLAY_ACTIONS, OBS_DIM,
+    encode_observation, guess_mask, play_mask,
+)
+from rl.evaluate import evaluate, play_round
+from rl.env import RoundEnv
+from rl.model import BridzikNet, mask_tensor, masked_categorical, obs_tensor
+from rl.players import RandomPlayer
+from rl.policy_player import NeuralPlayer
+from rl.selfplay import _assign_seats, collect_episodes
+from rl.train import ppo_update
+
+
+class ModelCase(unittest.TestCase):
+    def test_output_shapes(self):
+        net = BridzikNet(hidden=32)
+        obs = torch.zeros((5, OBS_DIM))
+        guess_logits, play_logits, value = net(obs)
+        self.assertEqual(guess_logits.shape, (5, N_GUESS_ACTIONS))
+        self.assertEqual(play_logits.shape, (5, N_PLAY_ACTIONS))
+        self.assertEqual(value.shape, (5,))
+
+    def test_masked_categorical_never_samples_illegal(self):
+        torch.manual_seed(0)
+        logits = torch.zeros((1, 9))
+        mask = torch.tensor([[False, True, False, True, False,
+                              False, False, False, False]])
+        dist = masked_categorical(logits, mask)
+        samples = dist.sample((200,))
+        self.assertTrue(set(samples.flatten().tolist()) <= {1, 3})
+        # entropia a log_prob su konecne aj s -inf logitmi
+        self.assertTrue(torch.isfinite(dist.entropy()).all())
+        self.assertTrue(torch.isfinite(dist.log_prob(torch.tensor([1]))).all())
+
+    def test_encoding_tensors(self):
+        r = Round(3, 0)
+        obs = obs_tensor(encode_observation(r, 0))
+        self.assertEqual(obs.shape, (OBS_DIM,))
+        self.assertEqual(mask_tensor(guess_mask(r)).shape, (N_GUESS_ACTIONS,))
+        self.assertEqual(mask_tensor(play_mask(r, 0)).shape, (N_PLAY_ACTIONS,))
+
+
+class NeuralPlayerCase(unittest.TestCase):
+    def test_untrained_net_plays_legal_full_rounds(self):
+        torch.manual_seed(1)
+        net = BridzikNet(hidden=32)
+        env = RoundEnv(Random(2))
+        players = [NeuralPlayer(net, greedy=True),
+                   NeuralPlayer(net, greedy=False),
+                   RandomPlayer(Random(3)), RandomPlayer(Random(4))]
+        # dohratie kola bez BridzikException = vsetky tahy legalne
+        for round_number in range(8):
+            rewards = play_round(players, env, round_number)
+            self.assertEqual(len(rewards), 4)
+
+
+class SelfPlayCase(unittest.TestCase):
+    def test_collect_episodes_batch_consistency(self):
+        torch.manual_seed(5)
+        net = BridzikNet(hidden=32)
+        batch = collect_episodes(net, n_episodes=6, rng=Random(6))
+
+        n = batch['obs'].shape[0]
+        self.assertGreater(n, 0)
+        for key, width in (('guess_mask', N_GUESS_ACTIONS),
+                           ('play_mask', N_PLAY_ACTIONS)):
+            self.assertEqual(batch[key].shape, (n, width))
+        for key in ('phase_play', 'action', 'logp', 'value', 'ret'):
+            self.assertEqual(batch[key].shape, (n,))
+
+        # kazda epizoda ma prave 4 guess kroky -> pocet guess krokov = 4*epizody
+        self.assertEqual(int((~batch['phase_play']).sum()), 4 * 6)
+        # return je bud 0 alebo (10+tip)/REWARD_SCALE, cize v (0.55, 1.0]
+        for r in batch['ret'].tolist():
+            self.assertTrue(r == 0.0 or 10.0 / 18.0 <= r <= 1.0)
+        # akcia bola vzdy legalna podla ulozenej masky svojej fazy
+        for i in range(n):
+            mask = batch['play_mask'][i] if batch['phase_play'][i] \
+                else batch['guess_mask'][i]
+            self.assertTrue(bool(mask[batch['action'][i]]))
+        self.assertGreaterEqual(batch['mean_points'], 0.0)
+
+
+class OpponentMixingCase(unittest.TestCase):
+    def test_assign_seats_always_keeps_a_net_seat(self):
+        rng = Random(20)
+        marker = object()
+        for _ in range(200):
+            seats = _assign_seats(rng, 1.0, 0.0, marker, marker)
+            self.assertIn(None, seats.values())  # aj pri mix_random=1.0
+            self.assertEqual(set(seats), {0, 1, 2, 3})
+
+    def test_mixed_episodes_record_only_net_seats(self):
+        torch.manual_seed(21)
+        net = BridzikNet(hidden=32)
+        # mix_random=1.0 -> presne jedno sietove sedadlo na epizodu
+        batch = collect_episodes(net, n_episodes=5, rng=Random(22),
+                                 mix_random=1.0)
+        # 1 sietove sedadlo = presne 1 guess krok na epizodu
+        self.assertEqual(int((~batch['phase_play']).sum()), 5)
+        for i in range(batch['obs'].shape[0]):
+            mask = batch['play_mask'][i] if batch['phase_play'][i] \
+                else batch['guess_mask'][i]
+            self.assertTrue(bool(mask[batch['action'][i]]))
+        for r in batch['ret'].tolist():
+            self.assertTrue(r == 0.0 or 10.0 / 18.0 <= r <= 1.0)
+
+    def test_mixed_episodes_with_heuristic(self):
+        torch.manual_seed(23)
+        net = BridzikNet(hidden=32)
+        batch = collect_episodes(net, n_episodes=4, rng=Random(24),
+                                 mix_heuristic=0.5, heuristic_samples=10)
+        n_guess = int((~batch['phase_play']).sum())
+        self.assertGreaterEqual(n_guess, 4)   # aspon 1 sietove sedadlo/epizodu
+        self.assertLessEqual(n_guess, 16)
+        self.assertGreaterEqual(batch['mean_points'], 0.0)
+
+
+class PpoUpdateCase(unittest.TestCase):
+    def test_update_changes_params_and_is_finite(self):
+        torch.manual_seed(7)
+        net = BridzikNet(hidden=32)
+        optimizer = torch.optim.Adam(net.parameters(), lr=1e-3)
+        batch = collect_episodes(net, n_episodes=8, rng=Random(8))
+
+        before = [p.detach().clone() for p in net.parameters()]
+        stats = ppo_update(net, optimizer, batch, epochs=2, minibatch=64)
+
+        for key in ('policy_loss', 'value_loss', 'entropy'):
+            self.assertTrue(torch.isfinite(torch.tensor(stats[key])))
+        changed = any(
+            not torch.equal(b, a.detach())
+            for b, a in zip(before, net.parameters())
+        )
+        self.assertTrue(changed)
+
+    def test_value_head_learns_constant_reward(self):
+        # sanity uciaceho kroku: na batchi s konstantnym returnom sa value
+        # loss po par updatoch zmensi
+        torch.manual_seed(9)
+        net = BridzikNet(hidden=32)
+        optimizer = torch.optim.Adam(net.parameters(), lr=3e-3)
+        batch = collect_episodes(net, n_episodes=8, rng=Random(10))
+        batch['ret'] = torch.full_like(batch['ret'], 12.0 / 18.0)
+
+        first = ppo_update(net, optimizer, batch, epochs=1, minibatch=4096)
+        for _ in range(10):
+            last = ppo_update(net, optimizer, batch, epochs=1, minibatch=4096)
+        self.assertLess(last['value_loss'], first['value_loss'])
+
+
+class CheckpointCase(unittest.TestCase):
+    def test_save_load_roundtrip_with_hidden(self):
+        import os
+        import tempfile
+        from rl.train import load_checkpoint, save_checkpoint
+
+        torch.manual_seed(13)
+        net = BridzikNet(hidden=48)
+        path = os.path.join(tempfile.gettempdir(), 'bridzik_ckpt_test.pt')
+        try:
+            save_checkpoint(net, 48, path)
+            loaded = load_checkpoint(path)
+            self.assertEqual(loaded.trunk[0].out_features, 48)
+            obs = torch.zeros((1, OBS_DIM))
+            for a, b in zip(net(obs), loaded(obs)):
+                self.assertTrue(torch.equal(a, b))
+        finally:
+            os.remove(path)
+
+
+class EvaluateIntegrationCase(unittest.TestCase):
+    def test_neural_player_in_harness(self):
+        torch.manual_seed(11)
+        net = BridzikNet(hidden=32)
+        rng = Random(12)
+        stats = evaluate(
+            [NeuralPlayer(net)] + [RandomPlayer(rng) for _ in range(3)],
+            n_rounds=20, rng=rng,
+        )
+        self.assertEqual(stats[0]['rounds'], 20)
+
+
+if __name__ == '__main__':
+    unittest.main(verbosity=2)