feat(SCOPONE-0009): complete iteration 0 dealer AI

docs/ARCHITECTURE.md

@@ -1,21 +1,21 @@
 # Architecture
 
-> Last Updated: 2026-04-02T19:05:00.000Z
+> Last Updated: 2026-04-08T19:48:08.000Z
 
 ## Overview
 
 | Attribute | Value |
 |-----------|-------|
 | Primary language | TypeScript |
-| Secondary language | Java (Capacitor Android shell) |
+| Secondary language | Java |
-| Project type | Browser card game with Android packaging |
+| Project type | Phaser browser game packaged for Android with Capacitor |
 | Framework | Phaser 3.87.0 |
 | Tooling | Vite 5, TypeScript 5.x, Capacitor 8.3 |
 | Runtime layout | 1280 x 720, `Phaser.Scale.FIT`, centered in `#game` |
 | Build command | `npm run build` |
 | Test command | `npx tsc --noEmit` |
 
-The repository is a TypeScript-first Scopone Scientifico implementation. The web client contains the real game logic and presentation. The Android tree is a Capacitor wrapper with a custom immersive `MainActivity`.
+The repository is a TypeScript-first implementation of Scopone Scientifico. Gameplay rules, scoring, inference, and AI live in framework-independent modules under `src/game/`. Phaser scenes under `src/scenes/` own rendering, input, UI, animation, and scene transitions. The `android/` tree is a Capacitor wrapper with a small custom `MainActivity` for immersive full-screen behavior.
 
 ## Project Structure
 
@@ -37,6 +37,8 @@ scopone-phaser/
 |     `- GameScene.ts
 |- public/
 |- android/
+|  |- app/
+|  `- variables.gradle
 |- docs/
 |- prompts/
 |- package.json
@@ -49,102 +51,102 @@ scopone-phaser/
 
 | Directory | Purpose |
 |-----------|---------|
-| `src/game/` | Framework-independent rules, scoring, imperfect-information tracking, and AI search |
+| `src/game/` | Rules engine, score calculation, imperfect-information tracking, AI heuristics, and master search |
-| `src/scenes/` | Phaser scene lifecycle, UI, animation, effects, and round orchestration |
+| `src/scenes/` | Phaser scene lifecycle, menus, board rendering, interaction, HUD, audio, and FX |
-| `public/` | Static web assets consumed by Phaser loaders |
+| `public/` | Atlas metadata and other static assets loaded by Phaser |
-| `android/` | Capacitor Android project, Gradle config, and immersive activity wrapper |
+| `android/` | Capacitor Android project, Gradle configuration, generated wrapper assets, and the native activity |
 | `docs/` | Architecture, code style, findings, and cache metadata |
-| `prompts/` | JIRA pipeline artifacts and iteration state |
+| `prompts/` | JIRA workflow artifacts and iteration state |
 
 ## Design Patterns
 
 No explicit GoF patterns were detected in the source or by semantic search.
 
-Observed architectural patterns in the current codebase:
+Observed architectural patterns:
 
 | Pattern | Where it appears |
 |---------|------------------|
 | Scene-based flow | `BootScene -> MenuScene -> GameScene` via Phaser scene registration |
-| Functional core / imperative shell | `src/game/` stays free of Phaser imports, while `src/scenes/` owns rendering and input |
+| Functional core / imperative shell | `src/game/` avoids Phaser imports while `src/scenes/` owns runtime side effects |
-| Clone-before-mutate state transitions | `applyMove()` clones `GameState` before applying move effects |
+| Immutable state transitions | `applyMove()` clones `GameState` before mutating round state |
-| Worker offload with fallback | `AIWorkerClient` runs heavy AI inside `ai.worker.ts` and falls back to in-thread `chooseMove()` if worker startup or messaging fails |
+| Worker offload with fallback | `AIWorkerClient` uses `ai.worker.ts` when available and falls back to direct `chooseMove()` otherwise |
-| Determinization search | Master AI samples hidden hands before alpha-beta evaluation |
+| Typed message protocol | `ai-worker-protocol.ts` defines worker request, progress, result, and error shapes |
-| Message-based progress reporting | Worker and main thread exchange typed request/result/progress messages through `ai-worker-protocol.ts` |
+| Imperfect-information search | `CardTracker` plus determinization sampling support the `master` AI tier |
 
 ## Key Components
 
 ### `src/main.ts`
 
-- Bootstraps `Phaser.Game`.
+- Creates the `Phaser.Game` instance.
 - Registers `BootScene`, `MenuScene`, and `GameScene`.
-- Installs a one-shot fullscreen request handler on first user interaction.
+- Installs a one-shot fullscreen request on first user input when supported.
 
 ### `src/game/types.ts`
 
-- Defines the game model: `Card`, `Capture`, `Player`, `GameState`, `TeamScore`, `ScoreBreakdown`.
+- Defines the core game model: `Card`, `Capture`, `Player`, `GameState`, `TeamScore`, and `ScoreBreakdown`.
-- Encodes difficulty tiers as `'beginner' | 'advanced' | 'master'`.
+- Models constrained domains with unions such as `PlayerIndex` and `Difficulty`.
-- Stores the `PRIMIERA_VALUES` lookup table.
+- Stores `PRIMIERA_VALUES` for end-of-round scoring.
 
-### `src/game/engine.ts` (371 lines)
+### `src/game/engine.ts`
 
 - Builds and shuffles the 40-card deck.
-- Creates the initial round state for four players.
+- Creates a round state for four players with dealer-relative opening order.
-- Implements capture selection rules: single direct matches take priority; subset sums are considered only when no direct match exists.
+- Implements capture rules where direct value matches take priority over subset-sum captures.
-- Applies moves immutably, detects scopas, assigns leftover table cards, and computes round scores.
+- Applies moves immutably, awards scope, assigns leftover table cards, and computes round and match scoring.
 
-### `src/game/card-tracker.ts` (89 lines)
+### `src/game/card-tracker.ts`
 
-- Tracks seen cards across a round without exposing hidden hands directly.
+- Tracks cards visible through play and capture events without exposing hidden hands.
-- Computes unseen cards from `played + myHand + table`.
+- Reconstructs unseen cards from `played + myHand + table`.
-- Supplies probability helpers used by the AI for value-based inference.
+- Supplies value and suit residue helpers used by AI inference and probability estimates.
 
 ### `src/game/ai.ts`
 
-- Exposes `chooseMove()` as an async entry point.
+- Exposes `chooseMove()` as the async AI entry point.
-- Implements three difficulty levels:
+- Implements three difficulty tiers: `beginner`, `advanced`, and `master`.
-    - `beginner`: noisy heuristic play.
+- Uses table-driven search profiles, role-aware heuristics, tracker-based inference, and determinization plus alpha-beta search.
-    - `advanced`: stronger heuristics with race awareness, partner setup, and card-tracker inference.
+- Configures the current master profile with a 4600 ms budget, 10 samples, depth 6, and batch size 2.
-    - `master`: determinization plus alpha-beta search with dynamic time budgets, batching, and progress callbacks.
-- Uses `yieldToBrowser()` between master-search batches so Phaser can repaint the think bar.
 
 ### `src/game/ai-worker-protocol.ts`
 
-- Defines the typed message contract between the main thread and the worker.
+- Defines a single `choose-move` worker request type.
-- Serializes requests around `GameState`, `Difficulty`, `PlayerIndex`, tracker snapshots, progress, results, and worker-safe errors.
+- Defines typed progress, result, and serialized error responses.
+- Keeps worker communication schema isolated from UI code.
 
 ### `src/game/ai-worker-client.ts`
 
-- Wraps the worker lifecycle behind the same `chooseMove()` API the scene needs.
+- Wraps worker lifecycle and pending-request tracking behind the same `chooseMove()` API that scenes consume.
-- Creates module workers with `new Worker(new URL('./ai.worker.ts', import.meta.url), { type: 'module' })`.
+- Creates the worker as an ES module with `new Worker(new URL('./ai.worker.ts', import.meta.url), { type: 'module' })`.
-- Streams progress callbacks back into `GameScene` and degrades to direct `chooseMove()` execution when workers are unavailable.
+- Fails over pending requests to in-thread AI execution if worker creation, messaging, or deserialization fails.
 
 ### `src/game/ai.worker.ts`
 
-- Rehydrates `CardTracker` from a snapshot, delegates move selection to `chooseMove()`, and posts progress/result/error messages back to the scene thread.
+- Rehydrates `CardTracker` from a serialized snapshot.
-- Keeps the expensive `master` search off the main rendering thread when worker support is available.
+- Delegates move selection to `chooseMove()`.
+- Posts progress, result, or serialized error messages back to the main thread.
 
 ### `src/scenes/BootScene.ts`
 
-- Loads the card atlas and card-back texture.
+- Loads the card atlas and card back.
-- Shows a simple progress bar and transitions into the menu.
+- Displays a simple loading bar.
+- Transitions into `MenuScene` after asset load.
 
 ### `src/scenes/MenuScene.ts`
 
-- Renders the title screen and rules summary.
+- Renders the title, rules summary, and difficulty selection.
-- Lets the player choose `beginner`, `advanced`, or `master` difficulty.
+- Starts `GameScene` with the chosen difficulty.
-- Starts `GameScene` with the selected difficulty in scene data.
 
 ### `src/scenes/GameScene.ts`
 
-- Owns the match loop, HUD, think bar, card interaction, animation, FX, audio, and round transitions.
+- Owns match flow, dealing, selection, capture resolution, AI turn orchestration, score HUD, status UI, think bar, particles, and procedural audio.
-- Uses `CardTracker` to record played and captured cards after each move.
+- Instantiates and disposes `AIWorkerClient` on scene lifecycle events.
-- Instantiates `AIWorkerClient`, bridges async AI progress into a visible top-of-screen think bar, and disposes worker resources on scene shutdown.
+- Updates `CardTracker` after play and capture events so AI inference remains derived from visible information.
-- Handles end-of-round overlays and full-match restart flow.
 
 ### `android/app/src/main/java/com/phaser/scopa/MainActivity.java`
 
 - Extends `BridgeActivity`.
-- Forces immersive mode by hiding status and navigation bars whenever the window gains focus.
+- Applies immersive mode during `onCreate()` and whenever window focus returns.
+- Hides status and navigation bars with transient swipe behavior.
 
 ## Dependencies
 
@@ -152,74 +154,77 @@ Observed architectural patterns in the current codebase:
 
 | Package | Version | Purpose |
 |---------|---------|---------|
-| `phaser` | `^3.87.0` | Game engine |
+| `phaser` | `^3.87.0` | Game engine runtime |
-| `@capacitor/core` | `^8.3.0` | Capacitor runtime |
+| `@capacitor/core` | `^8.3.0` | Capacitor runtime bridge |
-| `@capacitor/cli` | `^8.3.0` | Capacitor tooling |
+| `@capacitor/cli` | `^8.3.0` | Capacitor project tooling |
 | `@capacitor/android` | `^8.3.0` | Android platform integration |
 
 ### JavaScript development dependencies
 
 | Package | Version | Purpose |
 |---------|---------|---------|
-| `typescript` | `^5.0.0` | Type-checking and TS compilation for builds |
+| `typescript` | `^5.0.0` | Static type checking and TS compilation step |
-| `vite` | `^5.0.0` | Dev server and bundling |
+| `vite` | `^5.0.0` | Dev server and production bundler |
 
 ### Android / Gradle dependencies
 
-| Dependency | Source | Purpose |
+| Dependency | Version | Source | Purpose |
-|------------|--------|---------|
+|------------|---------|--------|---------|
-| `com.android.tools.build:gradle:8.13.0` | `android/build.gradle` | Android build plugin |
+| `com.android.tools.build:gradle` | `8.13.0` | `android/build.gradle` | Android Gradle plugin |
-| `com.google.gms:google-services:4.4.4` | `android/build.gradle` | Optional Google services integration |
+| `com.google.gms:google-services` | `4.4.4` | `android/build.gradle` | Optional Google services integration |
-| `androidx.appcompat:appcompat` | `android/app/build.gradle` | Android UI compatibility |
+| `androidx.appcompat:appcompat` | `1.7.1` | `android/variables.gradle` | Android app compatibility |
-| `androidx.coordinatorlayout:coordinatorlayout` | `android/app/build.gradle` | Android layout support |
+| `androidx.coordinatorlayout:coordinatorlayout` | `1.3.0` | `android/variables.gradle` | Layout coordination helpers |
-| `androidx.core:core-splashscreen` | `android/app/build.gradle` | Splash screen support |
+| `androidx.core:core-splashscreen` | `1.2.0` | `android/variables.gradle` | Splash screen support |
-| `junit:junit` | `android/app/build.gradle` | JVM-side Android tests |
+| `junit:junit` | `4.13.2` | `android/variables.gradle` | JVM Android tests |
-| `androidx.test.ext:junit` | `android/app/build.gradle` | Instrumented Android testing |
+| `androidx.test.ext:junit` | `1.3.0` | `android/variables.gradle` | Instrumented test runner |
-| `androidx.test.espresso:espresso-core` | `android/app/build.gradle` | Android UI testing |
+| `androidx.test.espresso:espresso-core` | `3.7.0` | `android/variables.gradle` | Instrumented UI testing |
+
+### Platform configuration
+
+- `compileSdkVersion`: 36
+- `targetSdkVersion`: 36
+- `minSdkVersion`: 24
 
 ## Module Organization
 
 ```text
 main.ts
-    -> BootScene
+  -> BootScene
-    -> MenuScene
+  -> MenuScene
-    -> GameScene
+  -> GameScene
-             -> engine.ts
+       -> engine.ts
-             -> ai.ts
+       -> types.ts
-           -> ai-worker-client.ts
+       -> card-tracker.ts
-               -> ai-worker-protocol.ts
+       -> ai-worker-client.ts
-               -> ai.worker.ts
+            -> ai-worker-protocol.ts
-                   -> ai.ts
+            -> ai.worker.ts
-             -> card-tracker.ts
+                 -> ai.ts
-             -> types.ts
 ```
 
-Dependencies are one-directional at the application level:
+Application-level dependency direction is one-way:
 
 - `src/game/` imports only from sibling game modules.
-- `src/scenes/` imports from `src/game/`.
+- `src/scenes/` imports from `src/game/` and Phaser.
 - `src/game/` never imports Phaser.
 
 ## Data Flow
 
 1. `main.ts` creates the Phaser app and registers all scenes.
-2. `BootScene` loads assets, then starts `MenuScene`.
+2. `BootScene` loads textures and starts `MenuScene`.
-3. `MenuScene` passes the chosen difficulty into `GameScene`.
+3. `MenuScene` passes the chosen difficulty to `GameScene`.
-4. `GameScene.create()` initializes a new `CardTracker`, creates the initial `GameState`, and animates the opening deal.
+4. `GameScene.create()` creates a fresh `CardTracker`, constructs a new `GameState`, and starts the opening deal.
-5. On each turn:
+5. Human turns use pointer-driven card selection and `findCaptures()` output to choose legal captures.
-     - Human turns use click-driven selection and capture highlighting.
+6. AI turns call `AIWorkerClient.chooseMove(state, playerIdx, difficulty, tracker, onProgress)`.
-    - AI turns call `AIWorkerClient.chooseMove(state, playerIdx, difficulty, tracker, onProgress)`.
+7. `AIWorkerClient` posts a typed request to `ai.worker.ts`; if workers are unavailable, it reruns the same request in-thread.
-6. `AIWorkerClient` posts a typed request into `ai.worker.ts`; if worker setup fails, it falls back to in-thread `chooseMove()`.
+8. `chooseMove()` returns a heuristic move for lower tiers or performs batched master search while emitting `AIDecisionProgress`.
-7. `chooseMove()` either returns immediately for heuristic tiers or performs batched master search while reporting `AIDecisionProgress`.
+9. `GameScene` updates the think bar from progress callbacks, executes the returned move, records tracker state, and advances turn order.
-8. Worker progress messages drive `GameScene.updateThinkBar()` until a result is posted back.
+10. When every hand is empty, `engine.ts` finalizes scoring and `GameScene` presents the round or match outcome.
-9. `GameScene.executeMove()` applies the move, updates the tracker, animates the result, refreshes the HUD, and advances the round.
-10. When all hands are empty, `engine.ts` finalizes scoring and `GameScene` displays the round summary or final match screen.
 
 ## Build System
 
 | Command | Source | Purpose |
 |---------|--------|---------|
-| `npm run dev` | `package.json` | Starts the Vite development server on port 3000 and opens the browser |
+| `npm run dev` | `package.json` | Starts the Vite dev server |
 | `npm run build` | user-provided build command | Runs `tsc && vite build` and writes web output to `dist/` |
-| `npm run preview` | `package.json` | Serves the built app locally via Vite preview |
+| `npm run preview` | `package.json` | Serves the built app with Vite preview |
 | `npx tsc --noEmit` | user-provided test command | Type-checks the TypeScript codebase without emitting files |

docs/FINDINGS.md

@@ -1,34 +1,35 @@
 # Findings
 
-> Last Updated: 2026-04-02T19:05:00.000Z
+> Last Updated: 2026-04-08T19:48:08.000Z
 
 ## Summary
 
-Initializer refresh for the current Scopone Scientifico codebase. The existing findings were stale relative to the current worker-backed AI execution path, so the observations below reflect the live source tree.
+Initializer refresh for SCOPONE-0009. The cached findings were stale relative to the live source tree, so the observations below reflect the current Phaser, worker, and AI implementation.
 
 ## Codebase Observations
 
-- Primary gameplay code lives in 8 TypeScript source files under `src/`; the Android wrapper adds 3 Java files.
+- Primary gameplay code currently lives in 10 TypeScript source files under `src/`; the Android wrapper adds 3 Java files.
-- The gameplay runtime now includes three AI transport files in addition to the rules engine: `ai-worker-protocol.ts`, `ai-worker-client.ts`, and `ai.worker.ts`.
+- The project is structurally split between framework-free gameplay modules in `src/game/` and Phaser scene code in `src/scenes/`.
-- The largest concentration of logic still sits in `src/scenes/GameScene.ts` and `src/game/ai.ts`.
+- `src/scenes/GameScene.ts` and `src/game/ai.ts` remain the two largest concentrations of application logic.
-- `src/game/` remains framework-independent and contains the rules engine, score calculation, card tracker, and AI logic.
+- The AI transport layer is now a stable three-file path: `ai-worker-protocol.ts`, `ai-worker-client.ts`, and `ai.worker.ts`.
-- The AI now has three distinct difficulty levels: `beginner`, `advanced`, and `master`.
+- The AI exposes three difficulty levels: `beginner`, `advanced`, and `master`.
-- The `advanced` and `master` tiers use `CardTracker` to reason about unseen cards instead of reading hidden hands directly.
+- `advanced` and `master` both use `CardTracker` to reason about unseen cards without directly reading hidden hands.
-- The `master` tier performs determinization plus alpha-beta search and reports progress back through `AIDecisionProgress`.
+- The current `master` search profile is `timeBudgetMs: 4600`, `sampleCount: 10`, `maxDepth: 6`, `batchSize: 2`.
-- `GameScene` displays AI progress through a top think bar and updates it from worker-forwarded progress messages.
+- `GameScene` consumes AI progress callbacks to update an on-screen think bar while a worker request is running.
-- `AIWorkerClient` degrades cleanly to in-thread `chooseMove()` execution when workers are unavailable or fail.
+- `AIWorkerClient` fails over pending work to in-thread `chooseMove()` if worker creation, posting, or deserialization fails.
-- Audio remains fully procedural via Web Audio; no audio asset pipeline is present.
+- The Android wrapper targets SDK 36 with `minSdkVersion` 24 and applies immersive mode from the native activity.
-- No ESLint or Prettier config is present.
+- Audio remains procedural via Web Audio; no dedicated audio asset pipeline is present in the source tree.
+- No ESLint or Prettier configuration is present.
 - The only repository-wide verification command supplied is `npx tsc --noEmit`.
 
 ## Potential Improvement Areas
 
-- `GameScene.ts` still centralizes scene layout, input, effects, audio, HUD, and round transitions in one file, which raises maintenance cost.
+- `GameScene.ts` still centralizes layout, turn flow, HUD updates, effects, and audio in one scene class, which raises maintenance cost.
-- `ai.ts` mixes heuristic tiers, inference helpers, determinization, and alpha-beta evaluation in one module.
+- `ai.ts` still combines heuristic tiers, inference helpers, determinization, and alpha-beta evaluation in one module.
-- Worker message types and fallback behavior are separated cleanly, but the UI still knows about AI progress presentation details directly.
+- Worker transport is isolated cleanly, but progress rendering remains coupled to scene-level UI concerns.
-- The `master` profile allows up to 9800 ms of search budget, which may be expensive on slower devices even with batch yielding.
+- A 4600 ms master search budget may still be noticeable on slower mobile devices even with batch yielding.
-- There is still no dedicated automated test suite for rules or AI behavior beyond type-checking.
+- There is no dedicated automated rules or AI test suite beyond type-checking.
-- Formatting rules are enforced socially rather than by a linter/formatter toolchain.
+- Formatting and style are enforced socially rather than by automated linting or formatting tools.
 
 ## Current Rule / Implementation Notes
 
@@ -36,30 +37,32 @@ Initializer refresh for the current Scopone Scientifico codebase. The existing f
 
 - Direct-match capture has priority over subset-sum capture.
 - When multiple direct matches exist, `findCaptures()` returns one single-card option per matching card.
-- Subset-sum captures are explored only when no direct match exists.
+- Subset-sum captures are considered only when no direct match exists.
 - `applyMove()` defaults to the first legal capture if no explicit capture choice is supplied.
+- Scope is awarded only when a capture clears the table before the final play of the round.
 
 ### AI implementation snapshot
 
-- `beginner` adds randomness around a basic heuristic to remain beatable.
+- `beginner` uses a simpler heuristic with noise to remain beatable.
 - `advanced` adds race awareness, anti-scopa logic, partner setup, anchor play, and tracker-based probability estimates.
-- `master` orders legal moves with a quick evaluator, samples hidden hands, then scores moves with alpha-beta search under a deadline.
+- `master` orders legal moves with a quick evaluator, samples hidden hands, and scores them with alpha-beta search under the active deadline.
-- `masterMove()` yields back to the browser between batches so Phaser can repaint the progress UI.
+- Progress is reported through `AIDecisionProgress` so the scene can keep the think bar responsive.
 
 ### Worker execution snapshot
 
-- `GameScene` creates a fresh `AIWorkerClient` on scene creation and disposes it on shutdown.
+- `GameScene` creates `AIWorkerClient` during `create()` and disposes it on both `shutdown` and `destroy`.
-- `AIWorkerClient` serializes a tracker snapshot instead of sending a live `CardTracker` instance across the worker boundary.
+- `AIWorkerClient` serializes `CardTracker` state through `toSnapshot()` instead of attempting to transfer the class instance.
-- `ai.worker.ts` reconstructs tracker state with `CardTracker.fromSnapshot()` before calling `chooseMove()`.
+- `ai.worker.ts` rebuilds tracker state with `CardTracker.fromSnapshot()` before calling `chooseMove()`.
-- Progress, results, and serialized worker errors all travel through `ai-worker-protocol.ts`.
+- Progress, result, and serialized error payloads all travel through `ai-worker-protocol.ts`.
-- If worker initialization, posting, or message deserialization fails, pending requests are rerun with the in-thread AI path.
+- If worker execution becomes unavailable, pending requests are rerun with the in-thread AI path rather than being dropped.
 
 ### Scene / UI implementation snapshot
 
+- `BootScene` loads atlas assets and presents a simple loading bar.
 - `MenuScene` exposes difficulty selection before match start.
-- `GameScene` records every played card and captured table card in `CardTracker`.
+- `GameScene` tracks played and captured cards in `CardTracker` as the round evolves.
-- The HUD continuously displays cards, denari, settebello, primiera, scope, and total points for both teams.
+- The scene owns score HUD rendering, player labels, status text, think-bar rendering, and procedural particle effects.
-- Round-end and game-over flows are managed in-scene rather than through separate overlay components.
+- Round-end and match-end flows remain managed inside the scene instead of separate overlay components.
 
 ## Research Performed
 
@@ -106,4 +109,11 @@ Initializer refresh for the current Scopone Scientifico codebase. The existing f
 - The current implementation does not use Phaser `TimerEvent` progress helpers.
 - Instead, `chooseMove()` emits its own normalized progress payload through `AIDecisionProgress`.
 - `GameScene.updateThinkBar()` renders remaining time from that callback.
-- The yielding behavior in `masterMove()` is necessary so the browser can repaint while search batches continue.
+- The yielding behavior in the master search path is necessary so the browser can repaint while search batches continue.
+
+### SCOPONE-0009: Phaser scene lifecycle notes (2026-04-08)
+
+- Source: Context7 `/websites/phaser_io_api-documentation`, query `Phaser 3.87 Scene lifecycle create restart shutdown destroy event listeners scene restart preserving external state`.
+- Phaser dispatches `shutdown` when a scene stops being active but may be re-used later; resource cleanup that should also cover final teardown can additionally listen to `destroy`.
+- The current `GameScene` pattern of registering one-shot shutdown and destroy handlers is aligned with Phaser guidance for worker disposal and UI cleanup.
+- Dealer rotation and next-round state changes can stay inside the existing in-scene orchestration without requiring a different Phaser lifecycle primitive.

src/game/ai.ts

@@ -1,5 +1,5 @@
-import { Card, GameState, PlayerIndex, Difficulty, PRIMIERA_VALUES, Suit, SUITS } from './types';
+import { Card, GameState, PlayerIndex, Difficulty, PRIMIERA_VALUES, Suit, SUITS, DealerRelativeRole } from './types';
-import { findCaptures, canCapture, teamOf, applyMove, buildDeck, cloneState } from './engine';
+import { findCaptures, canCapture, teamOf, applyMove, buildDeck, cloneState, getDealerRelativeRole } from './engine';
 import { CardTracker } from './card-tracker';
 
 export interface AIMove {
@@ -22,10 +22,59 @@ interface SearchProfile {
   batchSize: number;
 }
 
+interface DealerRoleContext {
+  role: DealerRelativeRole;
+  onDealerSide: boolean;
+  defendingDealerAdvantage: boolean;
+  attackingDealerAdvantage: boolean;
+  aggressionBias: number;
+  controlBias: number;
+  pairPreservingBias: number;
+  parityBreakingBias: number;
+  tablePressureBias: number;
+}
+
+interface ParitySnapshot {
+  unseenCounts: number[];
+  oddResidue: boolean[];
+  evenResidue: boolean[];
+}
+
+const DEALER_ROLE_WEIGHTS: Record<DealerRelativeRole, Omit<DealerRoleContext, 'role' | 'onDealerSide' | 'defendingDealerAdvantage' | 'attackingDealerAdvantage'>> = {
+  'first-hand': {
+    aggressionBias: 1.28,
+    controlBias: 0.9,
+    pairPreservingBias: 0.88,
+    parityBreakingBias: 1.26,
+    tablePressureBias: 1.3,
+  },
+  'second-hand': {
+    aggressionBias: 1,
+    controlBias: 1.08,
+    pairPreservingBias: 1.12,
+    parityBreakingBias: 0.96,
+    tablePressureBias: 1,
+  },
+  'third-hand': {
+    aggressionBias: 1.16,
+    controlBias: 0.94,
+    pairPreservingBias: 0.94,
+    parityBreakingBias: 1.16,
+    tablePressureBias: 1.12,
+  },
+  dealer: {
+    aggressionBias: 0.84,
+    controlBias: 1.32,
+    pairPreservingBias: 1.34,
+    parityBreakingBias: 0.82,
+    tablePressureBias: 0.78,
+  },
+};
+
 const SEARCH_PROFILES: Record<Difficulty, SearchProfile> = {
   beginner: { timeBudgetMs: 120, sampleCount: 0, maxDepth: 0, batchSize: 0 },
   advanced: { timeBudgetMs: 650, sampleCount: 0, maxDepth: 0, batchSize: 0 },
-  master: { timeBudgetMs: 9800, sampleCount: 12, maxDepth: 6, batchSize: 2 },
+  master: { timeBudgetMs: 4600, sampleCount: 10, maxDepth: 6, batchSize: 2 },
 };
 
 // ---------------------------------------------------------------------------
@@ -72,18 +121,192 @@ function countValueInHand(hand: Card[], value: number): number {
   return n;
 }
 
+function getDealerRoleContext(state: GameState, playerIdx: PlayerIndex): DealerRoleContext {
+  const role = getDealerRelativeRole(state.dealer, playerIdx);
+  const onDealerSide = role === 'dealer' || role === 'second-hand';
+  return {
+    role,
+    onDealerSide,
+    defendingDealerAdvantage: onDealerSide,
+    attackingDealerAdvantage: !onDealerSide,
+    ...DEALER_ROLE_WEIGHTS[role],
+  };
+}
+
+function getParitySnapshot(
+  tracker: CardTracker | undefined,
+  myHand: Card[],
+  table: Card[],
+): ParitySnapshot | null {
+  if (!tracker) return null;
+
+  const unseenCounts = Array.from({ length: 11 }, () => 0);
+  const oddResidue = Array.from({ length: 11 }, () => false);
+  const evenResidue = Array.from({ length: 11 }, () => false);
+  const summary = tracker.getValueParityResidueSummary(myHand, table);
+
+  for (const residue of summary) {
+    unseenCounts[residue.value] = residue.unseenCount;
+    oddResidue[residue.value] = residue.hasOddUnseenResidue;
+    evenResidue[residue.value] = residue.hasEvenUnseenResidue;
+  }
+
+  return { unseenCounts, oddResidue, evenResidue };
+}
+
+function countParityValuesOnTable(afterTable: Card[], parity: ParitySnapshot | null): { oddValues: number; evenValues: number } {
+  if (!parity || afterTable.length === 0) {
+    return { oddValues: 0, evenValues: 0 };
+  }
+
+  let oddValues = 0;
+  let evenValues = 0;
+  const seenValues = new Set<number>();
+
+  for (const card of afterTable) {
+    if (seenValues.has(card.value)) continue;
+    seenValues.add(card.value);
+    if (parity.oddResidue[card.value]) oddValues++;
+    else if (parity.evenResidue[card.value]) evenValues++;
+  }
+
+  return { oddValues, evenValues };
+}
+
+function scoreRoleTablePlan(
+  afterTable: Card[],
+  roleContext: DealerRoleContext,
+  nextIsOpp: boolean,
+): number {
+  if (afterTable.length === 0) return 0;
+
+  const tableSum = afterTable.reduce((sum, card) => sum + card.value, 0);
+  let score = 0;
+
+  if (roleContext.role === 'first-hand') {
+    if (afterTable.length >= 2) score += 22 * roleContext.tablePressureBias;
+    if (tableSum >= 8 && tableSum <= 15) score += 18 * roleContext.aggressionBias;
+  }
+
+  if (roleContext.role === 'third-hand') {
+    if (afterTable.length >= 2) score += 14 * roleContext.tablePressureBias;
+    if (tableSum >= 10) score += 10 * roleContext.aggressionBias;
+  }
+
+  if (roleContext.role === 'second-hand') {
+    if (nextIsOpp && tableSum >= 11) score += 16 * roleContext.controlBias;
+    if (!nextIsOpp && tableSum <= 10) score += 10 * roleContext.tablePressureBias;
+  }
+
+  if (roleContext.role === 'dealer') {
+    if (tableSum >= 11) score += 28 * roleContext.controlBias;
+    if (tableSum <= 10 && nextIsOpp) score -= 24 * roleContext.controlBias;
+    if (afterTable.length === 1 && nextIsOpp) score -= 16 * roleContext.controlBias;
+  }
+
+  return Math.round(score);
+}
+
+function scoreParityTableState(
+  afterTable: Card[],
+  parity: ParitySnapshot | null,
+  roleContext: DealerRoleContext,
+  nextIsOpp: boolean,
+): number {
+  const { oddValues, evenValues } = countParityValuesOnTable(afterTable, parity);
+  if (oddValues === 0 && evenValues === 0) return 0;
+
+  let score = 0;
+  if (roleContext.defendingDealerAdvantage) {
+    score += evenValues * 18 * roleContext.controlBias;
+    score -= oddValues * 22 * roleContext.controlBias;
+    if (nextIsOpp) score += evenValues * 8 - oddValues * 10;
+  } else {
+    score += oddValues * 20 * roleContext.tablePressureBias;
+    score -= evenValues * 10;
+    if (nextIsOpp) score += oddValues * 12;
+  }
+
+  return Math.round(score);
+}
+
+function scoreCaptureParityPlan(
+  played: Card,
+  captured: Card[],
+  afterTable: Card[],
+  parity: ParitySnapshot | null,
+  roleContext: DealerRoleContext,
+  nextIsOpp: boolean,
+): number {
+  if (!parity || captured.length === 0) return 0;
+
+  let score = 0;
+  const directCapture = captured.length === 1 && captured[0].value === played.value;
+
+  if (directCapture) {
+    const unseenCount = parity.unseenCounts[played.value] ?? 0;
+    const base = parity.oddResidue[played.value] ? 58 : 30;
+    score += base * roleContext.pairPreservingBias;
+    if (roleContext.defendingDealerAdvantage && unseenCount > 0) score += 18 * roleContext.controlBias;
+  } else {
+    let parityBreaks = 0;
+    let oddTargets = 0;
+    const seenValues = new Set<number>();
+
+    for (const card of captured) {
+      if (seenValues.has(card.value)) continue;
+      seenValues.add(card.value);
+      if ((parity.unseenCounts[card.value] ?? 0) > 0) parityBreaks++;
+      if (parity.oddResidue[card.value]) oddTargets++;
+    }
+
+    const disruption = parityBreaks * 20 + oddTargets * 18 + Math.max(0, captured.length - 1) * 12;
+    score += disruption * roleContext.parityBreakingBias;
+    if (roleContext.defendingDealerAdvantage) score -= 18 * roleContext.controlBias;
+  }
+
+  score += scoreParityTableState(afterTable, parity, roleContext, nextIsOpp);
+  return Math.round(score);
+}
+
+function scoreDumpParityPlan(
+  card: Card,
+  afterTable: Card[],
+  parity: ParitySnapshot | null,
+  roleContext: DealerRoleContext,
+  nextIsOpp: boolean,
+): number {
+  if (!parity) return 0;
+
+  let score = scoreParityTableState(afterTable, parity, roleContext, nextIsOpp);
+  if (parity.oddResidue[card.value]) {
+    score += roleContext.attackingDealerAdvantage ? 18 * roleContext.tablePressureBias : -20 * roleContext.controlBias;
+  }
+  if (parity.evenResidue[card.value]) {
+    score += roleContext.defendingDealerAdvantage ? 14 * roleContext.pairPreservingBias : 6;
+  }
+
+  return Math.round(score);
+}
+
 function getSearchProfile(state: GameState, difficulty: Difficulty): SearchProfile {
   if (difficulty !== 'master') return SEARCH_PROFILES[difficulty];
 
   const cardsRemaining = state.players.reduce((sum, player) => sum + player.hand.length, 0);
+  if (cardsRemaining <= 4) {
+    return { timeBudgetMs: 3200, sampleCount: 4, maxDepth: cardsRemaining, batchSize: 1 };
+  }
   if (cardsRemaining <= 6) {
-    return { timeBudgetMs: 9800, sampleCount: 18, maxDepth: Math.min(cardsRemaining, 8), batchSize: 1 };
+    return { timeBudgetMs: 3600, sampleCount: 6, maxDepth: cardsRemaining, batchSize: 1 };
+  }
+  if (cardsRemaining <= 8) {
+    return { timeBudgetMs: 3900, sampleCount: 8, maxDepth: cardsRemaining, batchSize: 1 };
   }
   if (cardsRemaining <= 12) {
-    return { timeBudgetMs: 9000, sampleCount: 16, maxDepth: 8, batchSize: 2 };
+    return { timeBudgetMs: 4200, sampleCount: 8, maxDepth: 8, batchSize: 1 };
   }
   if (cardsRemaining <= 20) {
-    return { timeBudgetMs: 8200, sampleCount: 14, maxDepth: 7, batchSize: 2 };
+    return { timeBudgetMs: 4400, sampleCount: 9, maxDepth: 7, batchSize: 2 };
   }
   return SEARCH_PROFILES.master;
 }
@@ -432,6 +655,8 @@ function advancedMove(state: GameState, playerIdx: PlayerIndex, tracker?: CardTr
   const table = state.table;
   const phase = gamePhase(state);
   const race = getRaceState(state, playerIdx);
+  const roleContext = getDealerRoleContext(state, playerIdx);
+  const parity = getParitySnapshot(tracker, player.hand, table);
   const next = nextPlayer(playerIdx);
   const nextIsOpp = isOpponent(playerIdx, next);
   const partner = partnerOf(playerIdx);
@@ -447,14 +672,14 @@ function advancedMove(state: GameState, playerIdx: PlayerIndex, tracker?: CardTr
       for (const captureSet of captures) {
         const score = scoreCaptureAdv(
           card, captureSet, table, state, playerIdx, race,
-          tracker, player.hand, phase, nextIsOpp, partnerHandSize, lastPlay,
+          tracker, player.hand, phase, nextIsOpp, partnerHandSize, lastPlay, roleContext, parity,
         );
         if (score > bestScore) { bestScore = score; bestMove = { card, capture: captureSet }; }
       }
     } else {
       const score = scoreDumpAdv(
         card, table, state, playerIdx, race,
-        tracker, player.hand, phase, nextIsOpp, partnerHandSize, lastPlay,
+        tracker, player.hand, phase, nextIsOpp, partnerHandSize, lastPlay, roleContext, parity,
       );
       if (score > bestScore) { bestScore = score; bestMove = { card, capture: [] }; }
     }
@@ -467,7 +692,7 @@ function scoreCaptureAdv(
   played: Card, captured: Card[], table: Card[], state: GameState,
   playerIdx: PlayerIndex, race: RaceState, tracker: CardTracker | undefined,
   myHand: Card[], phase: number, nextIsOpp: boolean, partnerHandSize: number,
-  lastPlay: boolean,
+  lastPlay: boolean, roleContext: DealerRoleContext, parity: ParitySnapshot | null,
 ): number {
   let score = 100;
   const allCaptured = [played, ...captured];
@@ -508,6 +733,8 @@ function scoreCaptureAdv(
     }
   }
 
+  score += scoreCaptureParityPlan(played, captured, afterTable, parity, roleContext, nextIsOpp);
+
   // --- ANCHOR STRATEGY ---
   // Prefer captures that leave table cards matching values we hold (we can recapture)
   if (!isScopa) {
@@ -543,6 +770,8 @@ function scoreCaptureAdv(
     }
   }
 
+  score += scoreRoleTablePlan(afterTable, roleContext, nextIsOpp);
+
   // --- PARTNER COOPERATION ---
   const next = nextPlayer(playerIdx);
   if (!isScopa && !isOpponent(playerIdx, next)) {
@@ -620,6 +849,14 @@ function scoreCaptureAdv(
     if (tableSum <= 5 && nextIsOpp) score -= 60;
   }
 
+  if (roleContext.role === 'first-hand' && !isScopa && afterTable.length >= 2) {
+    score += Math.round(24 * roleContext.tablePressureBias);
+  }
+  if (roleContext.role === 'dealer' && !isScopa) {
+    const tableSum = afterTable.reduce((s, c) => s + c.value, 0);
+    if (tableSum >= 11) score += Math.round(30 * roleContext.controlBias);
+  }
+
   return score;
 }
 
@@ -627,7 +864,7 @@ function scoreDumpAdv(
   card: Card, table: Card[], state: GameState,
   playerIdx: PlayerIndex, race: RaceState, tracker: CardTracker | undefined,
   myHand: Card[], phase: number, nextIsOpp: boolean, partnerHandSize: number,
-  lastPlay: boolean,
+  lastPlay: boolean, roleContext: DealerRoleContext, parity: ParitySnapshot | null,
 ): number {
   let score = 0;
   const afterTable = [...table, card];
@@ -653,6 +890,8 @@ function scoreDumpAdv(
   const partnerProb = partnerLikelyHolds(card.value, playerIdx, state, tracker, myHand, table);
   if (partnerProb > 0.4) score += 55; // partner can recapture what we dump
 
+  score += scoreDumpParityPlan(card, afterTable, parity, roleContext, nextIsOpp);
+
   // --- ANTI-SCOPA ---
   if (tableSum >= 11) {
     score += 150;
@@ -694,6 +933,8 @@ function scoreDumpAdv(
     score += 20; // safe dump before partner's turn, signals we don't need this suit
   }
 
+  score += scoreRoleTablePlan(afterTable, roleContext, nextIsOpp);
+
   // --- CARD TRACKING ---
   if (tracker) {
     const unseen = tracker.getUnseenCards(myHand, afterTable);
@@ -729,6 +970,13 @@ function scoreDumpAdv(
     if (card.value >= 8) score += 15;
   }
 
+  if (roleContext.role === 'first-hand' && afterTable.length >= 2 && tableSum >= 8) {
+    score += Math.round(18 * roleContext.tablePressureBias);
+  }
+  if (roleContext.role === 'dealer' && nextIsOpp && tableSum <= 10) {
+    score -= Math.round(24 * roleContext.controlBias);
+  }
+
   return score;
 }
 
@@ -744,6 +992,8 @@ function tableControlPressure(
   tracker: CardTracker | undefined,
   myHand: Card[],
   race: RaceState,
+  roleContext: DealerRoleContext,
+  parity: ParitySnapshot | null,
 ): number {
   if (afterTable.length === 0) return 0;
 
@@ -790,6 +1040,8 @@ function tableControlPressure(
   }
 
   if (race.aheadOverall && nextIsOpp && tableSum <= 10) score -= 60;
+  score += scoreRoleTablePlan(afterTable, roleContext, nextIsOpp);
+  score += scoreParityTableState(afterTable, parity, roleContext, nextIsOpp);
   return score;
 }
 
@@ -814,9 +1066,11 @@ async function masterMove(
   // Quick-eval move ordering for better pruning
   const lastPlay = isLastPlay(state, playerIdx);
   const race = getRaceState(state, playerIdx);
+  const roleContext = getDealerRoleContext(state, playerIdx);
+  const parity = getParitySnapshot(tracker, state.players[playerIdx].hand, state.table);
   const quickScored = legalMoves.map(m => ({
     move: m,
-    quick: quickEval(m, state, playerIdx, tracker, lastPlay, race),
+    quick: quickEval(m, state, playerIdx, tracker, lastPlay, race, roleContext, parity),
   }));
   quickScored.sort((a, b) => b.quick - a.quick);
   const sortedMoves = quickScored.map(qs => qs.move);
@@ -891,15 +1145,19 @@ function quickEval(
   move: AIMove, state: GameState, playerIdx: PlayerIndex,
   tracker: CardTracker | undefined, lastPlay: boolean,
   race: RaceState,
+  roleContext: DealerRoleContext,
+  parity: ParitySnapshot | null,
 ): number {
   let score = 0;
   const table = state.table;
-  const afterTable = table.filter(c => !move.capture.some(cc => cc.id === c.id));
+  const afterCaptureTable = table.filter(c => !move.capture.some(cc => cc.id === c.id));
+  const projectedTable = move.capture.length > 0 ? afterCaptureTable : [...afterCaptureTable, move.card];
+  const projectedHand = state.players[playerIdx].hand.filter(card => card.id !== move.card.id);
   const allCaptured = [move.card, ...move.capture];
   const nextIsOpp = isOpponent(playerIdx, nextPlayer(playerIdx));
 
   // Scopa (not on last play!)
-  if (move.capture.length > 0 && afterTable.length === 0) {
+  if (move.capture.length > 0 && projectedTable.length === 0) {
     score += lastPlay ? 50 : 1200;
   }
 
@@ -924,21 +1182,35 @@ function quickEval(
   }
 
   // Anti-scopa
-  if (afterTable.length > 0) {
+  if (projectedTable.length > 0) {
-    const sum = afterTable.reduce((s, c) => s + c.value, 0);
+    const sum = projectedTable.reduce((s, c) => s + c.value, 0);
     if (sum <= 10 && nextIsOpp) score -= 180;
     if (sum >= 11) score += 60;
-    if (afterTable.length === 1 && nextIsOpp) score -= 120;
+    if (projectedTable.length === 1 && nextIsOpp) score -= 120;
   }
 
   // Partner awareness
   const next = nextPlayer(playerIdx);
-  if (!isOpponent(playerIdx, next) && afterTable.length > 0) {
+  if (!isOpponent(playerIdx, next) && projectedTable.length > 0) {
-    const sum = afterTable.reduce((s, c) => s + c.value, 0);
+    const sum = projectedTable.reduce((s, c) => s + c.value, 0);
     if (sum >= 1 && sum <= 10) score += 40; // partner might scopa
   }
 
-  score += tableControlPressure(afterTable, state, playerIdx, tracker, state.players[playerIdx].hand, race);
+  score += scoreCaptureParityPlan(move.card, move.capture, projectedTable, parity, roleContext, nextIsOpp);
+  if (move.capture.length === 0) {
+    score += scoreDumpParityPlan(move.card, projectedTable, parity, roleContext, nextIsOpp);
+  }
+
+  score += tableControlPressure(
+    projectedTable,
+    state,
+    playerIdx,
+    tracker,
+    projectedHand,
+    race,
+    roleContext,
+    parity,
+  );
 
   return score;
 }
@@ -1036,20 +1308,22 @@ function alphaBeta(
   tracker: CardTracker | undefined,
 ): number {
   if (depth === 0 || state.roundOver || Date.now() > deadline) {
-    return evaluateFast(state, myTeam, phase);
+    return evaluateFast(state, myTeam, phase, tracker, rootPlayer);
   }
 
   const cur = state.currentPlayer;
   const isMyTeam = teamOf(cur) === myTeam;
   const moves = getLegalMoves(state, cur);
 
-  if (moves.length === 0) return evaluateFast(state, myTeam, phase);
+  if (moves.length === 0) return evaluateFast(state, myTeam, phase, tracker, rootPlayer);
 
   // Move ordering: settebello captures first, then scopa, then captures by size, then dumps
   if (moves.length > 2) {
     const race = getRaceState(state, cur);
     const lastPlay = isLastPlay(state, cur);
-    moves.sort((a, b) => quickEval(b, state, cur, tracker, lastPlay, race) - quickEval(a, state, cur, tracker, lastPlay, race));
+    const roleContext = getDealerRoleContext(state, cur);
+    const parity = getParitySnapshot(tracker, state.players[rootPlayer].hand, state.table);
+    moves.sort((a, b) => quickEval(b, state, cur, tracker, lastPlay, race, roleContext, parity) - quickEval(a, state, cur, tracker, lastPlay, race, roleContext, parity));
   }
 
   if (isMyTeam) {
@@ -1076,12 +1350,20 @@ function alphaBeta(
 }
 
 /** Fast evaluation: avoids flatMap/filter at every leaf node */
-function evaluateFast(state: GameState, myTeam: 0 | 1, phase: number): number {
+function evaluateFast(
+  state: GameState,
+  myTeam: 0 | 1,
+  phase: number,
+  tracker: CardTracker | undefined,
+  rootPlayer: PlayerIndex,
+): number {
   const p0 = state.players[0], p1 = state.players[1], p2 = state.players[2], p3 = state.players[3];
   const myA = myTeam === 0 ? p0 : p1;
   const myB = myTeam === 0 ? p2 : p3;
   const oppA = myTeam === 0 ? p1 : p0;
   const oppB = myTeam === 0 ? p3 : p2;
+  const roleContext = getDealerRoleContext(state, state.currentPlayer);
+  const parity = getParitySnapshot(tracker, state.players[rootPlayer].hand, state.table);
 
   // Single-pass pile scan — no flatMap/filter allocations
   let myCards = 0, oppCards = 0;
@@ -1204,6 +1486,21 @@ function evaluateFast(state: GameState, myTeam: 0 | 1, phase: number): number {
       // Good: table has cards we can capture
       score += state.table.length * 5;
     }
+
+    const parityPressure = scoreParityTableState(state.table, parity, roleContext, !myTurn);
+    score += myTurn ? parityPressure : -parityPressure;
+
+    const rolePlan = scoreRoleTablePlan(state.table, roleContext, !myTurn);
+    score += myTurn ? rolePlan : -rolePlan;
+
+    if (parity) {
+      const { oddValues, evenValues } = countParityValuesOnTable(state.table, parity);
+      if (roleContext.defendingDealerAdvantage) {
+        score += (myTurn ? evenValues : -oddValues) * 14;
+      } else {
+        score += (myTurn ? oddValues : -evenValues) * 14;
+      }
+    }
   }
 
   return score;

src/game/card-tracker.ts

@@ -4,12 +4,35 @@ export interface CardTrackerSnapshot {
   playedCardIds: string[];
 }
 
+export interface CardTrackerValueParityResidue {
+  value: number;
+  knownCount: number;
+  unseenCount: number;
+  hasOddUnseenResidue: boolean;
+  hasEvenUnseenResidue: boolean;
+}
+
+interface VisibleValueResidueKnowledge {
+  unseenCards: Card[];
+  unseenCountBySuit: Record<Suit, number>;
+  unseenCountByValue: number[];
+  valueParityResidues: CardTrackerValueParityResidue[];
+}
+
 function normalizeSnapshot(snapshot: CardTrackerSnapshot): CardTrackerSnapshot {
   return {
     playedCardIds: Array.from(new Set(snapshot.playedCardIds)),
   };
 }
 
+function createEmptySuitCounts(): Record<Suit, number> {
+  const counts = {} as Record<Suit, number>;
+  for (const suit of SUITS) {
+    counts[suit] = 0;
+  }
+  return counts;
+}
+
 /**
  * Tracks which cards have been played/captured during a round.
  * Used by AI to infer opponent hands WITHOUT cheating.
@@ -65,44 +88,96 @@ export class CardTracker {
     return !this.played.has('denara_7');
   }
 
+  private buildVisibleValueResidueKnowledge(myHand: Card[], table: Card[]): VisibleValueResidueKnowledge {
+    const knownCardIds = new Set<string>(this.played);
+    for (const card of myHand) {
+      knownCardIds.add(card.id);
+    }
+    for (const card of table) {
+      knownCardIds.add(card.id);
+    }
+
+    const knownCountByValue = Array.from({ length: 11 }, () => 0);
+    const unseenCountByValue = Array.from({ length: 11 }, () => 0);
+    const unseenCountBySuit = createEmptySuitCounts();
+    const unseenCards: Card[] = [];
+
+    for (const suit of SUITS) {
+      for (let value = 1; value <= 10; value++) {
+        const id = `${suit}_${value}`;
+        if (knownCardIds.has(id)) {
+          knownCountByValue[value] += 1;
+          continue;
+        }
+
+        unseenCountByValue[value] += 1;
+        unseenCountBySuit[suit] += 1;
+        unseenCards.push({ suit, value, id });
+      }
+    }
+
+    const valueParityResidues: CardTrackerValueParityResidue[] = [];
+    for (let value = 1; value <= 10; value++) {
+      const unseenCount = unseenCountByValue[value];
+      valueParityResidues.push({
+        value,
+        knownCount: knownCountByValue[value],
+        unseenCount,
+        hasOddUnseenResidue: unseenCount % 2 === 1,
+        hasEvenUnseenResidue: unseenCount % 2 === 0,
+      });
+    }
+
+    return {
+      unseenCards,
+      unseenCountBySuit,
+      unseenCountByValue,
+      valueParityResidues,
+    };
+  }
+
   /**
    * Get cards that could be in opponent hands.
    * = full 40-card deck minus: already played, my hand, currently on table
    */
   getUnseenCards(myHand: Card[], table: Card[]): Card[] {
-    const known = new Set<string>();
+    return this.buildVisibleValueResidueKnowledge(myHand, table).unseenCards;
-    for (const id of this.played) known.add(id);
-    for (const c of myHand) known.add(c.id);
-    for (const c of table) known.add(c.id);
-
-    const unseen: Card[] = [];
-    for (const suit of SUITS) {
-      for (let v = 1; v <= 10; v++) {
-        const id = `${suit}_${v}`;
-        if (!known.has(id)) {
-          unseen.push({ suit, value: v, id });
-        }
-      }
-    }
-    return unseen;
   }
 
   /** Count how many cards of a suit are still unseen */
   countRemainingSuit(suit: Suit, myHand: Card[], table: Card[]): number {
-    return this.getUnseenCards(myHand, table).filter(c => c.suit === suit).length;
+    return this.buildVisibleValueResidueKnowledge(myHand, table).unseenCountBySuit[suit];
   }
 
   /** Count how many unseen cards share a value */
   countRemainingValue(value: number, myHand: Card[], table: Card[]): number {
-    return this.getUnseenCards(myHand, table).filter(c => c.value === value).length;
+    return this.getValueParityResidue(value, myHand, table).unseenCount;
+  }
+
+  /** Get visible known-count, unseen-count, and parity residue for a single value */
+  getValueParityResidue(value: number, myHand: Card[], table: Card[]): CardTrackerValueParityResidue {
+    const valueParityResidues = this.buildVisibleValueResidueKnowledge(myHand, table).valueParityResidues;
+    return valueParityResidues[value - 1] ?? {
+      value,
+      knownCount: 0,
+      unseenCount: 0,
+      hasOddUnseenResidue: false,
+      hasEvenUnseenResidue: true,
+    };
+  }
+
+  /** Get visible known-count, unseen-count, and parity residue for all card values */
+  getValueParityResidueSummary(myHand: Card[], table: Card[]): CardTrackerValueParityResidue[] {
+    return this.buildVisibleValueResidueKnowledge(myHand, table).valueParityResidues;
   }
 
   /** Probability that a hidden hand contains at least one card with the requested value */
   probabilityHandHasValue(value: number, handSize: number, myHand: Card[], table: Card[]): number {
     if (handSize <= 0) return 0;
 
-    const unseen = this.getUnseenCards(myHand, table);
+    const visibleValueResidueKnowledge = this.buildVisibleValueResidueKnowledge(myHand, table);
-    const matching = unseen.filter(c => c.value === value).length;
+    const unseen = visibleValueResidueKnowledge.unseenCards;
+    const matching = visibleValueResidueKnowledge.unseenCountByValue[value] ?? 0;
     if (matching === 0) return 0;
     if (handSize >= unseen.length) return 1;
 

src/game/engine.ts

@@ -1,6 +1,6 @@
 import {
   Card, Suit, SUITS, Player, PlayerIndex, GameState,
-  TeamScore, ScoreBreakdown, PRIMIERA_VALUES, Capture
+  TeamScore, ScoreBreakdown, PRIMIERA_VALUES, Capture, DealerRelativeRole
 } from './types';
 
 // ---------------------------------------------------------------------------
@@ -76,8 +76,38 @@ export function canCapture(played: Card, table: Card[]): boolean {
 // Game state initialisation
 // ---------------------------------------------------------------------------
 
-export function createInitialState(startingPlayer: PlayerIndex = 0): GameState {
+export function nextPlayer(playerIdx: PlayerIndex, steps = 1): PlayerIndex {
+  return ((playerIdx + steps) % 4) as PlayerIndex;
+}
+
+export function getOpeningPlayerForDealer(dealer: PlayerIndex): PlayerIndex {
+  return nextPlayer(dealer);
+}
+
+export function getDealerRelativeOrder(
+  dealer: PlayerIndex
+): [PlayerIndex, PlayerIndex, PlayerIndex, PlayerIndex] {
+  const firstHand = getOpeningPlayerForDealer(dealer);
+  const secondHand = nextPlayer(firstHand);
+  const thirdHand = nextPlayer(secondHand);
+  return [firstHand, secondHand, thirdHand, dealer];
+}
+
+export function getDealerRelativeRole(
+  dealer: PlayerIndex,
+  playerIdx: PlayerIndex
+): DealerRelativeRole {
+  const [firstHand, secondHand, thirdHand] = getDealerRelativeOrder(dealer);
+
+  if (playerIdx === firstHand) return 'first-hand';
+  if (playerIdx === secondHand) return 'second-hand';
+  if (playerIdx === thirdHand) return 'third-hand';
+  return 'dealer';
+}
+
+export function createInitialState(dealer: PlayerIndex = 3): GameState {
   const deck = shuffle(buildDeck());
+  const startingPlayer = getOpeningPlayerForDealer(dealer);
 
   const players: [Player, Player, Player, Player] = [
     { index: 0, hand: [], pile: [], scope: 0, isHuman: true,  name: 'Tu' },
@@ -102,6 +132,7 @@ export function createInitialState(startingPlayer: PlayerIndex = 0): GameState {
     players,
     table,
     matchStartingPlayer: startingPlayer,
+    dealer,
     currentPlayer: startingPlayer,
     roundOver: false,
     gameOver: false,
@@ -164,7 +195,7 @@ export function applyMove(
   }
 
   // Advance turn
-  state2.currentPlayer = ((playerIdx + 1) % 4) as PlayerIndex;
+  state2.currentPlayer = nextPlayer(playerIdx);
 
   // Check if round is over (all hands empty)
   const allHandsEmpty = state2.players.every(p => p.hand.length === 0);
@@ -353,6 +384,7 @@ export function cloneState(state: GameState): GameState {
     ],
     table: state.table.map(cloneCard),
     matchStartingPlayer: state.matchStartingPlayer,
+    dealer: state.dealer,
     currentPlayer: state.currentPlayer,
     roundOver: state.roundOver,
     gameOver: state.gameOver,

src/game/types.ts

@@ -14,6 +14,8 @@ export interface Capture {
 
 export type PlayerIndex = 0 | 1 | 2 | 3;
 
+export type DealerRelativeRole = 'first-hand' | 'second-hand' | 'third-hand' | 'dealer';
+
 export type Difficulty = 'beginner' | 'advanced' | 'master';
 
 export interface Player {
@@ -29,6 +31,7 @@ export interface GameState {
   players: [Player, Player, Player, Player];
   table: Card[];
   matchStartingPlayer: PlayerIndex;
+  dealer: PlayerIndex;
   currentPlayer: PlayerIndex;
   roundOver: boolean;
   gameOver: boolean;

src/scenes/GameScene.ts

@@ -1,7 +1,8 @@
 import Phaser from 'phaser';
 import { Card, PlayerIndex, GameState, Difficulty } from '../game/types';
 import {
-  createInitialState, applyMove, findCaptures, getScoreBreakdown, teamOf, calcPrimiera, getMatchOutcome
+  createInitialState, applyMove, findCaptures, getScoreBreakdown, teamOf, calcPrimiera, getMatchOutcome,
+  nextPlayer
 } from '../game/engine';
 import { AIDecisionProgress } from '../game/ai';
 import { AIWorkerClient, AIWorkerClientLike } from '../game/ai-worker-client';
@@ -130,8 +131,8 @@ export class GameScene extends Phaser.Scene {
 
     this.input.once('pointerdown', () => this.startMusic());
 
-    const startingPlayer = Phaser.Math.Between(0, 3) as PlayerIndex;
+    const initialDealer = Phaser.Math.Between(0, 3) as PlayerIndex;
-    this.state = createInitialState(startingPlayer);
+    this.state = createInitialState(initialDealer);
     this.dealAnimation(() => {
       this.updateScoreBar();
       this.nextTurn();
@@ -1434,11 +1435,11 @@ export class GameScene extends Phaser.Scene {
     const totals = this.state.teamScores.map(t => t.totalPoints);
     const nextRound = (this.state.roundNumber ?? 1) + 1;
     const matchStartingPlayer = this.state.matchStartingPlayer;
-    const startingPlayer = ((matchStartingPlayer + nextRound - 1) % 4) as PlayerIndex;
+    const nextDealer = nextPlayer(this.state.dealer);
     for (const img of this.cardImages.values()) img.destroy();
     this.cardImages.clear();
     this.tracker.reset();
-    this.state = createInitialState(startingPlayer);
+    this.state = createInitialState(nextDealer);
     this.state.matchStartingPlayer = matchStartingPlayer;
     this.state.teamScores[0].totalPoints = totals[0];
     this.state.teamScores[1].totalPoints = totals[1];