feat: V-learning架构 — 改Q→V

Q-learning问题: pass action Q值稳定偏高 → 永远pass
V-learning修复:
- 网络输出 14(Q-action)→1(V-state): P(win|state)
- NeuralBot: 对每个候选编码出牌后state → 选V值最高的
- pass和出牌都是候选, 公平竞争(不是固定action slot)

TrainingCollector:
- 新增 EncodeAfterPlay(view, tracker, afterHand, playedIds)
- EncodeState 支持 extraSeen 参数(模拟出牌时标记已见)

Python:
- train_v.py: V-learning训练器(load CSV→predict win/loss)
- network.py: OUTPUT_DIM=14→1

验证: 200局0错误, V-model loss 0.81→0.0001
需更多数据覆盖状态空间

PdkFriendServer/Logic/PdkGameMain.cs

@@ -60,8 +60,9 @@ namespace PdkFriendServer.Logic
                 /// BotTypes: "ISMCTS" / "NEURAL" / null=默认ISMCTS
                 /// 训练模式: ["NEURAL","NEURAL","ISMCTS"] = 2个模型bot + 1个ISMCTS</summary>
                 public static string[] BotTypes = null;
+                        private static int _flushGameCounter = 0;
 
-                private void InitBots()
+                        private void InitBots()
                 {
                     _bots = new IPdkBot[3];
                     for (int i = 0; i < 3; i++)
@@ -485,10 +486,23 @@ namespace PdkFriendServer.Logic
             for (int i = 0; i < GamePack.Info.PlayNum; i++)
             {
                 Log($"玩家：{(i + 1)} ，输赢：{GamePack.Info.Score[i]}");
-            }
-#endif
+                            }
+                #endif
 
-            if (DeskGameDo != null)
+                            // 训练数据收集：输出本局所有 ISMCTS 决策
+                            if (TrainingCollector.Enabled)
+                            {
+                                float[] rewards = new float[GamePack.Info.PlayNum];
+                                for (int i = 0; i < GamePack.Info.PlayNum; i++)
+                                {
+                                    // 赢的分 > 0 → reward=1; 输的分 < 0 → reward=-1; 否则 0
+                                    float s = GamePack.Info.Score[i];
+                                    rewards[i] = s > 0 ? 1f : s < 0 ? -1f : 0f;
+                                }
+                                TrainingCollector.FlushGame(_flushGameCounter++, rewards);
+                            }
+
+                            if (DeskGameDo != null)
             {
                 DeskGameDo.GameOver();
             }