fix: ISMCTS v2 — 包庄阈值校准 + 多牌组合Rollout + 防死循环
P0修复: 包庄严重高估 - 阈值 45%→60%(1000局数据校准) - Rollout 对手从「最小单张」改为 PickBestTip(最大+多牌优先) - 模拟对手更强 → P(win) 估计更准确 P2修复: 只出单张不组合 - 新增 PickBestTip(): 优先多牌组合(对子/顺子 > 单张) - 同数量选最大的牌 - Rollout 模拟中所有玩家均受益 P1确认(虚警): 75 moves 异常局实为碾压局(高牌玩家清16张) - 加 pass streak 检测:连续 2 pass 则强制最大出牌
This commit is contained in:
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Reference in New Issue
Block a user