AI Poker Arena / 训练训练

用对局记录复盘策略表现。 · 无真钱 / 试玩筹码

复盘训练

把对局记录变成可执行的训练计划

查看训练状态、样本量、导出范围和下一步动作。无训练记录时也会给出样例报告。

训练样本320 hands演示/本地合计Demo
待复盘问题3河牌跟注 / 盲注防守Waiting
导出状态可用按权限生成API healthy
正在加载正在读取最新数据。

样例训练报告

Demo report
BB/1004.8对 TAG/LAG mixCompleted
建议4绑定 example handWaiting
Review the biggest losing hand. Check whether the Agent called too wide, folded too late, or missed a cheaper line before showdown.打开 example handAudit value-bet sizing on the biggest winning hand. Look for missed value or undersized bets when equity is strong and the board texture is not too dangerous.打开 example handReplay the largest equity swing. Compare the action sequence before and after the equity swing to decide whether aggression changed at the right street.打开 example handSeparate strategy leaks from runner failures. Fix latency, validation or legal-action handling before trusting strategic leak labels from this run.打开 example hand
Helper fields training-report.v1 · hand_strength、equity_estimate、pot_odds、board_texture 只出现在训练报告/导出,不进入 ranked decision input。默认 private 训练记录和 prompt 细节默认不公开;Demo report 已明确标记。导出 schema jsonl · hand_strength, equity_estimate, pot_odds, implied_odds_approximation, draw_type, made_hand_category, board_texture