冻结参数：lookback=4 / current<=0.790 / deepest<=0.790
执行口径：BTC/ETH/SOL 120d 15m + fib_retest_long + signal 当根及之前数据 + next-bar open + no-overlap + hold 8 bars
当前 hard verdict：park / evidence pool
改善主要靠缩样本，或连测试段的 hold-quality 指标都没同步变好；当前更诚实的处理是 park。
这轮只回答一件事：把 fib retest 的“当前回踩够浅吗”升级成“当前回踩 + 最近几根里最深曾经扎到哪”，能不能更诚实地过滤掉假 hold。
读法上，current_only 是只看信号当根 low 的回踩深度；current_plus_deepest 则额外要求最近 4 根里的最深 low 也不过阈值。
| variant | baseline_trades | variant_trades | trade_count_retention | baseline_return | variant_return | return_delta | baseline_failure_before_target | variant_failure_before_target | failure_delta | baseline_false_hold | variant_false_hold | false_hold_delta |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| current_only | 14.00 | 12.00 | 85.71% | 0.05% | 0.21% | 0.16% | 42.86% | 41.67% | -1.19% | 21.43% | 16.67% | -4.76% |
| current_plus_deepest | 14.00 | 8.00 | 57.14% | 0.05% | -0.06% | -0.12% | 42.86% | 50.00% | 7.14% | 21.43% | 12.50% | -8.93% |
| asset | variant | baseline_trades | variant_trades | trade_count_retention | baseline_return | variant_return | return_delta | baseline_failure_before_target | variant_failure_before_target | failure_delta | baseline_false_hold | variant_false_hold | false_hold_delta |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BTC-USD | current_only | 5.00 | 4.00 | 80.00% | 0.36% | 0.52% | 0.15% | 40.00% | 25.00% | -15.00% | 40.00% | 25.00% | -15.00% |
| BTC-USD | current_plus_deepest | 5.00 | 2.00 | 40.00% | 0.36% | 0.30% | -0.06% | 40.00% | 50.00% | 10.00% | 40.00% | 50.00% | 10.00% |
| ETH-USD | current_only | 4.00 | 4.00 | 100.00% | 0.29% | 0.29% | 0.00% | 50.00% | 50.00% | 0.00% | 25.00% | 25.00% | 0.00% |
| ETH-USD | current_plus_deepest | 4.00 | 2.00 | 50.00% | 0.29% | -0.22% | -0.52% | 50.00% | 50.00% | 0.00% | 25.00% | 0.00% | -25.00% |
| SOL-USD | current_only | 5.00 | 4.00 | 80.00% | -0.45% | -0.16% | 0.28% | 40.00% | 50.00% | 10.00% | 0.00% | 0.00% | 0.00% |
| SOL-USD | current_plus_deepest | 5.00 | 4.00 | 80.00% | -0.45% | -0.16% | 0.28% | 40.00% | 50.00% | 10.00% | 0.00% | 0.00% | 0.00% |
| variant | cost_bps_per_side | baseline_trades | variant_trades | trade_count_retention | baseline_return | variant_return | return_delta | baseline_failure_before_target | variant_failure_before_target | failure_delta |
|---|---|---|---|---|---|---|---|---|---|---|
| current_only | 6.00 | 14.00 | 12.00 | 85.71% | 0.05% | 0.21% | 0.16% | 42.86% | 41.67% | -1.19% |
| current_plus_deepest | 6.00 | 14.00 | 8.00 | 57.14% | 0.05% | -0.06% | -0.12% | 42.86% | 50.00% | 7.14% |
| current_only | 10.00 | 14.00 | 12.00 | 85.71% | -0.03% | 0.13% | 0.16% | 42.86% | 41.67% | -1.19% |
| current_plus_deepest | 10.00 | 14.00 | 8.00 | 57.14% | -0.03% | -0.14% | -0.12% | 42.86% | 50.00% | 7.14% |
| current_only | 15.00 | 14.00 | 12.00 | 85.71% | -0.13% | 0.03% | 0.16% | 42.86% | 41.67% | -1.19% |
| current_plus_deepest | 15.00 | 14.00 | 8.00 | 57.14% | -0.13% | -0.24% | -0.12% | 42.86% | 50.00% | 7.14% |
| lookback_bars | current_thr | deepest_thr | score | train_baseline_trades | train_gate_trades |
|---|---|---|---|---|---|
| 4.00 | 0.790 | 0.790 | -0.522 | 19.00 | 10.00 |
| 8.00 | 0.790 | 0.790 | -999.000 | 19.00 | 1.00 |
| 4.00 | 0.500 | 0.500 | -999.000 | 19.00 | 0.00 |
| 4.00 | 0.500 | 0.618 | -999.000 | 19.00 | 0.00 |
| 4.00 | 0.500 | 0.790 | -999.000 | 19.00 | 0.00 |
| 4.00 | 0.618 | 0.618 | -999.000 | 19.00 | 0.00 |
| 4.00 | 0.618 | 0.790 | -999.000 | 19.00 | 0.00 |
| 8.00 | 0.500 | 0.500 | -999.000 | 19.00 | 0.00 |
| 8.00 | 0.500 | 0.618 | -999.000 | 19.00 | 0.00 |
| 8.00 | 0.500 | 0.790 | -999.000 | 19.00 | 0.00 |
reports/artifacts/scout_rank126_deepest_retracement_hold_quality_15m/signal_catalog.csvreports/artifacts/scout_rank126_deepest_retracement_hold_quality_15m/trade_log.csvreports/artifacts/scout_rank126_deepest_retracement_hold_quality_15m/overall_summary.csvreports/artifacts/scout_rank126_deepest_retracement_hold_quality_15m/asset_summary.csvreports/artifacts/scout_rank126_deepest_retracement_hold_quality_15m/cost_summary.csvreports/artifacts/scout_rank126_deepest_retracement_hold_quality_15m/summary.json