生成时间：2026-03-18 13:47 UTC
问题：在最小 clean replication 之后，Rank 55 只剩 1 次便宜诚实检查预算。这次不再重跑数据，只用现有 trade_log.csv 把每个 asset × setup × variant 按时间顺序切成 3 个等样本窗口，检查收益是否只是偶然 pocket。
desk verdict：park / evidence pool。
一句话原因：唯一能把三段时间窗口都维持为正的是 ema_psar_long + binary_crash_gate，但它仍只在单一 archetype 上成立，且每桶平均 trades 仅约 1.7~2.7；作为共享 crash-risk overlay，这还不足以诚实升格到 P2 / paper candidate。
breakout_short 三个变体在三个时间窗口里全部非正，说明这条 shared crash-risk overlay 没有把 short archetype 稳定救回来。fib_retest_long 只在零碎窗口小幅转正，幅度基本贴近噪音，不足以支撑升格。ema_psar_long + binary_crash_gate 虽然是唯一三桶都为正的组合，但它仍只服务单一 archetype，而且每桶平均 trades 偏薄，不能把“局部少亏/小幅转正”误写成 desk 级 shared overlay 通过。| setup | variant | time_bucket | mean_total_return | positive_asset_ratio | mean_trades | mean_false_hold_4bars_rate | min_asset_return | max_asset_return |
|---|---|---|---|---|---|---|---|---|
| breakout_short | base | bucket_1 | -0.67% | 66.67% | 3.00 | 55.56% | -3.38% | 1.27% |
| breakout_short | base | bucket_2 | -0.76% | 0.00% | 3.00 | 44.44% | -1.47% | -0.28% |
| breakout_short | base | bucket_3 | -1.08% | 0.00% | 2.00 | 66.67% | -1.30% | -0.92% |
| breakout_short | binary_crash_gate | bucket_1 | -0.30% | 33.33% | 2.00 | 77.78% | -1.81% | 1.27% |
| breakout_short | binary_crash_gate | bucket_2 | -1.08% | 0.00% | 2.00 | 55.56% | -2.77% | -0.19% |
| breakout_short | binary_crash_gate | bucket_3 | -0.51% | 33.33% | 1.67 | 50.00% | -1.15% | 0.66% |
| breakout_short | size_haircut | bucket_1 | -0.88% | 33.33% | 3.00 | 55.56% | -4.04% | 1.53% |
| breakout_short | size_haircut | bucket_2 | -0.90% | 0.00% | 3.00 | 44.44% | -1.70% | -0.34% |
| breakout_short | size_haircut | bucket_3 | -0.93% | 0.00% | 2.00 | 66.67% | -1.23% | -0.64% |
| ema_psar_long | base | bucket_1 | -1.43% | 33.33% | 3.00 | 66.67% | -7.96% | 8.17% |
| ema_psar_long | base | bucket_2 | 1.50% | 100.00% | 3.00 | 44.44% | 0.82% | 2.74% |
| ema_psar_long | base | bucket_3 | 1.41% | 100.00% | 2.00 | 50.00% | 0.26% | 3.37% |
| ema_psar_long | binary_crash_gate | bucket_1 | 0.54% | 33.33% | 2.67 | 50.00% | -4.72% | 8.17% |
| ema_psar_long | binary_crash_gate | bucket_2 | 0.99% | 66.67% | 2.00 | 50.00% | -1.14% | 3.86% |
| ema_psar_long | binary_crash_gate | bucket_3 | 1.36% | 100.00% | 1.67 | 50.00% | 0.26% | 3.37% |
| ema_psar_long | size_haircut | bucket_1 | -0.56% | 33.33% | 3.00 | 66.67% | -6.34% | 8.17% |
| ema_psar_long | size_haircut | bucket_2 | 1.36% | 66.67% | 3.00 | 44.44% | -0.16% | 3.30% |
| ema_psar_long | size_haircut | bucket_3 | 1.39% | 100.00% | 2.00 | 50.00% | 0.26% | 3.37% |
| fib_retest_long | base | bucket_1 | 0.03% | 33.33% | 3.00 | 55.56% | -0.27% | 0.64% |
| fib_retest_long | base | bucket_2 | -0.21% | 33.33% | 3.00 | 77.78% | -1.68% | 1.20% |
| fib_retest_long | base | bucket_3 | 0.21% | 66.67% | 2.00 | 33.33% | -0.57% | 0.61% |
| fib_retest_long | binary_crash_gate | bucket_1 | -0.29% | 33.33% | 2.67 | 66.67% | -1.24% | 0.64% |
| fib_retest_long | binary_crash_gate | bucket_2 | 0.50% | 66.67% | 2.00 | 66.67% | -2.03% | 2.08% |
| fib_retest_long | binary_crash_gate | bucket_3 | -0.20% | 66.67% | 1.67 | 50.00% | -1.22% | 0.61% |
| fib_retest_long | size_haircut | bucket_1 | -0.01% | 33.33% | 3.00 | 55.56% | -0.40% | 0.64% |
| fib_retest_long | size_haircut | bucket_2 | -0.02% | 66.67% | 3.00 | 77.78% | -1.68% | 1.33% |
| fib_retest_long | size_haircut | bucket_3 | 0.04% | 66.67% | 2.00 | 33.33% | -1.07% | 0.61% |
| setup | variant | overall_mean_total_return | overall_positive_asset_ratio | overall_mean_trades | positive_bucket_count | weak_bucket_count | min_bucket_return | max_bucket_return | bucket_return_spread | min_bucket_positive_asset_ratio | min_bucket_mean_trades | verdict | reason |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| breakout_short | base | -2.49% | 0.00% | 8.00 | 0 | 3 | -1.08% | -0.67% | 0.41% | 0.00% | 2.00 | fails_time_check | most buckets non-positive |
| breakout_short | binary_crash_gate | -1.88% | 0.00% | 5.67 | 0 | 3 | -1.08% | -0.30% | 0.79% | 0.00% | 1.67 | fails_time_check | most buckets non-positive |
| breakout_short | size_haircut | -2.68% | 0.00% | 8.00 | 0 | 3 | -0.93% | -0.88% | 0.05% | 0.00% | 2.00 | fails_time_check | most buckets non-positive |
| ema_psar_long | base | 1.63% | 33.33% | 8.00 | 2 | 1 | -1.43% | 1.50% | 2.93% | 33.33% | 2.00 | watch_only | mixed |
| ema_psar_long | binary_crash_gate | 3.15% | 33.33% | 6.33 | 3 | 0 | 0.54% | 1.36% | 0.82% | 33.33% | 1.67 | survives_time_check | all buckets positive but still needs cross-setup honesty |
| ema_psar_long | size_haircut | 2.40% | 33.33% | 8.00 | 2 | 1 | -0.56% | 1.39% | 1.94% | 33.33% | 2.00 | watch_only | mixed |
| fib_retest_long | base | 0.03% | 66.67% | 8.00 | 2 | 1 | -0.21% | 0.21% | 0.42% | 33.33% | 2.00 | watch_only | mixed |
| fib_retest_long | binary_crash_gate | 0.00% | 66.67% | 6.33 | 1 | 2 | -0.29% | 0.50% | 0.79% | 33.33% | 1.67 | fails_time_check | most buckets non-positive |
| fib_retest_long | size_haircut | 0.02% | 66.67% | 8.00 | 1 | 2 | -0.02% | 0.04% | 0.06% | 33.33% | 2.00 | fails_time_check | most buckets non-positive |
当前样本里最能留下来的是 fib_retest_long / size_haircut：整体回报 0.02%，正资产占比 66.67%，每桶最少平均 trades 2.00。但它仍不足以把 Rank 55 从 P1 weak candidate 升到 P2。
产物：reports/artifacts/scout_rank55_order_imbalance_crash_risk_15m/time_stability_window_summary.csv、reports/artifacts/scout_rank55_order_imbalance_crash_risk_15m/time_stability_verdict_summary.csv