这轮不追新 bar、不改规则，只复用上轮 clean replication 同一份代理事件样本，检查 strict CLV 的 short 改善是不是只是局部 pocket。
Hard verdict：park / evidence pool
short strict-CLV 只剩单一时间 pocket，long 侧各臂三桶全负；时间稳定性已把它打回 park。
short_clv080：只有 1/3 时间桶为正；第二桶重新明显转负，最差桶约 -9.32%。short_clv070_plus_volume：同样只有 1/3 时间桶为正；volume 没把中段稳定救回来。long_volume_only / long_volume_plus_clv：三桶全负，long 侧仍不支持把 CLV-only 或 volume+CLV 写成 continuation 充分条件。换成人话：Rank 99 留下的是 short-biased bar-quality 线索，不是能穿过时间稳定性的 shared hard gate。因此这轮应把它压回 park / evidence pool，然后切 fresh repo intake。
| variant | time_bucket | mean_total_return | positive_asset_ratio | mean_trade_count | mean_net_ret_h4 |
|---|---|---|---|---|---|
| long_baseline | bucket_1 | -9.88% | 0.00% | 159.7 | -5.91 bps |
| long_baseline | bucket_2 | -27.66% | 0.00% | 159.7 | -20.45 bps |
| long_baseline | bucket_3 | -17.44% | 0.00% | 159.7 | -11.46 bps |
| long_volume_only | bucket_1 | -4.15% | 33.33% | 114.3 | -3.07 bps |
| long_volume_only | bucket_2 | -20.06% | 0.00% | 114.0 | -19.63 bps |
| long_volume_only | bucket_3 | -8.48% | 0.00% | 113.7 | -7.26 bps |
| long_volume_plus_clv | bucket_1 | -7.46% | 0.00% | 73.7 | -9.60 bps |
| long_volume_plus_clv | bucket_2 | -13.15% | 0.00% | 73.7 | -19.20 bps |
| long_volume_plus_clv | bucket_3 | -4.16% | 0.00% | 73.7 | -4.97 bps |
| short_baseline | bucket_1 | 5.54% | 66.67% | 165.3 | 3.45 bps |
| short_baseline | bucket_2 | -20.23% | 0.00% | 165.0 | -13.60 bps |
| short_baseline | bucket_3 | -11.90% | 33.33% | 164.7 | -7.52 bps |
| short_clv070_plus_volume | bucket_1 | 10.07% | 66.67% | 88.7 | 11.48 bps |
| short_clv070_plus_volume | bucket_2 | -11.75% | 0.00% | 88.3 | -13.48 bps |
| short_clv070_plus_volume | bucket_3 | -1.92% | 66.67% | 88.3 | -1.68 bps |
| short_clv080 | bucket_1 | 8.35% | 66.67% | 82.0 | 12.09 bps |
| short_clv080 | bucket_2 | -9.32% | 0.00% | 81.7 | -12.10 bps |
| short_clv080 | bucket_3 | -2.27% | 33.33% | 81.7 | -1.31 bps |