Rank213 / 213c 第五轮：真实成本下利润厚度
对象：rank213_age90_14d_skip1d_voladj / 213c family。样本使用 monthly-volume causal universe、age≥90d、daily next-day hold，不重新下载行情。
生成时间：2026-05-07 05:37:32 UTC。成本表统一为 cost_bps_per_1x_turnover，覆盖 4/8/12/16bps；funding 只做 sensitivity，不计作真实收益。
结论
第五轮没有找到可以直接 Pass 的候选。 基准 daily short_every_day 在 12bps/1x turnover 后为 -2.19 bps，累计 -66.53%，最大回撤 -84.08%。它说明第四轮 flat 12bps 的表观余量，落到换手成本口径后并不够厚。
唯一 Watch 是 rank_buffer_extra_8：12bps 后日均 4.11 bps，累计 52.78%，换手 0.210x/day，但最大回撤仍有 -48.88%，不能作为放大 notional 的依据。
本轮最有用的方向是：rank buffer / 慢 rebalance / 持仓 carry 这类低换手结构。short gate 能改善部分均值，但回撤没有被压住；prior30 dispersion 仍更适合解释和监控，不适合作为主交易开关。
给新研究者的读法
- 本页回答的问题：如果把 213c 放到更接近实盘的换手成本口径，策略还有没有足够利润厚度。
- 先看 12bps 候选筛选。 Pass 表示可以进入下一轮严格验证，Watch 表示有研究价值但不能上线放大，Fail 表示当前形态不值得继续。
- 不要只看日均 bps。 还要同时看累计收益、最大回撤、换手率和 active_short_days。一个变体均值变高但回撤扩大，通常不能算进步。
- funding sensitivity 只是敏感性。 它不是已经确认的真实收益，不能拿来抵消 12bps 成本压力。
- 本页最重要的结论：213c 的下一轮优化应该优先降低换手和重构 short leg，而不是继续扩大参数网格。
核心读数
-2.19 bps基准 12bps 日均
4.11 bps唯一 Watch 12bps 日均
0.210x唯一 Watch 平均换手
0 / 1 / 31Pass / Watch / Fail
12bps 候选筛选
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days | reason |
|---|---|---|---|---|---|---|---|---|---|
| Watch | turnover_reduction | rank_buffer_extra_8 | 4.11 bps | 52.78% | -48.88% | 0.368 | 0.210 | 2257 | positive at 12bps, but drawdown/turnover still needs execution proof |
| Fail | turnover_reduction | rebalance_every_7d | 4.56 bps | 68.05% | -66.11% | 0.405 | 0.200 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | btc_prior7_positive | 3.40 bps | -7.74% | -72.77% | 0.237 | 0.501 | 1203 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | btc_above_ma20 | 2.98 bps | -17.76% | -77.30% | 0.205 | 0.475 | 1197 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | prior30_dispersion_high | 1.61 bps | -38.31% | -86.99% | 0.113 | 0.385 | 756 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | turnover_reduction | rebalance_every_2d | 1.06 bps | -28.67% | -63.68% | 0.089 | 0.406 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | dynamic_universe | static_top30 | 0.40 bps | -34.19% | -51.23% | 0.036 | 0.528 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | eligible_prior7_negative | 0.35 bps | -46.22% | -90.39% | 0.027 | 0.488 | 1075 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | signal_variant | retvol_21d_skip1d | 0.02 bps | -43.73% | -79.24% | 0.002 | 0.501 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | turnover_reduction | rank_buffer_extra_2 | -0.39 bps | -47.80% | -76.84% | -0.033 | 0.385 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | turnover_reduction | rebalance_every_3d | -0.74 bps | -52.44% | -69.64% | -0.063 | 0.325 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | eligible_prior7_positive | -1.03 bps | -66.82% | -89.78% | -0.072 | 0.493 | 1177 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | prior30_dispersion_mid_high | -1.07 bps | -63.31% | -84.60% | -0.079 | 0.495 | 1510 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | eligible_ew_above_ma20 | -1.21 bps | -68.18% | -87.96% | -0.084 | 0.456 | 1083 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | btc_prior30_positive | -1.37 bps | -68.65% | -83.56% | -0.096 | 0.469 | 1217 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | turnover_reduction | replacement_cap_1_per_side | -1.43 bps | -58.83% | -77.20% | -0.123 | 0.419 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | btc_above_ma50 | -1.44 bps | -69.20% | -83.81% | -0.101 | 0.458 | 1216 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | turnover_reduction | rank_buffer_extra_4 | -1.67 bps | -60.02% | -78.05% | -0.147 | 0.292 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | dynamic_universe | static_top20 | -1.91 bps | -60.93% | -77.61% | -0.174 | 0.478 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | turnover_reduction | replacement_cap_2_per_side | -2.16 bps | -65.96% | -82.81% | -0.181 | 0.572 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | short_every_day | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | signal_variant | retvol_14d_skip1d | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | dynamic_universe | static_top50 | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | signal_variant | rank_ensemble_7_14_21 | -2.56 bps | -68.90% | -92.10% | -0.216 | 0.635 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | eligible_ew_above_ma50 | -3.01 bps | -78.45% | -91.09% | -0.211 | 0.423 | 983 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | dynamic_universe | prior30_dispersion_low20_mid50_high80 | -3.13 bps | -73.44% | -88.45% | -0.258 | 0.594 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | signal_variant | downside_vol_14d_skip1d | -3.25 bps | -73.82% | -88.94% | -0.271 | 0.592 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | btc_prior7_negative | -4.10 bps | -80.70% | -96.43% | -0.313 | 0.487 | 1044 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | dynamic_universe | static_top80 | -4.16 bps | -81.13% | -94.32% | -0.314 | 0.642 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | short_gate | eligible_prior30_positive | -4.40 bps | -84.33% | -90.24% | -0.308 | 0.436 | 1015 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | signal_variant | retvol_7d_skip1d | -5.97 bps | -85.66% | -93.08% | -0.506 | 0.814 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
| Fail | signal_variant | raw_return_14d_skip1d | -8.90 bps | -93.78% | -95.14% | -0.669 | 0.557 | 2257 | does not clear positive 12bps net with acceptable drawdown/turnover |
筛选规则偏保守：必须在 12bps 后为正，同时回撤和换手不能太差；funding sensitivity 不参与真实候选排序。
12bps 全部实验 Top 24
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days | gross_mean_bps | long_half_mean_bps | short_half_mean_bps |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Watch | turnover_reduction | rank_buffer_extra_8 | 4.11 bps | 52.78% | -48.88% | 0.368 | 0.210 | 2257 | 6.62 bps | 5.50 bps | 1.13 bps |
| Fail | turnover_reduction | rebalance_every_7d | 4.56 bps | 68.05% | -66.11% | 0.405 | 0.200 | 2257 | 6.97 bps | 6.29 bps | 0.67 bps |
| Fail | short_gate | btc_prior7_positive | 3.40 bps | -7.74% | -72.77% | 0.237 | 0.501 | 1203 | 9.42 bps | 6.10 bps | 3.31 bps |
| Fail | short_gate | btc_above_ma20 | 2.98 bps | -17.76% | -77.30% | 0.205 | 0.475 | 1197 | 8.68 bps | 6.10 bps | 2.58 bps |
| Fail | short_gate | prior30_dispersion_high | 1.61 bps | -38.31% | -86.99% | 0.113 | 0.385 | 756 | 6.24 bps | 6.10 bps | 0.14 bps |
| Fail | turnover_reduction | rebalance_every_2d | 1.06 bps | -28.67% | -63.68% | 0.089 | 0.406 | 2257 | 5.92 bps | 5.75 bps | 0.17 bps |
| Fail | dynamic_universe | static_top30 | 0.40 bps | -34.19% | -51.23% | 0.036 | 0.528 | 2257 | 6.73 bps | 5.86 bps | 0.87 bps |
| Fail | short_gate | eligible_prior7_negative | 0.35 bps | -46.22% | -90.39% | 0.027 | 0.488 | 1075 | 6.20 bps | 6.10 bps | 0.10 bps |
| Fail | signal_variant | retvol_21d_skip1d | 0.02 bps | -43.73% | -79.24% | 0.002 | 0.501 | 2257 | 6.04 bps | 6.74 bps | -0.70 bps |
| Fail | turnover_reduction | rank_buffer_extra_2 | -0.39 bps | -47.80% | -76.84% | -0.033 | 0.385 | 2257 | 4.23 bps | 3.33 bps | 0.90 bps |
| Fail | turnover_reduction | rebalance_every_3d | -0.74 bps | -52.44% | -69.64% | -0.063 | 0.325 | 2257 | 3.16 bps | 6.86 bps | -3.70 bps |
| Fail | short_gate | eligible_prior7_positive | -1.03 bps | -66.82% | -89.78% | -0.072 | 0.493 | 1177 | 4.88 bps | 6.10 bps | -1.22 bps |
| Fail | short_gate | prior30_dispersion_mid_high | -1.07 bps | -63.31% | -84.60% | -0.079 | 0.495 | 1510 | 4.87 bps | 6.10 bps | -1.23 bps |
| Fail | short_gate | eligible_ew_above_ma20 | -1.21 bps | -68.18% | -87.96% | -0.084 | 0.456 | 1083 | 4.26 bps | 6.10 bps | -1.84 bps |
| Fail | short_gate | btc_prior30_positive | -1.37 bps | -68.65% | -83.56% | -0.096 | 0.469 | 1217 | 4.26 bps | 6.10 bps | -1.84 bps |
| Fail | turnover_reduction | replacement_cap_1_per_side | -1.43 bps | -58.83% | -77.20% | -0.123 | 0.419 | 2257 | 3.59 bps | 4.88 bps | -1.29 bps |
| Fail | short_gate | btc_above_ma50 | -1.44 bps | -69.20% | -83.81% | -0.101 | 0.458 | 1216 | 4.05 bps | 6.10 bps | -2.05 bps |
| Fail | turnover_reduction | rank_buffer_extra_4 | -1.67 bps | -60.02% | -78.05% | -0.147 | 0.292 | 2257 | 1.83 bps | 2.46 bps | -0.64 bps |
| Fail | dynamic_universe | static_top20 | -1.91 bps | -60.93% | -77.61% | -0.174 | 0.478 | 2257 | 3.83 bps | 5.36 bps | -1.53 bps |
| Fail | turnover_reduction | replacement_cap_2_per_side | -2.16 bps | -65.96% | -82.81% | -0.181 | 0.572 | 2257 | 4.71 bps | 5.95 bps | -1.24 bps |
| Fail | short_gate | short_every_day | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | 4.98 bps | 6.10 bps | -1.12 bps |
| Fail | signal_variant | retvol_14d_skip1d | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | 4.98 bps | 6.10 bps | -1.12 bps |
| Fail | dynamic_universe | static_top50 | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | 4.98 bps | 6.10 bps | -1.12 bps |
| Fail | signal_variant | rank_ensemble_7_14_21 | -2.56 bps | -68.90% | -92.10% | -0.216 | 0.635 | 2257 | 5.06 bps | 5.66 bps | -0.60 bps |
Short Leg 条件化
long half 每天保留；short gate OFF 时 short half 空仓。表中 active_short_days 是实际做空天数。
最好的 short gate 是 btc_prior7_positive：active short days 1203，short half 日均 3.31 bps，12bps 后组合日均 3.40 bps，但累计仍为 -7.74%、最大回撤 -72.77%。推论是：short 条件化能减少坏 short，但它没有解决组合路径风险。
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days | gross_mean_bps | long_half_mean_bps | short_half_mean_bps |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Fail | short_gate | btc_prior7_positive | 3.40 bps | -7.74% | -72.77% | 0.237 | 0.501 | 1203 | 9.42 bps | 6.10 bps | 3.31 bps |
| Fail | short_gate | btc_above_ma20 | 2.98 bps | -17.76% | -77.30% | 0.205 | 0.475 | 1197 | 8.68 bps | 6.10 bps | 2.58 bps |
| Fail | short_gate | prior30_dispersion_high | 1.61 bps | -38.31% | -86.99% | 0.113 | 0.385 | 756 | 6.24 bps | 6.10 bps | 0.14 bps |
| Fail | short_gate | eligible_prior7_negative | 0.35 bps | -46.22% | -90.39% | 0.027 | 0.488 | 1075 | 6.20 bps | 6.10 bps | 0.10 bps |
| Fail | short_gate | eligible_prior7_positive | -1.03 bps | -66.82% | -89.78% | -0.072 | 0.493 | 1177 | 4.88 bps | 6.10 bps | -1.22 bps |
| Fail | short_gate | prior30_dispersion_mid_high | -1.07 bps | -63.31% | -84.60% | -0.079 | 0.495 | 1510 | 4.87 bps | 6.10 bps | -1.23 bps |
| Fail | short_gate | eligible_ew_above_ma20 | -1.21 bps | -68.18% | -87.96% | -0.084 | 0.456 | 1083 | 4.26 bps | 6.10 bps | -1.84 bps |
| Fail | short_gate | btc_prior30_positive | -1.37 bps | -68.65% | -83.56% | -0.096 | 0.469 | 1217 | 4.26 bps | 6.10 bps | -1.84 bps |
| Fail | short_gate | btc_above_ma50 | -1.44 bps | -69.20% | -83.81% | -0.101 | 0.458 | 1216 | 4.05 bps | 6.10 bps | -2.05 bps |
| Fail | short_gate | short_every_day | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | 4.98 bps | 6.10 bps | -1.12 bps |
| Fail | short_gate | eligible_ew_above_ma50 | -3.01 bps | -78.45% | -91.09% | -0.211 | 0.423 | 983 | 2.06 bps | 6.10 bps | -4.04 bps |
| Fail | short_gate | btc_prior7_negative | -4.10 bps | -80.70% | -96.43% | -0.313 | 0.487 | 1044 | 1.75 bps | 6.10 bps | -4.35 bps |
| Fail | short_gate | eligible_prior30_positive | -4.40 bps | -84.33% | -90.24% | -0.308 | 0.436 | 1015 | 0.83 bps | 6.10 bps | -5.27 bps |
降换手增厚
这组是本轮唯一有正向可用信息的方向。最强均值来自 rebalance_every_7d，12bps 后 4.56 bps，平均换手 0.200x/day；但它最大回撤 -66.11%。唯一 Watch 的 rank_buffer_extra_8 稍弱但回撤更低，说明“少换一点、让持仓有惯性”比“每天完全追 rank”更接近真实成本下可活的形态。
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days | gross_mean_bps | long_half_mean_bps | short_half_mean_bps |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Fail | turnover_reduction | rebalance_every_7d | 4.56 bps | 68.05% | -66.11% | 0.405 | 0.200 | 2257 | 6.97 bps | 6.29 bps | 0.67 bps |
| Watch | turnover_reduction | rank_buffer_extra_8 | 4.11 bps | 52.78% | -48.88% | 0.368 | 0.210 | 2257 | 6.62 bps | 5.50 bps | 1.13 bps |
| Fail | turnover_reduction | rebalance_every_2d | 1.06 bps | -28.67% | -63.68% | 0.089 | 0.406 | 2257 | 5.92 bps | 5.75 bps | 0.17 bps |
| Fail | turnover_reduction | rank_buffer_extra_2 | -0.39 bps | -47.80% | -76.84% | -0.033 | 0.385 | 2257 | 4.23 bps | 3.33 bps | 0.90 bps |
| Fail | turnover_reduction | rebalance_every_3d | -0.74 bps | -52.44% | -69.64% | -0.063 | 0.325 | 2257 | 3.16 bps | 6.86 bps | -3.70 bps |
| Fail | turnover_reduction | replacement_cap_1_per_side | -1.43 bps | -58.83% | -77.20% | -0.123 | 0.419 | 2257 | 3.59 bps | 4.88 bps | -1.29 bps |
| Fail | turnover_reduction | rank_buffer_extra_4 | -1.67 bps | -60.02% | -78.05% | -0.147 | 0.292 | 2257 | 1.83 bps | 2.46 bps | -0.64 bps |
| Fail | turnover_reduction | replacement_cap_2_per_side | -2.16 bps | -65.96% | -82.81% | -0.181 | 0.572 | 2257 | 4.71 bps | 5.95 bps | -1.24 bps |
信号小范围优化
信号微调没有稳定增厚。最好的 retvol_21d_skip1d 在 12bps 后只有 0.02 bps，其余 7d、ensemble、downside vol、raw return 都更差。推论是：当前问题不是再找一个更锋利的 return-vol 变体，而是成本、换手和 short-leg 结构。
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days | gross_mean_bps | long_half_mean_bps | short_half_mean_bps |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Fail | signal_variant | retvol_21d_skip1d | 0.02 bps | -43.73% | -79.24% | 0.002 | 0.501 | 2257 | 6.04 bps | 6.74 bps | -0.70 bps |
| Fail | signal_variant | retvol_14d_skip1d | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | 4.98 bps | 6.10 bps | -1.12 bps |
| Fail | signal_variant | rank_ensemble_7_14_21 | -2.56 bps | -68.90% | -92.10% | -0.216 | 0.635 | 2257 | 5.06 bps | 5.66 bps | -0.60 bps |
| Fail | signal_variant | downside_vol_14d_skip1d | -3.25 bps | -73.82% | -88.94% | -0.271 | 0.592 | 2257 | 3.85 bps | 5.25 bps | -1.39 bps |
| Fail | signal_variant | retvol_7d_skip1d | -5.97 bps | -85.66% | -93.08% | -0.506 | 0.814 | 2257 | 3.80 bps | 4.10 bps | -0.30 bps |
| Fail | signal_variant | raw_return_14d_skip1d | -8.90 bps | -93.78% | -95.14% | -0.669 | 0.557 | 2257 | -2.22 bps | 1.05 bps | -3.26 bps |
Dynamic Universe
静态 static_top30 是 universe 组里最不差的版本，12bps 后 0.40 bps、最大回撤 -51.23%。Top80 和 dispersion-driven low20/mid50/high80 都更差，说明扩大池子不是免费增厚，容易把小币暴露和 short 噪声一起放大。
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days | gross_mean_bps | long_half_mean_bps | short_half_mean_bps | avg_universe_size | avg_eligible_count |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Fail | dynamic_universe | static_top30 | 0.40 bps | -34.19% | -51.23% | 0.036 | 0.528 | 2257 | 6.73 bps | 5.86 bps | 0.87 bps | 30.000 | 24.933 |
| Fail | dynamic_universe | static_top20 | -1.91 bps | -60.93% | -77.61% | -0.174 | 0.478 | 2257 | 3.83 bps | 5.36 bps | -1.53 bps | 20.000 | 16.875 |
| Fail | dynamic_universe | static_top50 | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 | 4.98 bps | 6.10 bps | -1.12 bps | 50.000 | 40.161 |
| Fail | dynamic_universe | prior30_dispersion_low20_mid50_high80 | -3.13 bps | -73.44% | -88.45% | -0.258 | 0.594 | 2257 | 3.99 bps | 5.72 bps | -1.73 bps | 50.000 | 38.127 |
| Fail | dynamic_universe | static_top80 | -4.16 bps | -81.13% | -94.32% | -0.314 | 0.642 | 2257 | 3.55 bps | 5.01 bps | -1.46 bps | 80.000 | 60.378 |
Funding Sensitivity
没有在当前 monthly-volume age90 Top50 逐日样本里发现可直接对齐每条 short leg 的完整 funding 序列；本节只把 short active day 加 0/+1/+2/+3bps/day 做敏感性，不作为真实 PnL。即使 +3bps/day，12bps 后也只有 0.77 bps，累计仍为 -34.09%，因此 funding 不能被当作这轮 cover 12bps 的主答案。
| variant | cost_bps_per_1x_turnover | funding_credit_bps_per_short_active_day | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | active_short_days |
|---|---|---|---|---|---|---|---|---|
| short_every_day | 4.000 | 0.000 | 2.59 bps | -0.02% | -70.30% | 0.216 | 0.598 | 2257 |
| short_every_day | 8.000 | 0.000 | 0.20 bps | -42.15% | -78.25% | 0.017 | 0.598 | 2257 |
| short_every_day | 12.000 | 0.000 | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 2257 |
| short_every_day | 16.000 | 0.000 | -4.58 bps | -80.64% | -88.43% | -0.382 | 0.598 | 2257 |
| short_every_day | 4.000 | 1.000 | 3.58 bps | 25.30% | -66.40% | 0.299 | 0.598 | 2257 |
| short_every_day | 8.000 | 1.000 | 1.19 bps | -27.49% | -75.40% | 0.099 | 0.598 | 2257 |
| short_every_day | 12.000 | 1.000 | -1.20 bps | -58.05% | -81.99% | -0.101 | 0.598 | 2257 |
| short_every_day | 16.000 | 1.000 | -3.60 bps | -75.73% | -86.81% | -0.300 | 0.598 | 2257 |
| short_every_day | 4.000 | 2.000 | 4.57 bps | 57.02% | -61.99% | 0.381 | 0.598 | 2257 |
| short_every_day | 8.000 | 2.000 | 2.17 bps | -9.12% | -72.17% | 0.181 | 0.598 | 2257 |
| short_every_day | 12.000 | 2.000 | -0.22 bps | -47.42% | -79.62% | -0.018 | 0.598 | 2257 |
| short_every_day | 16.000 | 2.000 | -2.61 bps | -69.58% | -85.08% | -0.218 | 0.598 | 2257 |
| short_every_day | 4.000 | 3.000 | 5.55 bps | 96.78% | -57.00% | 0.463 | 0.598 | 2257 |
| short_every_day | 8.000 | 3.000 | 3.16 bps | 13.89% | -68.51% | 0.264 | 0.598 | 2257 |
| short_every_day | 12.000 | 3.000 | 0.77 bps | -34.09% | -76.95% | 0.064 | 0.598 | 2257 |
| short_every_day | 16.000 | 3.000 | -1.62 bps | -61.87% | -83.12% | -0.135 | 0.598 | 2257 |
推论与操作建议
- 不要把 213c 直接放大。 换手成本口径下，基准 12bps 后已经转负；唯一 Watch 也有接近 50% 的历史回撤。
- 下一轮若继续研究，只围绕低换手结构。 rank buffer、持仓 carry、rebalance 7d 是这轮唯一能同时保留正均值和降低成本的方向。
- short leg 不应简单每日满配。 BTC prior7 positive / BTC MA20 这类 gate 有改善均值的迹象，但路径仍差；更合理的是把 short 作为风控/机会模块，而不是强制 0.5 capital 每天开。
- dispersion 继续做监控，不做主开关。 prior30 dispersion high 只把 short active days 降到 756，12bps 后仍累计为负；它解释收益环境，但不足以因果增厚。
- 实盘 canary 的任务更明确。 重点收集真实 maker fill、taker fallback、exit slippage、funding 和 live-vs-shadow drift，用真实成本替换本页的 12bps 假设。
产物
reports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213_age90_14d_fifth_round_profit_thickness_results.csvreports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213_age90_14d_fifth_round_profit_thickness_daily.csvreports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213_age90_14d_fifth_round_profit_thickness_candidates.csvreports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213_age90_14d_fifth_round_profit_thickness_summary.json