Rank213c 试探性拓展研究
目标：基于 213c 的 causal monthly-volume Top50 / age90 / ret-vol 选币框架，尝试寻找下一步可能优化收益的方向。这里是探索性研究，不是上线参数选择。
生成时间：2026-05-07 05:35:42 UTC。成本统一用 cost_bps_per_1x_turnover，主读法看 12bps。
一句话结论
试探性拓展里最值得继续的是“低换手 + 降低 short capital”，而不是继续扩大 universe 或换信号。 基准 base_50_50_daily 在 12bps 后为 -2.19 bps、累计 -66.53%；最好候选 rank_buffer8_50_50 为 4.11 bps、累计 52.78%、换手 0.210x/day。
推论：213c 的 long 排序仍有价值，但 short leg 更像需要降权/条件化的风险模块。若继续推进，应把 213c2/213d 的研究收窄到“低换手持仓惯性 + 0.25 short capital + 明确风险开关”。
给新研究者的读法
- 本页是方向筛选，不是上线报告。 Promising 表示值得进入下一轮验证，不表示可以直接替换 213c 实盘参数。
- 主读法看 12bps。 这个口径比 flat 4bps 更接近实盘摩擦，因此比漂亮的低成本曲线更重要。
- 先比较基准和最好候选。 如果候选只是牺牲回撤或减少交易天数换来均值，不能简单视为改进。
- 本页最重要的推论：213c 的 alpha 主要还在 long 排序，short half 应该更像风险模块，下一轮应围绕低换手、short 降权和明确风控开关做小样本深挖。
核心读数
-2.19 bps基准 12bps 日均
4.11 bps最好候选 12bps 日均
0.210x最好候选换手
1 / 4Promising / Watch
12bps Shortlist
| verdict | experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | sharpe | avg_turnover_x | gross_mean_bps | long_mean_bps | short_mean_bps | active_short_days | reason |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Promising | low_turnover | rank_buffer8_50_50 | 4.11 bps | 52.78% | -48.88% | 0.368 | 0.210 | 6.62 bps | 5.50 bps | 1.13 bps | 2257 | positive 12bps net with lower turnover; needs robustness follow-up |
| Watch | low_turnover | weekly_buffer8_long75_short25 | 9.92 bps | 168.50% | -85.72% | 0.564 | 0.129 | 11.47 bps | 10.91 bps | 0.55 bps | 2257 | positive 12bps net but drawdown/turnover still not clean |
| Watch | low_turnover | weekly_long75_short25 | 7.46 bps | 39.51% | -94.60% | 0.409 | 0.193 | 9.78 bps | 9.44 bps | 0.34 bps | 2257 | positive 12bps net but drawdown/turnover still not clean |
| Watch | low_turnover | rank_buffer8_long75_short25 | 6.38 bps | 7.79% | -93.31% | 0.348 | 0.202 | 8.81 bps | 8.24 bps | 0.56 bps | 2257 | positive 12bps net but drawdown/turnover still not clean |
| Watch | low_turnover | weekly_50_50 | 4.56 bps | 68.05% | -66.11% | 0.405 | 0.200 | 6.97 bps | 6.29 bps | 0.67 bps | 2257 | positive 12bps net but drawdown/turnover still not clean |
| Fail | capital_mix | long100_no_short | 4.94 bps | -94.12% | -99.83% | 0.161 | 0.543 | 11.45 bps | 11.45 bps | 0.00 bps | 0 | does not preserve positive 12bps net |
| Fail | short_optional | long75_short25_btc_prior7_positive | 4.57 bps | -59.03% | -97.19% | 0.212 | 0.520 | 10.81 bps | 9.15 bps | 1.66 bps | 1203 | does not preserve positive 12bps net |
| Fail | short_optional | long75_short25_btc_ma20 | 4.36 bps | -61.38% | -97.43% | 0.202 | 0.507 | 10.44 bps | 9.15 bps | 1.29 bps | 1197 | does not preserve positive 12bps net |
| Fail | leg_shape | top6_long_only | 3.53 bps | -92.36% | -99.77% | 0.125 | 0.490 | 9.41 bps | 9.41 bps | 0.00 bps | 0 | does not preserve positive 12bps net |
| Fail | risk_guard | weekly_long75_short25_avoid_prior7_crash | 3.20 bps | -24.72% | -92.29% | 0.204 | 0.223 | 5.88 bps | 3.54 bps | 2.34 bps | 1884 | does not preserve positive 12bps net |
| Fail | short_optional | long50_short25_dispersion_high | 2.24 bps | -27.01% | -89.80% | 0.159 | 0.327 | 6.17 bps | 6.10 bps | 0.07 bps | 756 | does not preserve positive 12bps net |
| Fail | capital_mix | long75_short25_daily | 1.77 bps | -66.50% | -97.34% | 0.093 | 0.568 | 8.59 bps | 9.15 bps | -0.56 bps | 2257 | does not preserve positive 12bps net |
| Fail | leg_shape | top3_long_short3_75_25 | 1.74 bps | -74.92% | -98.23% | 0.084 | 0.610 | 9.06 bps | 9.27 bps | -0.21 bps | 2257 | does not preserve positive 12bps net |
| Fail | risk_guard | base_avoid_eligible_prior7_crash | 0.86 bps | -26.80% | -75.37% | 0.078 | 0.557 | 7.55 bps | 3.42 bps | 4.13 bps | 1884 | does not preserve positive 12bps net |
| Fail | risk_guard | long75_short25_avoid_prior7_crash | 0.74 bps | -62.53% | -96.44% | 0.044 | 0.537 | 7.19 bps | 5.13 bps | 2.06 bps | 1884 | does not preserve positive 12bps net |
| Fail | capital_mix | long50_short25 | 0.34 bps | -39.70% | -87.38% | 0.028 | 0.434 | 5.54 bps | 6.10 bps | -0.56 bps | 2257 | does not preserve positive 12bps net |
| Fail | leg_shape | top6_long_short2_75_25 | -0.53 bps | -74.13% | -97.49% | -0.031 | 0.571 | 6.32 bps | 7.04 bps | -0.71 bps | 2257 | does not preserve positive 12bps net |
| Fail | capital_mix | base_50_50_daily | -2.19 bps | -66.53% | -84.08% | -0.183 | 0.598 | 4.98 bps | 6.10 bps | -1.12 bps | 2257 | does not preserve positive 12bps net |
分组最好结果
| experiment_group | variant | net_mean_bps | net_cum_pct | max_drawdown_pct | avg_turnover_x | long_capital | short_capital | short_gate | market_guard |
|---|---|---|---|---|---|---|---|---|---|
| capital_mix | long100_no_short | 4.94 bps | -94.12% | -99.83% | 0.543 | 1.000 | 0.000 | never | none |
| capital_mix | long75_short25_daily | 1.77 bps | -66.50% | -97.34% | 0.568 | 0.750 | 0.250 | always | none |
| capital_mix | long50_short25 | 0.34 bps | -39.70% | -87.38% | 0.434 | 0.500 | 0.250 | always | none |
| capital_mix | base_50_50_daily | -2.19 bps | -66.53% | -84.08% | 0.598 | 0.500 | 0.500 | always | none |
| leg_shape | top6_long_only | 3.53 bps | -92.36% | -99.77% | 0.490 | 1.000 | 0.000 | never | none |
| leg_shape | top3_long_short3_75_25 | 1.74 bps | -74.92% | -98.23% | 0.610 | 0.750 | 0.250 | always | none |
| leg_shape | top6_long_short2_75_25 | -0.53 bps | -74.13% | -97.49% | 0.571 | 0.750 | 0.250 | always | none |
| low_turnover | weekly_buffer8_long75_short25 | 9.92 bps | 168.50% | -85.72% | 0.129 | 0.750 | 0.250 | always | none |
| low_turnover | weekly_long75_short25 | 7.46 bps | 39.51% | -94.60% | 0.193 | 0.750 | 0.250 | always | none |
| low_turnover | rank_buffer8_long75_short25 | 6.38 bps | 7.79% | -93.31% | 0.202 | 0.750 | 0.250 | always | none |
| low_turnover | weekly_50_50 | 4.56 bps | 68.05% | -66.11% | 0.200 | 0.500 | 0.500 | always | none |
| risk_guard | weekly_long75_short25_avoid_prior7_crash | 3.20 bps | -24.72% | -92.29% | 0.223 | 0.750 | 0.250 | always | avoid_eligible_prior7_crash |
| risk_guard | base_avoid_eligible_prior7_crash | 0.86 bps | -26.80% | -75.37% | 0.557 | 0.500 | 0.500 | always | avoid_eligible_prior7_crash |
| risk_guard | long75_short25_avoid_prior7_crash | 0.74 bps | -62.53% | -96.44% | 0.537 | 0.750 | 0.250 | always | avoid_eligible_prior7_crash |
| short_optional | long75_short25_btc_prior7_positive | 4.57 bps | -59.03% | -97.19% | 0.520 | 0.750 | 0.250 | btc_prior7_positive | none |
| short_optional | long75_short25_btc_ma20 | 4.36 bps | -61.38% | -97.43% | 0.507 | 0.750 | 0.250 | btc_above_ma20 | none |
| short_optional | long50_short25_dispersion_high | 2.24 bps | -27.01% | -89.80% | 0.327 | 0.500 | 0.250 | prior30_dispersion_high | none |
推论讨论
- 资本配比是第一优先级。 long-only 或 0.25 short capital 往往比 0.5/0.5 market-neutral 更接近保留 alpha，同时减少 short 噪声。
- 低换手是第二优先级。 rank buffer 和 weekly rebalance 不是单纯省成本，也在减少每日 rank 抖动带来的误换仓。
- 风险开关只能做 guard。 eligible/BTC crash guard 对回撤有帮助，但容易牺牲复利路径，不能独立构成收益来源。
- 下一轮不要做大参数网格。 更合理的是选 2-3 个候选做 walk-forward、年度稳定性、live-vs-shadow 成本复核。
产物
reports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213c_exploratory_expansion_results.csvreports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213c_exploratory_expansion_daily.csvreports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213c_exploratory_expansion_shortlist.csvreports/artifacts/paper_rank213_largecap_xs_jump_veto/rank213c_exploratory_expansion_summary.json