Every factor test and strategy backtest from our low-frequency quant program on China A-shares — published with the verdict, including the failures. Numbers come from backtest engine output, never hand-calculated.低频量化项目中每一个因子测试和策略回测——带着裁决一起公开，包括失败的。所有数字来自回测引擎输出，绝非手工计算。
Monthly-rebalanced cross-sectional tests, 2016–2026. RankIC = Spearman(factor, next-month return). Rule: ICIR > 0.3 with no out-of-sample reversal → candidate for strategy use.月度调仓截面检验，2016–2026。RankIC = Spearman（因子值，下月收益）。规则：ICIR > 0.3 且样本外不反转 → 进入策略候选。
| Factor因子 | RankIC | t | ICIR | Verdict结论 |
|---|---|---|---|---|
| Size proxy (1 / turnover)规模代理（1/成交额） | 0.068 | 4.52 | 0.443 | Valid → S005有效 → S005 |
| Short-term reversal (1M)短期反转（1M） | 0.0475 | 3.07 | 0.285 | Factor valid, strategy failed on turnover因子有效，策略化失败（换手） |
| Momentum (12-1M)动量（12-1M） | -0.0177 | -0.91 | -0.089 | Eliminated (inverted in A-shares)淘汰（A股呈反向） |
| Value (pure PE/PB)价值（纯 PE/PB） | — | — | — | Eliminated (S001)淘汰（S001） |
| Quality (pure ROE)质量（纯 ROE） | weakly negative弱负向 | — | — | Eliminated; composite version worth testing淘汰；合成版值得测 |
| Growth (OCF growth)成长（OCF 增速） | IR 0.9793 | p=1.04% | — | Valid standalone (+1.33pp in S007)独立有效（S007 内 +1.33pp） |
| Dividend (TTM yield + continuity)红利（TTM 股息率 + 连续性） | — | — | — | Valid → S009 (13.98% ann.)有效 → S009（年化 13.98%） |
Full backtests, ¥500k initial capital, real-world trading costs, T+1 and price-limit constraints. Eliminated strategies stay on the board — the graveyard is data too.完整回测，50万初始资金，实盘费率，T+1 与涨跌停约束。被淘汰的策略也留在榜上——墓园也是数据。
| Strategy策略 | Ann.年化 | Max DD最大回撤 | Turnover/yr年换手 | Sharpe | Excess超额 | Verdict裁决 | Failed on触发证伪 |
|---|---|---|---|---|---|---|---|
| S009 Pure dividendS009 纯红利单因子 | 13.98% | -24.65% | 62% | 0.94 | +8.50pp | PASS通过 | — |
| S005b Small-cap + quality (true mkt-cap)S005b 小市值+质量（真实市值版） | 12.32% | -25.34% | 52% | 0.83 | +6.83pp | PASS通过 | — |
| S005 Small-cap + qualityS005 小市值+质量复合 | 11.51% | -21.65% | 50% | 0.94 | +6.02pp | PASS通过 | — |
| S008 Dividend + qualityS008 红利+质量复合 | 10.74% | -22.41% | 49% | 0.77 | +5.26pp | OUT淘汰 | F4, F5 |
| S006 Short-term reversalS006 短期反转 | 7.44% | -24.90% | 449% | 0.54 | +1.95pp | OUT淘汰 | F3 |
| S004 Quality + trend filterS004 质量+趋势过滤 | 5.39% | -24.79% | 81% | 0.48 | -0.10pp | OUT淘汰 | F1,F4,F6,F7,F8 |
| S007 GARPS007 GARP成长估值复合 | 4.48% | -23.65% | 71% | 0.44 | -1.01pp | OUT淘汰 | F1,F4,F7,F8 |
| S002 Quality + low-volS002 质量+低波动复合 | 1.82% | -21.88% | 41% | 0.24 | -3.67pp | OUT淘汰 | F1,F4,F5,F8 |
| S001 Value + momentumS001 低估值动量双因子 | — | -31.4% | 161% | — | — | OUT淘汰 | F1, F5 |
| Benchmark equal-weight基准等权 | 5.48% | -16.98% | 2% | 0.65 | — | REF基准 | — |
Last updated: 2026-10-10 · S010 (momentum control + low-frequency reversal) queued for testing最后更新：2026-10-10 · S010（动量对照 + 降频反转）已排队待测
Fixed before any backtest. A strategy is eliminated if it triggers any single one.全部在回测前写死。触发任意一条即淘汰。
Excess return: full-period annualized return below the equal-weight benchmark.超额收益：全期年化低于等权基准。
Drawdown: maximum drawdown deeper than −35%.回撤：最大回撤超过 −35%。
Turnover: annual turnover above 300%.换手：年换手率超过 300%。
Persistence: underperforms the benchmark in any 3 consecutive years.持续性：任意连续 3 年跑输基准。
Incrementality: removing any single factor improves returns — the composite adds nothing.增量性：去掉任一因子后收益反而更高——复合没有增量。
Overfitting: out-of-sample Sharpe / in-sample Sharpe < 0.3.过拟合：样本外夏普 / 样本内夏普 < 0.3。
Opportunity cost: longest underwater period over 3 years.机会成本：最长水下期超过 3 年。
Cost coverage: gross alpha / total two-way costs < 1.5×.成本覆盖：毛 alpha / 双边总成本 < 1.5 倍。
Data snooping: deflated Sharpe ratio (trial count N) below 0.95.数据窥探：经试验次数 N 调整的 DSR 低于 0.95。
Research publication only. Nothing on this page is investment advice, a recommendation, or an offer. Backtested performance does not predict future results.仅为研究发表。本页任何内容都不构成投资建议、推荐或要约。回测表现不预示未来收益。