Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股Alpha衰减/因子拥挤度分析。当用户说"Alpha衰减"、"alpha decay"、"因子拥挤"、"策略容量"、"因子失效"、"XX因子还有效吗"、"crowding"、"策略拥挤"时触发。分析因子/策略的Alpha随时间的衰减趋势、因子拥挤度指标、策略容量限制。支持研报风格（formal）和快速分析风格（brief）。
.claude/skills/aifinlab-a-share-alpha-decay/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -38% | 0% |
bashSCRIPTS="$SKILLS_ROOT/cn-stock-data/scripts" python "$SCRIPTS/cn_stock_data.py" kline --code [CODE] --freq daily --start [日期] python "$SCRIPTS/cn_stock_data.py" finance --code [CODE] python "$SCRIPTS/cn_stock_data.py" quote --code [CODE1],[CODE2],...
确定要分析 Alpha 衰减的目标：特定因子（如低 PE）或策略（如动量）
| 维度 | formal | brief | |------|--------|-------| | 衰减分析 | 完整时序+断点检验 | 当前 IC vs 历史均值 | | 拥挤度 | 多维指标矩阵 | 拥挤/正常/低估 | | 容量 | 详细估算 | 大/中/小 |
默认风格：brief。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 20,214 | 13,497 | -33% | 1 | 1 | 0% | 2,871 | 2,561 | -11% | 0 | 0 | — |
case-02 | fail→fail | 15,378 | 14,727 | -4% | 1 | 1 | 0% | 2,523 | 2,796 | +11% | 0 | 0 | — |
case-17 | fail→pass | 11,573 | 1,766 | -85% | 1 | 1 | 0% | 1,637 | 811 | -50% | 0 | 0 | — |
case-03 | fail→pass | 33,906 | 36,618 | +8% | 1 | 1 | 0% | 5,307 | 6,671 | +26% | 0 | 0 | — |
case-04 | pass→pass | 21,280 | 16,522 | -22% | 1 | 1 | 0% | 3,195 | 3,274 | +2% | 0 | 0 | — |
case-05 | fail→pass | 18,375 | 16,002 | -13% | 1 | 1 | 0% | 2,983 | 2,869 | -4% | 0 | 0 | — |
case-06 | pass→pass | 26,775 | 19,588 | -27% | 1 | 1 | 0% | 3,852 | 3,728 | -3% | 0 | 0 | — |
case-07 | pass→pass | 18,924 | 15,763 | -17% | 1 | 1 | 0% | 2,436 | 3,089 | +27% | 0 | 0 | — |
case-08 | pass→pass | 16,845 | 14,395 | -15% | 1 | 1 | 0% | 2,215 | 2,469 | +11% | 0 | 0 | — |
case-09 | pass→pass | 16,547 | 16,501 | -0% | 1 | 1 | 0% | 2,674 | 2,947 | +10% | 0 | 0 | — |
case-10 | fail→pass | 15,678 | 5,113 | -67% | 1 | 1 | 0% | 1,935 | 1,251 | -35% | 0 | 0 | — |
case-11 | pass→pass | 19,201 | 13,947 | -27% | 1 | 1 | 0% | 2,893 | 2,381 | -18% | 0 | 0 | — |
case-12 | pass→pass | 13,085 | 11,374 | -13% | 1 | 1 | 0% | 1,984 | 2,072 | +4% | 0 | 0 | — |
case-13 | fail→pass | 10,727 | 4,089 | -62% | 1 | 1 | 0% | 1,834 | 1,145 | -38% | 0 | 0 | — |
case-14 | fail→pass | 15,270 | 2,677 | -82% | 1 | 1 | 0% | 2,255 | 985 | -56% | 0 | 0 | — |
case-15 | pass→pass | 19,689 | 4,593 | -77% | 1 | 1 | 0% | 2,880 | 1,209 | -58% | 0 | 0 | — |
case-16 | pass→pass | 25,309 | 18,688 | -26% | 1 | 1 | 0% | 3,106 | 3,153 | +2% | 0 | 0 | — |
case-18 | fail→pass | 14,447 | 7,709 | -47% | 1 | 1 | 0% | 2,298 | 1,771 | -23% | 0 | 0 | — |
case-19 | fail→pass | 9,015 | 2,274 | -75% | 1 | 1 | 0% | 1,441 | 854 | -41% | 0 | 0 | — |
case-20 | pass→fail | 16,527 | 7,675 | -54% | 1 | 1 | 0% | 3,488 | 987 | -72% | 0 | 0 | — |
case-21 | pass→fail | 24,469 | 7,846 | -68% | 1 | 1 | 0% | 4,341 | 1,101 | -75% | 0 | 0 | — |
case-22 | pass→pass | 21,460 | 22,610 | +5% | 1 | 1 | 0% | 2,659 | 4,223 | +59% | 0 | 0 | — |
case-23 | fail→fail | 11,919 | 5,739 | -52% | 1 | 1 | 0% | 1,613 | 1,425 | -12% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 21 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +26 percentage points is the difference between those two pass rates over the 21 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/28/2026 | +36% |
Other measured skills in the registry, with their headline benchmark lift.