Solorank.ai
Signal Lab
Do these short-term signals actually make money?
Fixed rules, next-day open entry, costs deducted, non-overlapping periods, judged on the latest third of the data (out-of-sample) against SPY. In-sample "best" numbers are not shown as proof.
Lv3 track sorting study · Track record monitor
Point-in-time constituent history is incomplete, so this run uses today's constituents for the full history. Results carry survivorship bias: stocks removed from the index in the past are missing.
| Signal | Hold | Periods | Mean / period | Excess vs SPY (out-of-sample) | t-stat | Win rate | Beat SPY | Max drawdown | Verdict |
|---|---|---|---|---|---|---|---|---|---|
| 新：12-1 動量最高 10 隻 New: 12-1 momentum Top 10 |
1D | 563 | +0.02% | -0.03% | -0.15 | 48.9% | 50.5% | -35.21% | No edge |
| 新：12-1 動量最高 10 隻 New: 12-1 momentum Top 10 |
3D | 191 | +0.69% | +0.64% | 1.09 | 59.4% | 53.1% | -25.83% | Unproven |
| 新：12-1 動量最高 10 隻 New: 12-1 momentum Top 10 |
5D | 115 | +0.90% | +0.51% | 0.49 | 61.5% | 61.5% | -40.43% | Unproven |
| 新：12-1 動量最高 10 隻 New: 12-1 momentum Top 10 |
10D | 57 | +2.81% | +2.10% | 1.09 | 52.6% | 63.2% | -30.52% | Unproven |
| 新：強勢股回調 New: momentum pullback |
1D | 551 | -0.04% | -0.04% | -0.34 | 46.2% | 47.3% | -31.89% | No edge |
| 新：強勢股回調 New: momentum pullback |
3D | 187 | +0.43% | +0.44% | 1.10 | 54.0% | 54.0% | -24.35% | Unproven |
| 新：強勢股回調 New: momentum pullback |
5D | 113 | +0.44% | +0.31% | 0.46 | 63.2% | 65.8% | -30.21% | Unproven |
| 新：強勢股回調 New: momentum pullback |
10D | 56 | +0.84% | +2.08% | 1.65 | 79.0% | 57.9% | -28.20% | Unproven |
| 現有：五日動能最高 10 隻 Existing: 5D momentum Top 10 |
1D | 87 | -0.26% | -0.16% | -0.87 | 44.8% | 48.3% | -21.95% | No edge |
| 現有：五日動能最高 10 隻 Existing: 5D momentum Top 10 |
3D | 29 | -0.16% | +0.60% | 0.95 | 50.0% | 60.0% | -13.13% | Unproven |
| 現有：五日動能最高 10 隻 Existing: 5D momentum Top 10 |
5D | 17 | -1.05% | -0.62% | -0.60 | 33.3% | 50.0% | -19.83% | Not enough data |
| 現有：五日動能最高 10 隻 Existing: 5D momentum Top 10 |
10D | 8 | -0.62% | +0.46% | 0.34 | 66.7% | 66.7% | -9.69% | Not enough data |
| 現有：投影中位數最高 10 隻 Existing: projection median Top 10 |
1D | 197 | +0.02% | -0.12% | -0.84 | 42.4% | 40.9% | -20.13% | No edge |
| 現有：投影中位數最高 10 隻 Existing: projection median Top 10 |
3D | 66 | +0.57% | -0.33% | -0.75 | 54.5% | 50.0% | -12.99% | No edge |
| 現有：投影中位數最高 10 隻 Existing: projection median Top 10 |
5D | 40 | +1.23% | +0.52% | 0.63 | 64.3% | 57.1% | -14.04% | Unproven |
| 現有：投影中位數最高 10 隻 Existing: projection median Top 10 |
10D | 20 | +2.04% | +0.52% | 0.36 | 57.1% | 57.1% | -8.91% | Not enough data |
| 新：上升趨勢中短期反轉 New: short-term reversal in uptrend |
1D | 615 | -0.05% | -0.03% | -0.29 | 48.8% | 50.2% | -41.49% | No edge |
| 新：上升趨勢中短期反轉 New: short-term reversal in uptrend |
3D | 209 | +0.30% | +0.26% | 0.83 | 44.3% | 50.0% | -24.37% | Unproven |
| 新：上升趨勢中短期反轉 New: short-term reversal in uptrend |
5D | 126 | +0.41% | +0.08% | 0.14 | 61.9% | 57.1% | -26.14% | Unproven |
| 新：上升趨勢中短期反轉 New: short-term reversal in uptrend |
10D | 63 | +1.10% | +0.10% | 0.10 | 52.4% | 47.6% | -23.87% | Unproven |
| 現有：總分最高 10 隻 Existing: total score Top 10 |
1D | 87 | -0.17% | -0.01% | -0.04 | 48.3% | 58.6% | -18.15% | No edge |
| 現有：總分最高 10 隻 Existing: total score Top 10 |
3D | 29 | +0.20% | +0.86% | 1.16 | 60.0% | 60.0% | -6.76% | Unproven |
| 現有：總分最高 10 隻 Existing: total score Top 10 |
5D | 17 | -1.06% | -0.06% | -0.04 | 50.0% | 50.0% | -21.81% | Not enough data |
| 現有：總分最高 10 隻 Existing: total score Top 10 |
10D | 8 | -0.25% | +0.72% | 1.01 | 66.7% | 66.7% | -7.80% | Not enough data |
Mean, drawdown and periods use the full sample; excess, t-stat, win rate and beat-SPY use only the out-of-sample third. "Evidence" requires excess > 0 with t > 2 out-of-sample.
Latest candidates (evidence-backed signals only)
No signal currently clears the out-of-sample evidence bar, so no candidate list is shown. That is the point of this page.
Caveats
- Existing-signal history uses score snapshots stored in the database; fundamentals and news inside those scores were not point-in-time, so existing-signal results are optimistic.
- Backfilled score snapshots (source 'database-backfill') are excluded because they were recomputed with later enrichment and today's constituents.
- Candidates and evidenced signals are suppressed while point-in-time membership is unreliable (survivorship bias).
- Signals are fixed rules chosen before this backtest; no parameter search was performed.
- Returns use next-day open entry, close exit, equal weights, and a flat round-trip cost.
2026-10-27T03:01:58.070Z