Does this thing actually work?
Realized forward returns of past candidates, bucketed by composite-score decile (10 = strongest signal). Computed from real cached bars — never modelled.
| Composite decile | Candidates | avg +5d % | avg +20d % |
|---|---|---|---|
| 10 | 2383 | +0.78 | +3.13 |
| 9 | 2237 | -0.13 | +1.67 |
| 8 | 2076 | +0.16 | +2.77 |
| 7 | 2130 | +0.07 | +1.02 |
| 6 | 2066 | +0.47 | +0.57 |
| 5 | 2181 | +0.49 | +1.24 |
| 4 | 2219 | +0.44 | +0.76 |
| 3 | 2194 | +0.62 | +0.71 |
| 2 | 2236 | -0.16 | -0.48 |
| 1 | 2136 | +0.41 | -0.71 |
If the top deciles don't outperform the bottom, the signal isn't working — that is the point of this page.