Scoreboard
Every BUY signal the scanner emitted, scored on what price did afterwards. No look-ahead — a forward return only reads snapshots recorded after the call.
Forward return by horizon
| horizon | cov | n | median | win% |
|---|---|---|---|---|
| 15m | 100% | 448 | -3% | 27% |
| 1h | 97% | 436 | -4% | 33% |
| 6h | 83% | 372 | -10% | 33% |
| 24h | 47% | 210 | -18% | 30% |
Median = the middle outcome across all calls held to that horizon, after the cost haircut. The average is left out here because a few price-feed glitches distort it. Coverage below 100% = some tokens stopped being tracked before then.
Exit-rule simulation
| win rate | 33% |
| avg win / loss | +19% / -14% |
| expectancy / call | -3% |
| 95% confidence | -5% … -2% |
| profit factor | 0.66 |
| median hold | 139.1 min |
Each call walked forward snapshot by snapshot; first of +30% / −15% / timeout is the outcome.
Does a higher score pay?
| opp band | n | +1h | +6h | exit | win% |
|---|---|---|---|---|---|
| 75+ | 0 | — | — | — | — |
| 65–74 | 1 | +8% | -10% | -18% | 0% |
| 55–64 | 447 | -4% | -10% | -10% | 33% |
| 45–54 | 0 | — | — | — | — |
Every scored call grouped by its opportunity score at signal time. If the score carries predictive weight, the higher bands should read better than the lower ones.
Best calls
The six highest peaks among winning calls older than 24 hours. 148 of 448 calls hit the profit target before the stop; every other call, losers included, is counted in the numbers above. Peak is the highest price reached after the call, not what the exit rule banks. Peaks above +1,000% are left out as probable price-feed errors.
B
E
Z
B
M
C