sweep: match.comeback_happened (bool)
← lab · AUC 0.494 (no real pattern — honest result) · ran 7/5/2026
What this is: Asks which pre-match factors drive one specific outcome, using a walk-forward model and permutation importance.
| Factor | Importance | Direction | Survives all eras? |
|---|
| away__t3__team.elo_momentum_l5 | 0.0001 | ↑ +0.003 | no |
| away__team.blown_lead_rate_l20 | 0.0000 | ↑ +0.003 | no |
| away__team.comeback_rate_l20 | 0.0000 | ↓ -0.002 | no |
| away__team.cards_avg_l5 | 0.0000 | ↑ +0.012 | no |
| away__team.conversion_l5 | 0.0000 | ↓ -0.016 | no |
| away__team.corner_diff_avg_l5 | 0.0000 | ↑ +0.007 | no |
| away__team.corners_against_avg_l5 | 0.0000 | ↓ -0.008 | no |
| away__team.congestion_21d | 0.0000 | ↑ +0.008 | no |
| away__team.corners_volatility_l10 | 0.0000 | ↑ +0.004 | no |
| away__team.defensive_leak_l5 | 0.0000 | ↑ +0.004 | no |
| away__team.dominance_l5 | 0.0000 | ↑ +0.013 | no |
| away__team.elo | 0.0000 | ↑ +0.009 | no |
Reading the columnswhat each number actually means
| AUC | predictability: 0.50 = coin flip, ~0.70 = ceiling for sports |
| Importance | how much the model leans on this factor (permutation importance) |
| Direction | sign of the raw correlation with the outcome |
| Survives all eras | effect points the same way in every historical era |
Spec · the reproducible recipe
{
"name": "sweep: match.comeback_happened (bool)",
"sport": "football",
"target": {
"metric": "match.comeback_happened"
},
"features": "all"
}