what drives 4+ cards? (refs + closeness)
← lab · AUC 0.547 (weak signal) · ran 7/3/2026
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? |
|---|
| home__team.conversion_l5 | 0.0039 | ↑ +0.012 | no |
| away__team.cards_avg_l5 | 0.0030 | ↑ +0.067 | no |
| home__team.state_index | 0.0029 | ↓ -0.003 | no |
| away__team.congestion_21d | 0.0026 | ↓ -0.030 | ⏳ yes |
| home__team.corners_for_avg_l5 | 0.0007 | ↓ -0.011 | no |
| away__team.shot_diff_avg_l5 | 0.0006 | ↑ +0.041 | no |
| away__team.corner_diff_avg_l5 | 0.0001 | ↑ +0.027 | no |
| away__team.matches_since_blank | 0.0000 | ↑ +0.037 | ⏳ yes |
| match.mismatch | 0.0000 | ↓ -0.081 | ⏳ yes |
| home__team.matches_since_win | -0.0002 | ↑ +0.006 | no |
| match.form_gap | -0.0002 | ↓ -0.019 | no |
| match.ref_cards_avg | -0.0002 | ↑ +0.150 | ⏳ yes |
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": "what drives 4+ cards? (refs + closeness)",
"sport": "football",
"target": {
"op": ">=",
"value": 4,
"metric": "match.total_cards"
},
"features": "all"
}