lead-lag: which trends move first (football)
← lab · sport football · n metrics 31 · n team contests 104,910 · ran 7/4/2026
What this is: A one-off study — its numbers below are exactly what the run reported.
| note | within-competitor first differences; dX_t vs dY_(t+1); rolling-window overlap inflates same-family pairs — judge cross-family pairs only |
| sport | football |
| n metrics | 31 |
| n team contests | 104,910 |
top lead lag
| n | r | leads | follows |
|---|
| 53,336 | 0.505 | team.congestion_21d | team.rest_days |
| 51,040 | -0.2328 | team.venue_ppg_l5 | team.state_index |
| 53,336 | 0.1791 | team.elo | team.matches_since_win |
| 51,173 | -0.167 | team.state_index | team.venue_ppg_l5 |
| 52,250 | 0.1407 | team.form_points_l5 | team.matches_since_win |
| 52,836 | 0.1372 | team.state_index | team.matches_since_win |
| 51,002 | -0.1226 | team.elo | team.season_ppg |
| 52,250 | 0.1105 | team.goal_diff_avg_l5 | team.matches_since_win |
| 52,250 | 0.1065 | team.elo_momentum_l5 | team.matches_since_win |
| 23,369 | 0.0969 | team.h2h_win_rate_l5 | team.shot_diff_avg_l5 |
| 51,493 | 0.0965 | team.elo | team.venue_ppg_l5 |
| 53,111 | -0.0929 | team.rest_days | team.congestion_21d |
| 52,250 | 0.0909 | team.goals_for_avg_l5 | team.matches_since_win |
| 23,369 | 0.0903 | team.h2h_win_rate_l5 | team.dominance_l5 |
| 51,493 | 0.0863 | team.form_points_l5 | team.venue_ppg_l5 |
| 53,336 | 0.0857 | team.elo | team.matches_since_clean_sheet |
| 51,493 | -0.0845 | team.matches_since_win | team.venue_ppg_l5 |
| 50,777 | 0.0844 | team.season_ppg | team.matches_since_win |
| 53,336 | -0.0843 | team.elo | team.matches_since_blank |
| 52,836 | -0.0836 | team.state_index | team.rest_days |
| 49,926 | -0.0814 | team.form_points_l5 | team.season_ppg |
| 51,341 | 0.0813 | team.h2_goals_l10 | team.matches_since_win |
| 52,250 | -0.0795 | team.goals_against_avg_l5 | team.matches_since_clean_sheet |
| 51,493 | 0.079 | team.goal_diff_avg_l5 | team.venue_ppg_l5 |
| 49,742 | 0.079 | team.ht_lead_rate_l20 | team.matches_since_win |
| 50,607 | 0.078 | team.venue_ppg_l5 | team.shot_diff_avg_l5 |
| 52,250 | -0.0773 | team.goals_for_avg_l5 | team.matches_since_blank |
| 50,607 | 0.0763 | team.venue_ppg_l5 | team.dominance_l5 |
| 23,369 | 0.0747 | team.h2h_win_rate_l5 | team.shots_for_avg_l5 |
| 23,420 | 0.0746 | team.h2h_win_rate_l5 | team.goal_diff_avg_l5 |
| 51,192 | 0.0745 | team.conversion_l5 | team.matches_since_win |
| 49,171 | 0.0745 | team.season_ppg | team.venue_ppg_l5 |
| 51,192 | -0.0733 | team.conversion_l5 | team.matches_since_blank |
| 50,502 | -0.0728 | team.state_index | team.season_ppg |
| 23,369 | 0.0724 | team.h2h_win_rate_l5 | team.corner_diff_avg_l5 |
| 51,192 | -0.0706 | team.defensive_leak_l5 | team.matches_since_clean_sheet |
| 49,926 | -0.0696 | team.goal_diff_avg_l5 | team.season_ppg |
| 52,250 | -0.0686 | team.goals_against_avg_l5 | team.matches_since_win |
| 51,337 | 0.0667 | team.h2_goals_l10 | team.venue_ppg_l5 |
| 22,242 | 0.0664 | team.h2h_win_rate_l5 | team.season_ppg |
trend vs level
| n | metric | r level | r trend | trend beats level |
|---|
| 103,809 | team.elo | 0.2268 | 0.0293 | no |
| 51,320 | team.venue_ppg_l5 | 0.1141 | -0.0183 | no |
| 51,250 | team.corner_diff_avg_l5 | 0.1443 | 0.0118 | no |
| 51,250 | team.dominance_l5 | 0.1784 | 0.0109 | no |
| 51,250 | team.corners_against_avg_l5 | -0.1191 | -0.0094 | no |
| 51,250 | team.corners_for_avg_l5 | 0.115 | 0.009 | no |
| 51,250 | team.shot_diff_avg_l5 | 0.1805 | 0.0083 | no |
| 51,341 | team.goals_volatility_l10 | 0.0957 | -0.0052 | no |
| 52,250 | team.goals_against_avg_l5 | -0.1089 | -0.0042 | no |
| 53,336 | team.congestion_21d | -0.0056 | -0.0038 | no |
| 49,742 | team.comeback_rate_l20 | 0.0305 | -0.0036 | no |
| 49,742 | team.blown_lead_rate_l20 | -0.025 | 0.0035 | 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
{
"kind": "lead_lag",
"sport": "football",
"top_n": 40
}