RadarWhich analysis gets the result right most often?
Before every match there are many ways to say who will win. Here they are measured on real results, judged on the same matches, with the margin of chance.
Ranking on the same matches based on 1703
For every played match we took what each analysis said BEFORE kick-off and counted how often it got the final result right: 1 (home win), X (draw) or 2 (away win). All analyses are judged on the same matches. The interval shows how much the percentage could move by chance alone.
Our model and Elo rating are level within the margin of chance. No analysis gets much past 1 match in 2: draws (about 25.5% of matches) are almost never predicted.
| # | Analysis | Got it right | 95% interval |
|---|---|---|---|
| 1 | Our model | 51.4%875 of 1703 | 49.0% … 53.7% |
| 2 | Elo rating | 50.7%863 of 1703 | 48.3% … 53.0% |
| 3 | Points per match this season | 46.9%799 of 1703 | 44.6% … 49.3% |
| 4 | Form, last 5 matches | 44.6%760 of 1703 | 42.3% … 47.0% |
| 5 | Head-to-head record | 44.3%754 of 1703 | 41.9% … 46.6% |
| 6 | “The home side wins” | 43.7%745 of 1703 | 41.4% … 46.1% |
Matches: 2024-25, 2025-26, 2026-27, every league we follow.
League by league
The same count, league by league: an analysis can work better in one league than in another. Under each percentage: how many times it was right, out of how many matches.
| League | Our model | Elo rating | Points per match this season | Form, last 5 matches | Head-to-head record | “The home side wins” |
|---|---|---|---|---|---|---|
| Premier League | 49.2%155 of 315 | 49.2%155 of 315 | 47.3%149 of 315 | 47.6%150 of 315 | 46.0%145 of 315 | 42.5%134 of 315 |
| La Liga | 53.2%149 of 280 | 52.5%147 of 280 | 46.1%129 of 280 | 42.9%120 of 280 | 43.9%123 of 280 | 48.6%136 of 280 |
| Serie A | 54.0%177 of 328 | 52.4%172 of 328 | 50.0%164 of 328 | 45.4%149 of 328 | 43.3%142 of 328 | 40.2%132 of 328 |
| Bundesliga | 55.1%150 of 272 | 55.1%150 of 272 | 52.6%143 of 272 | 47.8%130 of 272 | 48.5%132 of 272 | 44.5%121 of 272 |
| Ligue 1 | 53.5%144 of 269 | 51.7%139 of 269 | 45.0%121 of 269 | 46.1%124 of 269 | 48.3%130 of 269 | 47.2%127 of 269 |
| Liga 1 | 41.8%100 of 239 | 41.8%100 of 239 | 38.9%93 of 239 | 36.4%87 of 239 | 34.3%82 of 239 | 39.7%95 of 239 |
How each analysis picks
What each analysis is and how it picks a result. All of them use only matches played before kick-off; if an analysis has too little data for a match, that match does not count for it.
- Our model: The most likely of 1 / X / 2 in the model's estimate (frozen before kick-off; ↺ = rebuilt the same way for past seasons).
- Elo rating: The team with the higher expected score (home advantage included) wins; Elo never picks a draw.
- Points per match this season: The team with more points per match this season wins (at least 3 matches each); a gap of 0.05 or less ⇒ draw.
- Form, last 5 matches: The team with more points in its last 5 matches wins; level ⇒ draw.
- Head-to-head record: The team that won more of their previous meetings wins (at least 2); level ⇒ draw.
- “The home side wins”: Always 1: the simplest benchmark — every analysis should beat it.
Updated 8 Nov 2026