Prediction accuracy: national teams
What is checked: the result: home win, draw or away win. An honest probability means outcomes at 70% happen about 70% of the time.
Check on history
The national teams’ model was refitted at the start of every month on past matches only; the final formula of each football year was trained on the six years before it. Friendlies are in the check, with a row of their own in the table. That is how 1,461 matches from 21 September 2022 to 6 October 2026 were checked.
| Model estimate | Estimates | Average | Happened |
|---|---|---|---|
| 0–10% | 644 | 4.8% | 6.8% |
| 10–20% | 682 | 15.2% | 13.2% |
| 20–30% | 1,126 | 25.7% | 24.5% |
| 30–40% | 656 | 33.5% | 33.4% |
| 40–50% | 324 | 44.8% | 45.1% |
| 50–60% | 269 | 54.8% | 61.0% |
| 60–70% | 219 | 64.9% | 68.5% |
| 70–80% | 208 | 74.7% | 71.2% |
| 80–90% | 147 | 85.0% | 82.3% |
| 90–100% | 108 | 94.8% | 95.4% |
How to read probabilities and calibration →
Only the model's favourite
| Confidence from | Share of matches | Estimate | Came true |
|---|---|---|---|
| 40% | 87% | 64.1% | 65.3% |
| 50% | 65% | 70.7% | 72.1% |
| 60% | 47% | 76.9% | 76.5% |
| 70% | 32% | 82.6% | 80.3% |
Overall the model's most likely outcome came true in 62% of matches, with the favourite estimated at 61% on average. When the two are close, the model neither overstates nor understates its confidence; and even the most likely outcome coming true only about half the time is normal where a draw is possible.
By competition
«Honest»: the gap between the favourite’s estimate and how often it won is within the random spread for that many matches.
| Competition | Matches | Favourite came true | Favourite’s estimate | Probabilities |
|---|---|---|---|---|
| International Friendlies | 449 | 59.2% | 58.4% | honest |
| UEFA Nations League | 350 | 58.0% | 53.9% | honest |
| Euro Qualifiers | 239 | 67.4% | 72.5% | honest |
| World Cup Qualifiers | 204 | 71.1% | 71.4% | honest |
| World Cup | 168 | 60.1% | 53.7% | honest |
| Euro | 51 | 49.0% | 51.1% | honest |
Since launch
We store the last estimate before each match and never change it afterwards. Since 28 September 2026 the most likely outcome came true in 44 of 75 matches with an expected share of 56%. The sample is still small: conclusions need several hundred matches.
| Competition | Matches | Came true | Expected |
|---|---|---|---|
| UEFA Nations League | 70 | 41 | 56% |
| International Friendlies | 5 | 3 | 57% |
Latest 12 matches
- LuxembourgBulgaria6 October · UEFA Nations League343136XThe most likely outcome did not happen: X0:2Final score 0:2. Model's score: 1:1.2.1
- EstoniaIceland6 October · UEFA Nations League1423632The most likely outcome did not happen: 20:0Final score 0:0. Model's score: 0:2.2.6
- MoldovaSlovakia6 October · UEFA Nations League920712The most likely outcome came true: 20:4Final score 0:4. Model's score: 0:2.2.6
- BelarusFinland6 October · UEFA Nations League2129502The most likely outcome did not happen: 21:0Final score 1:0. Model's score: 1:2.2.3
- AlbaniaSan Marino6 October · UEFA Nations League95401The most likely outcome came true: 12:1Final score 2:1. Model's score: 2:0.3.4
- SwitzerlandNorth Macedonia6 October · UEFA Nations League841241The most likely outcome came true: 13:0Final score 3:0. Model's score: 3:1.3.3
- ScotlandSlovenia6 October · UEFA Nations League4530251The most likely outcome did not happen: 11:2Final score 1:2. Model's score: 1:0.2.2
- EnglandCzechia6 October · UEFA Nations League831241The most likely outcome came true: 13:0Final score 3:0. Model's score: 2:0.3.3
- CroatiaSpain6 October · UEFA Nations League918742The most likely outcome came true: 21:2Final score 1:2. Model's score: 0:3.3.2
- RussiaNigeria6 October · International Friendlies5527181The most likely outcome did not happen: 13:3Final score 3:3. Model's score: 3:0.2.7
- KazakhstanFaroe Islands6 October · UEFA Nations League383131XThe most likely outcome came true: X2:2Final score 2:2. Model's score: 1:1.2.1
- LiechtensteinGibraltar5 October · International Friendlies283438XThe most likely outcome did not happen: X0:2Final score 0:2. Model's score: 1:1.1.6