Model elo-v2
Methodology
The Footiqra power ranking estimates how strong every team in Europe's top leagues is right now, on one common scale. It is based only on match results. This page documents exactly how it is calculated (model elo-v2).
1. Team ratings
Every team has a rating. Before each match the model estimates each side's chance from the rating difference, using the standard Elo formula: a 100-point advantage corresponds to an expected score of about 64%. After the match, both teams move towards the result: the more surprising the result, the bigger the change. Whatever one team gains, the other loses.
- Base rating: 1500 (plus league strength, see below)
- Update step (K-factor): 20 rating points
- Home advantage: 65 rating points
- Margin of victory: the step is multiplied by 1 for a draw or one-goal win, 1.5 for a two-goal win and (11 + n) / 8 for a win by n ≥ 3 goals
2. New season
When a new season starts, clubs are not reset to 1500. Each club keeps half of its distance from the base rating: a club that finished last season on 1700 starts on 1600. Promoted clubs and clubs that did not play in the same league last season start 50 points below the base rating. League strength carries on unchanged.
So last season gives a sensible starting point, but the current season quickly takes over: every match moves the rating again, and older results fade. The data starts with the 2024/25 season, when every club started at the base rating.
Why half? We replayed every match from 2024/25 to October 2026 and scored the forecast made before each one. Carrying ratings over predicted the results of 2026/27 clearly better than resetting every club each summer. Carrying them over in full predicted slightly better still, but let last season dominate the current ranking for too long; half is the compromise.
3. Comparing leagues
Within a league, every team plays every other team, so ratings inside a league are well connected. Between leagues they are not: Bayern and Arsenal only meet in European competition. Domestic results alone cannot tell whether 80 points in one league are worth more than 80 points in another.
Footiqra therefore adds a league strength value to every team's rating. It starts from the official UEFA association coefficients, which summarise five seasons of European results: each league starts 2 rating points above or below average for every coefficient point it is above or below the average of all covered leagues.
| League | UEFA coefficient | Starting strength |
|---|---|---|
| Premier League | 119.519 | +57.9 |
| Serie A | 99.946 | +18.7 |
| La Liga | 97.046 | +12.9 |
| Bundesliga | 92.902 | +4.7 |
| Ligue 1 | 83.498 | −14.1 |
| Primeira Liga | 73.166 | −34.8 |
| Eredivisie | 67.929 | −45.3 |
Source: UEFA association club coefficients, season 2025/26 (read 2026-10-28).
These starting values applied at the start of 2024/25 and are only a prior. Every match between clubs from two different leagues covered by Footiqra also moves the two league strengths, by 12 points times the same surprise factor used for teams. Domestic matches never change league strength, because both teams share it.
4. Which matches count
- All finished league matches since the 2024/25 season; earlier seasons only shape the starting ratings (see “New season”)
- Champions League matches in which both clubs play in one of the covered leagues. Matches against clubs from other countries are not used, because their league strength is not modelled
- Matches decided in extra time or on penalties are not used yet; neither are cancelled or postponed matches
Results come from football-data.org and are updated several times a day. The calculation is deterministic: the same results always produce the same ranking.
5. Match forecasts
Forecasts use the same ratings (model forecast-v1). From the rating difference d between home and away side, home advantage included, the Elo formula gives the home side an expected score E = 1 / (1 + 10-d/400). A draw counts as half a win in E, so the three probabilities are:
- Draw: 25% × (1 − |2E − 1|2), highest for evenly matched sides
- Home win: E − draw / 2
- Away win: 1 − E − draw / 2
- No outcome is shown below 2%
The two draw parameters were fitted to this season's league and Champions League results, using the ratings each club had before kick-off. On those matches the forecasts score clearly better than always predicting the average share of home wins, draws and away wins. The forecasts are estimates for information, not betting advice.
6. Predicted final tables
The predicted tables (model sim-v1) play the rest of each league season 10,000 times. Every simulation starts from the official table and decides each remaining match at random with the forecast probabilities above. Winning margins follow the margins seen in that league this season.
After each simulated match the two ratings are updated exactly like after a real result. A club that starts a simulated run well therefore also gets stronger in that run, which keeps the uncertainty about true team strength in the results. Ties on points are split by goal difference, then at random; head-to-head and other league specific tie-breakers are not modelled. The simulation uses a fixed random seed, so the same data always gives the same table.
7. Limitations
- Early in a season, ratings still lean on last season. Summer transfers or a new coach only show up once the results change.
- Only a small number of matches per season connect the leagues, so league strength rests on comparatively few results.
- The model only sees scores. Injuries, transfers, expected goals or squad rotation are not taken into account.