Actual − expected points deviation, whole Liga
As % of expected points, per season
The 2019-20 and 2020-21 seasons (matches behind closed doors) are slightly below expectations. Probabilities = average market odds, bookmaker margin removed.
Study
The odds as a benchmark
Why bookmakers' odds are a fair yardstick for results.
Key takeaways
- Across the whole Liga, the deviation between actual and expected points ranges from −1.3% to +1.8% depending on the season (2000-01 → 2025-26). See
- 26 seasons out of 26 stay within 2% of expected points. See
- 2 matches without odds are excluded from odds-based measures. See
- An Elo model was tested for the archives and rejected: its expected points differ too much from those derived from the odds. See
In short: The whole "points − expected" measure rests on the odds. Across the Liga, actual points deviate from expected points by −1.3% to +1.8% depending on the season.
Calibration Across the whole Liga, actual points and points expected from the odds differ by only a few % per season: the odds are a reliable benchmark.
By club, all seasons 2000-01 → 2025-26
Actual against expected points; sorted by number of matches, not by deviation
| Club | Matches | Points | Expected | Deviation |
|---|---|---|---|---|
 Athletic Bilbao | 988 | 1,375 | 1,398.0 | −23.0 (−1.6%) |
 FC Barcelona | 988 | 2,168 | 2,119.5 | +48.5 (+2.3%) |
 Valencia CF | 988 | 1,529 | 1,509.3 | +19.7 (+1.3%) |
 Real Madrid CF | 987 | 2,161 | 2,063.6 | +97.4 (+4.7%) |
 Sevilla FC | 950 | 1,483 | 1,451.5 | +31.5 (+2.2%) |
 Villarreal CF | 950 | 1,475 | 1,397.9 | +77.1 (+5.5%) |
 Atlético Madrid | 912 | 1,646 | 1,582.1 | +63.9 (+4.0%) |
 RCD Espanyol | 912 | 1,096 | 1,112.4 | −16.4 (−1.5%) |
 Real Sociedad | 874 | 1,211 | 1,206.8 | +4.2 (+0.3%) |
 CA Osasuna | 836 | 1,004 | 993.0 | +11.0 (+1.1%) |
 Real Betis | 835 | 1,123 | 1,098.0 | +25.0 (+2.3%) |
 Getafe CF | 798 | 976 | 966.2 | +9.8 (+1.0%) |
 RC Celta de Vigo | 759 | 982 | 1,025.7 | −43.7 (−4.3%) |
 RCD Mallorca | 722 | 904 | 875.3 | +28.7 (+3.3%) |
 Deportivo de A Coruña | 608 | 814 | 822.2 | −8.2 (−1.0%) |
 Málaga CF | 608 | 743 | 751.5 | −8.5 (−1.1%) |
 Levante UD | 570 | 625 | 597.4 | +27.6 (+4.6%) |
 Rayo Vallecano | 532 | 613 | 622.9 | −9.9 (−1.6%) |
 Real Valladolid | 532 | 550 | 593.4 | −43.4 (−7.3%) |
 Deportivo Alavés | 494 | 568 | 567.4 | +0.6 (+0.1%) |
 Racing de Santander | 418 | 478 | 486.7 | −8.7 (−1.8%) |
 Real Zaragoza | 417 | 487 | 523.0 | −36.0 (−6.9%) |
 Granada CF | 380 | 380 | 396.1 | −16.1 (−4.1%) |
 UD Almería | 304 | 304 | 322.0 | −18.0 (−5.6%) |
 SD Eibar | 266 | 302 | 310.7 | −8.7 (−2.8%) |
 UD Las Palmas | 266 | 263 | 286.1 | −23.1 (−8.1%) |
 Elche CF | 228 | 227 | 221.3 | +5.7 (+2.6%) |
 Girona FC | 228 | 300 | 289.9 | +10.1 (+3.5%) |
 Sporting de Gijón | 228 | 237 | 243.6 | −6.6 (−2.7%) |
 Cádiz CF | 190 | 194 | 190.7 | +3.3 (+1.7%) |
 CD Leganés | 190 | 199 | 207.0 | −8.0 (−3.9%) |
 Recreativo de Huelva | 152 | 167 | 169.3 | −2.3 (−1.4%) |
 CD Numancia | 114 | 103 | 118.7 | −15.7 (−13.2%) |
 Albacete Balompié | 76 | 75 | 82.2 | −7.2 (−8.7%) |
 SD Huesca | 76 | 67 | 79.7 | −12.7 (−15.9%) |
 Real Murcia CF | 76 | 56 | 76.9 | −20.9 (−27.1%) |
 Real Oviedo | 76 | 70 | 80.1 | −10.1 (−12.6%) |
 CD Tenerife | 76 | 74 | 84.6 | −10.6 (−12.5%) |
 Córdoba CF | 38 | 20 | 33.0 | −13.0 (−39.4%) |
 Gimnàstic de Tarragona | 38 | 28 | 35.3 | −7.3 (−20.7%) |
 Hércules CF | 38 | 35 | 42.5 | −7.5 (−17.7%) |
 Xerez CD | 38 | 34 | 34.3 | −0.3 (−0.9%) |
Method
- Source: football-data.co.uk, pre-match market average (not closing odds).
- Margin removed by proportional normalisation: p = (1/odds) / Σ(1/odds).
- Expected points = 3 × P(win) + P(draw), match by match.
A rejected trial: the Elo model
For the archives (1990-91 → 1999-00), which have no odds, an Elo model (expected points computed from past results only) was built and tested against the odds over 2000-01 → 2025-26. Its expected points differ by several points per club and season from those derived from the odds: it was rejected as a benchmark, and the archives remain descriptive.
Limitations
- Proportional normalisation may leave a slight favourite-longshot bias.
- The odds include what the market knew before the match; a bias the market knew about would already be "in the odds".
- 2 matches without odds are excluded from odds-based measures.
Definitions
- Expected points
- Before each match, the bookmakers' odds give the probability of a win, a draw and a defeat. Expected points = 3 × P(win) + P(draw). A positive deviation means the club took more points than the market predicted. Because the team's strength is already built into the odds, a big club has no "automatic" positive deviation.
- q-value
- With hundreds of comparisons, some look extreme by chance. The Benjamini-Hochberg correction accounts for this: a deviation is only called notable if q < 0.05. Otherwise: inconclusive. Inconclusive does not mean "no effect": with 11 matches, only a deviation of about ±1.57 points per match could have been detected.
- Difference-in-differences
- We measure the club's change between two periods, then subtract the change of a comparison group over the same periods. What remains is the change specific to the club.
- Strength-adjusted measures
- Each match is compared with Liga matches from the same season, at the same venue type (home/away) and between teams of comparable strength according to the odds. A positive value is favourable to the club.