# A heavy favourite only wins 69% of league matches
> Across 6,580 Big Five league fixtures, heavy favourites win 69% of matches. Almost one third of heavy mismatches don't go to the favourite at all.
- URL: https://goalsforge.com/blog/a-heavy-favourite-only-wins-69-percent-of-league-matches/
- Markdown URL: https://goalsforge.com/blog/a-heavy-favourite-only-wins-69-percent-of-league-matches.md
- Author: Goalsforge
- Published: 2026-10-14T00:00:00.000Z
- Tags: Statistics, Probability, Leagues
---
In the 486 Big Five league matches across 2020-21 through 2024-25 where one team finished 11 or more places above the other the prior season, the supposed favourite won 68.9% of the time, drew 19.1%, and lost 11.9%. Almost a third of those heavy mismatches did not go to the favourite, and the pre-match framing usually assumes closer to 85%.

## What we measured

For each of the Premier League, La Liga, Serie A, the Bundesliga, and Ligue 1 across 2020-21 through 2024-25, we identified the pre-match favourite for every league fixture using a mechanical rule: the team with the better (lower-numbered) final league position in the prior season. Ties go to the home side. We bucketed every fixture by the resulting position gap into seven bands: ±11 or more, ±6 to ±10, ±1 to ±5. The 0-delta bucket is empty across the dataset because Wikipedia tables publish positions after all tiebreakers are applied. We excluded matches involving any newly-promoted side in its first top-flight season (the prior-season position is not directly comparable), removing 27% of fixtures and leaving **6,580 eligible matches** across the 25 league-seasons.

## The data

| Bucket (favourite vs. underdog by prior-season position delta) | n | Favourite win rate | Draw rate | Underdog win rate |
| --- | ---: | ---: | ---: | ---: |
| +11 or more | 486 | 68.9% | 19.1% | 11.9% |
| +6 to +10 | 1,090 | 60.3% | 21.7% | 18.0% |
| +1 to +5 | 1,714 | 48.8% | 26.4% | 24.7% |
| -1 to -5 | 1,714 | 34.0% | 28.6% | 37.4% |
| -6 to -10 | 1,090 | 28.1% | 26.3% | 45.6% |
| -11 or less | 486 | 18.3% | 25.3% | 56.4% |
| **All buckets** | **6,580** | **42.7%** | **25.6%** | **31.8%** |

The aggregate favourite-win rate sits below 50% because the methodology labels one team in every fixture as the favourite, and the same two teams play twice a season (once at each side's ground). Both legs count a "favourite win" when the same side wins both, and an "underdog win" when the home side springs the upset. The bucket structure is therefore symmetric around zero, with the most-populated near-zero buckets pulling the aggregate toward 50%.

The favourite win rate scales cleanly and monotonically with position gap. That is the first thing to check, and the curve goes in the only direction it should. It also stops well short of where the pre-match framing usually sits. Heavy favourites lose or draw 31% of matches, roughly one in three.

### Why the slight favourite is a coin flip

The most striking reading of the table is the **+1 to +5** bucket. Slight favourites win 48.8% of matches. They are below 50%. Once the home-away direction is accounted for, the slight-favourite bucket is functionally a coin flip, with the 1.2 percentage point gap from 50% well within sampling noise (around ±1 percentage point at this bucket size).

Matches between two sides that finished within a few places of each other the prior season behave almost indistinguishably from random. Fans often describe this kind of fixture as "should be close", and they are right, but the data says closer than that. The two sides are roughly evenly matched, and the differences in form, home advantage, and current-season PPG that emerge between August and May are large enough to wash out the small prior-season gap.

### A worked example

Real Madrid's 2020-21 La Liga home fixtures illustrate the heavy-favourite bucket cleanly. Real Madrid sat first in the prior-season standings. In their six home matches against bottom-half opposition (+11 or higher position gap), they won four, drew one, and lost two. The two losses, against Alavés and Levante, and the draw against Real Betis, were the matches they were expected to win comfortably, and they did not. The aggregate 68.9% favourite-win rate for the +11-or-more bucket lines up with a single team-season's small sample at 66.7%. The miss rate is the story.

### Per league

| League | Eligible matches | Aggregate favourite-win rate | Draw rate | Underdog-win rate |
| --- | ---: | ---: | ---: | ---: |
| Premier League | 1,360 | 43.8% | 23.5% | 32.7% |
| La Liga | 1,360 | 42.5% | 27.5% | 30.0% |
| Serie A | 1,360 | 41.2% | 26.3% | 32.5% |
| Bundesliga | 1,200 | 44.6% | 25.3% | 30.1% |
| Ligue 1 | 1,300 | 41.5% | 25.2% | 33.4% |

Aggregate favourite-win rates sit within a narrow band, between 41% and 45% across the five leagues. Where the leagues diverge most is in the heavy-favourite bucket itself: La Liga's heavy favourites win 78.1% of matches, Serie A's 72.6%, the Bundesliga's 70.7%, the Premier League's 61.9%, and Ligue 1's 61.1%. La Liga is the league where the strongest clubs are most dominant over the weakest. The Premier League and Ligue 1 are the leagues where weak clubs beat top clubs most often, a 17 percentage-point spread between the highest and lowest. That spread is large enough to be a league-feature finding, not a sample-size artefact.

## What we tried that did not work

We first tried using current-season points-per-game at the moment of each fixture as the favourite proxy. That produced a tautology: the team with more points has by definition won more matches, so the favourite always wins at least 50%, and the bucket curve collapsed to a flat 50%. The prior-season-position methodology removes the circularity.

We also tried using prior-season points rather than prior-season position as the proxy. That improved on position slightly in the heavy-favourite bucket, but produced uneven buckets (a +6 position gap in the Bundesliga is a larger proportional gap than +6 in the Premier League). The position-based method is less normalised but cleaner to bucket, and conveys the same story.

## Known limitations

Sample sizes in the heaviest-favourite bucket (±11 or more, n=486) are smaller than the aggregate. The Bundesliga has 18 teams (306 fixtures per season), so its bucket sizes run at about 75 per band versus 105 for the 20-team leagues. The Premier League and Ligue 1 heavy-favourites win at lower rates, between 61% and 62%, than the Spanish, Italian, and German equivalents, between 70% and 78%, with a spread of 17 percentage points between La Liga's 78.1% and the Premier League's 61.9%. That spread is large enough to be a real league-feature finding.

The 27% of matches excluded for promoted-side involvement are not random. They skew toward the early-season matches and toward the kind of "favourite at home vs. promoted side" mismatch where the data would most likely have pushed the heavy-favourite bucket higher. Excluding them makes the headline number a lower-bound estimate of true favouritism intensity.

The bucket thresholds (+6, +11) are arbitrary. Choosing +7 or +5 would shift the bucket sizes by 10-20% per band. The shape of the curve is preserved across reasonable alternative thresholds, but the precise bucket counts are not.

## What this means for a forecast

A pre-match analysis that treats a heavy favourite as a near-certain winner is mispricing roughly one third of those matches. The correct read is closer to 69% than to 90%. A 70-30 call on a heavy-favourite match is the honest calibration; a 90-10 call is over-confident by roughly 20 percentage points.

For a slight favourite, the correct read is essentially 50-25-25. The favourite framework contributes almost no information beyond the home-away direction in this bucket. The pre-match model has to lean on other inputs, current-season form, head-to-head, available players, to find any signal. The prior-season position is a wash.

A model with a 50% baseline favourite win rate, then nudged up or down by 19 percentage points for each bucket of position gap, recovers the table well. That is a deliberately simple functional form, but it is enough to reproduce the bucket structure, and the simplicity makes it auditable.

For a reader following [coverage across leagues](/blog/which-european-league-is-hardest-to-predict/), the same heavy-favourite bucket that wins 78% in La Liga wins only 62% in the Premier League. The model that catches this split is the model that knows the favourite framework is one input among several, not a fixed pre-match assumption. The Bundesliga and Ligue 1 separation in [home-field advantage](/blog/home-field-advantage-is-splitting-apart/) is the same kind of league-specific finding applied to a different input: a feature that looks uniform across leagues is actually moving differently in each one.

A read is a lens for separating signal from noise. The favourite framework is signal. The pre-match take treats it as a much stronger signal than it actually is. The 31% of heavy-favourite matches that do not go as expected are where the model earns its keep.


## Continue with the league coverage

The analysis above is one slice of the per-fixture reads Goalsforge publishes pre-kickoff across [Premier League](/coverage/premier-league/), [La Liga](/coverage/la-liga/), [Serie A](/coverage/serie-a/), [Bundesliga](/coverage/bundesliga/), [Ligue 1](/coverage/ligue-1/). Each league hub carries the per-fixture probability distribution, expected scoreline, and the full published record.