1,900 EPL matches · Five seasons · Frozen closing-market sample

How Often Do Premier League Favourites Win?

  • 1,899 unique favourites
  • Six probability bands
  • Wilson 95% intervals

Direct answer: the unique closing favourite won 1,063 of 1,899 matches— 55.98%. It did not win 836 times (44.02%). One tied top probability was excluded. This is a closing-market reference, not AI model accuracy or a profit claim.

Published by Football Proof AI · Published · premier-league-favourite-win-rate-benchmark/1.0.0 · Editorial technical note; not externally peer reviewed
Premier League favourite win rate across 1899 unique closing favourites
OPEN DATASET · EPL FAVOURITESCLOSING-MARKET REFERENCE · NOT AI ACCURACY OR A PROFIT CLAIM

Canonical publication record

Abstract

A reproducible, aggregate-only answer to how often Premier League favourites win: 1,063 wins from 1,899 unique closing favourites across 1,900 completed matches and five seasons, with probability bands, Wilson intervals, season variation, threshold coverage, source hashes and frozen cross-checks.

Author and publisher
Football Proof AI
Publication version
premier-league-favourite-win-rate-benchmark/1.0.0
Published
Last modified
Release status
Current release
Review status
Editorial technical note; not externally peer reviewed
Production method
Editorial dataset maintained from the declared method, linked sources and any listed reproducible artifacts.
Software and AI assistance
Software, including generative AI, may assist drafting, transformation or quality checks. It is not treated as a source, author or independent reviewer; factual and quantitative claims must remain bound to cited sources, declared methods or reproducible artifacts.
Why this exists
Enable readers to inspect assumptions, reproduce calculations and reject claims that exceed the published evidence.
Commercial boundary
No bookmaker, prediction provider or paid placement controls the evidence rules, calculations or conclusions on this page.
Version history
  1. premier-league-favourite-win-rate-benchmark/1.0.0 : Initial public release.
Immutable artifacts
  1. 1.0.0.json sha256:2c9317706d5222207f9153b9d891f1a17db46b52a78eaf3a4776becdf16e7b24
  2. premier-league-favourite-win-rate-v1.csv sha256:9900769c13bb3bc7e51f771db9b0f70d2ac1666b3518928086b9089d7e16716b
  3. 1.0.0-manifest.json sha256:e9d5a0c5bb76b447bd66da985dc492cc083c4fbfe303d6c751adebebd3bcc043
  4. premier-league-favourite-win-rate-generator-1.0.0.mjs sha256:15fc2cb095ca1e337df62eceffb230ce741504ec56c88db84672b1037515cb9f
Release resources
  1. Frozen EPL 1X2 empirical scoring benchmark
  2. Frozen EPL probability calibration benchmark
References
  1. https://www.football-data.co.uk/data.php
  2. https://www.football-data.co.uk/notes.txt
  3. https://doi.org/10.1080/01621459.1927.10502953
  4. https://footballproofai.com/research/football-1x2-scoring-benchmark
  5. https://footballproofai.com/research/football-1x2-probability-calibration-benchmark
Favourite win rate55.98%

1,063 wins from 1,899 unique favourites.

Favourite non-win rate44.02%

836 draws or other-outcome wins.

Mean no-vig probability54.06%

Observed minus mean probability: +1.92 pp.

Wilson 95% interval53.7%58.2%

Sampling uncertainty, not an independence guarantee.

One quotable result · One visible denominator

The favourite still failed to win 44.02% of the time.

1,063 / 1,899

“Favourite” means the unique maximum probability after proportional margin removal from average closing H/D/A odds. “Win” means that exact regulation-time outcome occurred. The frozen aggregate covers 2021/22 to 2025/26.

Included
1,899 matches with one unique top closing probability.
Excluded
1 match with a tied top probability; no arbitrary tie-break.
Observed interval
53.73%–58.20% under a Wilson 95% score interval.
Not established
AI accuracy, future certainty, value betting, profit or a universal cross-league rate.

Favourite strength · Exact disclosed bands

Stronger closing favourites won more often—but coverage fell fast.

The bars show observed win rate, not the probability assigned to an upcoming match. Every point retains its denominator and interval; no band was chosen after seeing the result.

  1. 30% to below 40%n=280 · 129 wins
    46.07%95% CI 40.33%–51.92%
  2. 40% to below 50%n=581 · 260 wins
    44.75%95% CI 40.75%–48.81%
  3. 50% to below 60%n=467 · 265 wins
    56.75%95% CI 52.21%–61.17%
  4. 60% to below 70%n=301 · 194 wins
    64.45%95% CI 58.89%–69.65%
  5. 70% to below 80%n=192 · 146 wins
    76.04%95% CI 69.53%–81.53%
  6. 80% to below 90%n=78 · 69 wins
    88.46%95% CI 79.50%–93.81%
Observed regulation-time hit rate for the unique closing favourite. Empty 0–30% and 90–100% fixed-width bands remain in the downloadable data but are omitted from this visual because they contain no matches.
Exact probability-band aggregates
No-vig favourite bandnWinsObserved win rateMean probabilityGapWilson 95% CI
30% to below 40%28012946.07%37.84%+8.23 pp40.33%–51.92%
40% to below 50%58126044.75%44.70%+0.05 pp40.75%–48.81%
50% to below 60%46726556.75%55.12%+1.62 pp52.21%–61.17%
60% to below 70%30119464.45%64.80%-0.35 pp58.89%–69.65%
70% to below 80%19214676.04%74.57%+1.48 pp69.53%–81.53%
80% to below 90%786988.46%83.69%+4.78 pp79.50%–93.81%

Confidence threshold · Accuracy and coverage together

A higher hit rate is not free: the sample gets smaller.

Quoting only the last card would hide 95.9% of unique favourites. Threshold results must disclose both observed accuracy and the fraction of the full eligible sample retained.

Below 50% favourite probability45.18%

389 / 861 won · 45.3% of eligible favourites · 95% CI 41.88%–48.52%

At least 50% favourite probability64.93%

674 / 1,038 won · 54.7% of eligible favourites · 95% CI 61.98%–67.78%

At least 70% favourite probability79.63%

215 / 270 won · 14.2% of eligible favourites · 95% CI 74.43%–84.00%

At least 80% favourite probability88.46%

69 / 78 won · 4.1% of eligible favourites · 95% CI 79.50%–93.81%

Season variation · Same definition every year

One season ranged from 49.47% to 60.00% across this window.

The overall 55.98% is a five-season weighted aggregate, not a stable annual constant. The same frozen rule is applied to every season, including the single excluded tie.

Five-season stability table
SeasonEligibleWinsNon-winsWin rateMean probabilityWilson 95% CITies excluded
2021/2238022515559.21%54.38%54.20%–64.04%0
2022/2338021116955.53%53.66%50.50%–60.44%0
2023/2438022815260.00%55.91%55.00%–64.80%0
2024/2537921116855.67%54.60%50.64%–60.59%1
2025/2638018819249.47%51.74%44.48%–54.48%0

Simple outcome rates · A label is not a fair contest

Always choosing home hit 44.16%; the closing favourite hit 55.98%.

That difference is descriptive, not a model-vs-market claim. Closing odds identify a different outcome match by match and can use information unavailable 24 hours earlier.

Always choose home44.16%

839 of 1,900 completed matches · 95% CI 41.94%–46.40%

Always choose draw23.89%

454 of 1,900 completed matches · 95% CI 22.03%–25.86%

Always choose away31.95%

607 of 1,900 completed matches · 95% CI 29.89%–34.08%

Unique closing favourite55.98%

1,063 of 1,899 eligible matches · one top-probability tie excluded.

Which outcome was the unique closing favourite?
Favourite labelnWinsWin rateMean probabilityWilson 95% CI
Home1,19068457.48%55.82%54.65%–60.26%
Draw00No unique draw favouritesNot available
Away70937953.46%51.11%49.78%–57.10%

Critical comparison boundary

Closing prices know more than a 24-hour-ahead forecast.

Football-Data labels AvgCH, AvgCD and AvgCA as average closing odds. They can reflect late line-ups, injuries and market movement. This sample therefore cannot be used as a same-information contest with the site's planned forecasts, which must be locked at least 24 hours before kick-off.

Late-information market reference ≠ live AI accuracy ≠ betting profit.

Reproducible method · No hand-edited scorecard

Five source hashes, one fixed favourite rule, two benchmark cross-checks.

  1. 01

    Verify source bytes

    Five EPL evaluation CSVs must match the SHA-256 values inherited from the frozen manifests.

  2. 02

    Remove margin

    Invert all three closing average odds and divide each inverse by their sum.

  3. 03

    Require one top label

    Exclude a tied maximum instead of picking home, draw or away by arbitrary class order.

  4. 04

    Keep every denominator

    Publish wins, non-wins, mean probability, gap, interval and coverage for each declared group.

  5. 05

    Cross-check releases

    Overall, season and band totals must match the existing scoring and calibration benchmarks.

  6. 06

    Freeze the output

    Versioned JSON, CSV, manifest and generator bytes expose every aggregate used by this page.

Open aggregates · Exact bytes · Rights kept separate

Download, reproduce and challenge the result.

The release republishes no raw match or odds rows. Its original aggregate analysis is offered under CC BY 4.0 only to the extent Football Proof AI holds rights in it; source files remain with their publisher.

JSON SHA-256
2c9317706d5222207f9153b9d891f1a17db46b52a78eaf3a4776becdf16e7b24
CSV SHA-256
9900769c13bb3bc7e51f771db9b0f70d2ac1666b3518928086b9089d7e16716b
Manifest SHA-256
e9d5a0c5bb76b447bd66da985dc492cc083c4fbfe303d6c751adebebd3bcc043
Generator SHA-256
15fc2cb095ca1e337df62eceffb230ce741504ec56c88db84672b1037515cb9f

Favourite win-rate FAQ

Five boundaries that keep 55.98% honest.

How often do Premier League favourites or favorites win?

In this frozen 2021/22–2025/26 sample, the unique closing favourite won 1,063 of 1,899 matches: 55.98%. One tied top probability was excluded. This is one league and one closing-market definition, not a universal football constant.

How often did 50%+ no-vig favourites win?

Favourites with at least 50% proportional no-vig probability won 674 of 1,038 matches: 64.93%. That is not identical to every bookmaker use of “odds-on”, because quoted odds include margin and can differ by bookmaker.

Is a 70% Premier League favourite guaranteed to win?

No. The at-least-70% group won 215 of 270 matches, or 79.63%. 55 still did not win, and the result describes a historical group rather than the next match.

What counts as an upset in this dataset?

The release uses the narrower label ‘favourite non-win’: the uniquely most-probable H/D/A outcome did not occur. It can include a draw or the other team winning. It does not establish a profitable underdog bet, market inefficiency or a causal shock.

Does a 55.98% favourite win rate imply betting profit?

No. Hit rate ignores price, margin, stake availability, limits, commission, slippage and selection rules. A market can correctly identify the likelier outcome while offering no positive expected value.