How accurate is ScoreGPT at football predictions?

meta

Real predictions, graded against real results — a live experiment, not a proven edge. Last 30 days, recomputed after every graded match (updated 17 Aug 2026, 06:41 UTC).

Result accuracy
47.8%
Exact score
8.7%
Calibration
52.7%
Graded predictions
312
01 — Confidence & calibration

Across 172 predictions with a stated confidence, ScoreGPT averaged 47.1% confidence — 50.1% when it was right vs 44.7% when it was wrong. Calibration score: 52.7%.

02 — By competition (30 days)
CompetitionGradedResult acc.Exact score
Major League SoccerUSA6149.2%11.5%
Serie ABrazil4233.3%4.8%
UEFA Europa LeagueWorld3557.1%5.7%
UEFA Champions LeagueWorld3452.9%11.8%
Liga MXMexico2839.3%3.6%
2. BundesligaGermany1838.9%11.1%
EredivisieNetherlands1861.1%16.7%
AllsvenskanSweden1353.8%0.0%
ChampionshipEngland1136.4%0.0%
EliteserienNorway1050.0%20.0%
Pro LeagueSaudi-Arabialow sample966.7%0.0%
BundesligaAustrialow sample862.5%25.0%
Süper LigTurkeylow sample850.0%12.5%
CONMEBOL LibertadoresWorldlow sample70.0%0.0%
Super LeagueSwitzerlandlow sample475.0%0.0%
La LigaSpainlow sample475.0%0.0%
World CupWorldlow sample250.0%50.0%

Low-sample rows are dimmed, not hidden — transparency beats tidiness.

03 — Betting experiment (30 days)

1 virtual unit per pick, tracked in units — a live experiment, not betting advice. 18+.

Win rate
44.1%
ROI
-4.6%
P/L
-14.4u
Avg odds
2.33
Bets
313

← Full model leaderboard · how grading works · which AI is best? · compare prediction tools