Every pick. Graded.
Today’s picksAI football prediction accuracy.
Every model’s pick and match-outcome accuracy, public and graded. Exact scores and pick value in the app.
Football · rolling 7 days
81 / 145correct outcomes
ResultsPublic consensus outcomes across covered leagues, separate from individual model pick accuracy below.
A count of the last 7 days of graded picks — not an accuracy rate; a week this size swings.
Football · all covered leagues
Pick accuracy Last 30 days
How often a model’s recorded pick was correct. Only graded selections with a recorded market price count.
Pick accuracy
Match-outcome record
199 of 262 right since 20 July 2026
Strong tier: the match-outcome record on every graded match the panel of AI models backed at Strong-tier firmness since 20 July 2026. It is a separate sample with no rolling window, so it is not comparable with the ScoreGPT figure beside it.
Compare the models
Live order. Real samples. Every model’s accuracy, public.
- 1Grok 4.6931 graded picks61.9%
- 2GLM-5.3935 graded picks59.1%
- 3Claude Opus 5.5940 graded picks58.6%
- 4Kimi K3917 graded picks57.8%
- 5GPT-5.6939 graded picks56.3%
Pick accuracy and match-outcome accuracy are public for every model. Exact-score accuracy, pick value and league breakdowns are in the app with Pro.
Sample sizes vary by model. A higher rate in one window does not establish a lasting advantage.
Know the record behind your next pick.
Get the full picture with ScoreGPT Pro in the app: compare model performance, follow the Matchday tiers — Strong, Lean and Open — and see how each model performs across leagues.
- Full model rankings
- Matchday tiers (Strong / Lean / Open)
- League breakdowns
Get the free app
Exact-score accuracy, pick value and league breakdowns require a Pro subscription in the app. Monthly and annual plans available.
3 free predictions each week · no sign-up
Picks are recorded before kickoff and checked against final results. Correct calls and misses both count. Pick value is a historical unit-based measure. It is not a forecast of future returns.
By competition · 30 days
Find your league. Follow its record.
See how the models compare in each competition with Pro in the app.
Every model receives the same frozen pre-match briefing. Predictions lock before kickoff and are graded automatically from final scores — wins and losses alike. Read the full methodology.
Keep exploring: how accurate are AI football predictions? · compare AI prediction tools · what confidence means · World Cup 2026 predictions
AI prediction accuracy, explained
What is the difference between pick accuracy and match-outcome accuracy?
Pick accuracy checks whether each model’s recorded selection was correct. That selection may differ from its home-win, draw or away-win prediction. Pick metrics include only graded selections with a market price recorded at prediction time; outcome metrics use the separate match-prediction sample. The page shows each metric’s own sample size.
How accurate are AI football predictions?
The last 30 days include 4,442 graded model predictions. Multiple models can cover the same match, so this is not a count of unique matches. Match-outcome accuracy across the independent models ranged between 50.6% and 51.6% in that window. Result accuracy means the model called the match outcome (home win, draw or away win) correctly. Every figure on this page comes from real predictions graded against final scores — never simulations.
How is accuracy measured?
Each model receives the same frozen pre-match briefing and locks its prediction before kickoff. After the final whistle the prediction is graded automatically: result accuracy checks the predicted outcome, exact-score accuracy checks the precise scoreline, and calibration checks whether a model’s stated confidence matches how often it is actually right. Statistics are recomputed after every graded match.
Why do the models have different sample sizes?
Models can cover different numbers of matches. Pick metrics also require a recorded market price. Windows are rolling 7, 14 and 30 days — a LOW SAMPLE flag identifies thin samples rather than hiding them. The Strong-tier record is the exception: it counts every graded match the panel of AI models backed at Strong-tier firmness since 20 July 2026, and it is a separate sample from the model figures.
Which AI model is best at football predictions?
The lead changes between time windows — that is exactly why this page exists. Compare the 7-, 14- and 30-day leaderboards above; treat any single window as a snapshot, not proof of a permanent edge.
Which performance insights are free?
The website shows every model’s pick accuracy and match-outcome accuracy, the live model order and graded sample sizes for each time window. Exact-score accuracy, pick value, league breakdowns and confidence analysis require ScoreGPT Pro in the app. Individual finished matches remain publicly graded. Downloading the app is free; Pro is a separate subscription.
Is this betting advice?
No. Pick value is a transparent experiment: every pick is scored against the real market price recorded at prediction time, and value is tracked in units, not money. Nothing here is a guarantee of future results. For information and entertainment only — 18+. Please play responsibly.