— Live · updated every 6 hours
AI prediction accuracy report
10 AI models, 1,244 games, 12,440 graded picks (22 Sep – 22 Oct 2026).
Every pick is made before the game and graded after it. Wins and losses both count.
Key findings
- Gemini 3.8 Flash picked the right winner most often: 54.8% of 1,244 picks. Mistral Medium 3.5 was last at 51%, so first to last was only 3.8 points.
- When every model agreed (629 of 1,244 games), the pick was right 60.3% of the time, against 54% for the majority pick across all games.
- Confidence means something: picks made with 70% or more confidence were right 70.6%, picks under 50% only 41.1%.
- In football (soccer), where draws make it a three-way call, the majority of the models was right 51.9% of 490 matches, against 46.3% for always backing the home team and 33.3% for a random pick.
Accuracy by AI model
Share of picks where the model called the winner (or the draw) correctly, all sports.
| # | Model | Right winner | Picks | Exact score | Backed home team |
|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash |
54.8%
|
1,244 | 6.8% | 61.2% |
| 2 | Claude Sonnet 5 |
54.3%
|
1,244 | 6.4% | 75% |
| 3 | GLM 5.3 |
54.1%
|
1,244 | 5.8% | 66.8% |
| 4 | Kimi K3 |
53.7%
|
1,244 | 6.8% | 59.6% |
| 5 | GPT-5.6 Luna |
53.5%
|
1,244 | 6.3% | 60.8% |
| 6 | Grok 4.6 |
53.1%
|
1,244 | 5.8% | 67.6% |
| 7 | Qwen 3.8 Max |
53.1%
|
1,244 | 5.7% | 60.2% |
| 8 | MiMo V2.6 Pro |
52.3%
|
1,244 | 5.7% | 63.7% |
| 9 | DeepSeek V4 Pro |
52.2%
|
1,244 | 4.7% | 63.3% |
| 10 | Mistral Medium 3.5 |
51%
|
1,244 | 5% | 48.8% |
When the models agree
- All models agreed (629 games)
- 60.3%
- Majority pick, all games
- 54%
- Best single model (Gemini 3.8 Flash)
- 54.8%
- Home team won
- 50.4%
Does AI confidence mean anything?
- Stated confidence under 50% (1,295 picks)
- 41.1% right
- Stated confidence 50–59% (6,921 picks)
- 52.5% right
- Stated confidence 60–69% (3,904 picks)
- 57% right
- Stated confidence 70% and up (320 picks)
- 70.6% right
Accuracy by sport
Sports with fewer than 30 graded games are left out of this table but count in the totals.
| Sport | Games | Majority right | All agreed, right | Home won | Best model |
|---|---|---|---|---|---|
| Football (soccer) | 490 | 51.9% | 65% (197) | 46.3% | Gemini 3.8 Flash · 54.9% |
| Ice hockey | 407 | 60.1% | 62.6% (235) | 56% | MiMo V2.6 Pro · 60% |
| Baseball | 207 | 46.2% | 52.4% (126) | 47.8% | Mistral Medium 3.5 · 52.7% |
| NBA | 79 | 53.2% | 51.5% (33) | 55.7% | Claude Sonnet 5 · 54.4% |
| NFL | 61 | 58.3% | 55.3% (38) | 47.5% | Kimi K3 · 59% |
Use the data
Free to republish under CC BY 4.0. Please credit TuringStats with a link to this page.
Cite as: TuringStats AI Prediction Accuracy Report (last 30 days to 22 October 2026), turingstats.com/ai-accuracy-report
Download every graded pick (CSV)Embed the live leaderboard
Paste this into any article or blog. It updates on its own.
Method
- All 10 models get the same match data for every game (form, venue, head-to-head and, where available, team news), never the betting odds. When the prompt changes, it changes for every model at once.
- Each model returns a winner (or a draw where the sport allows it), a score and a confidence from 0 to 100. Only its latest pick before the game counts.
- A pick is right when the result matches. "Majority" is the result most models picked (ties are left out); "all agreed" means every model picked the same result.
- Results come from official match data. The line-up has been fixed since 7 September 2026; earlier picks are excluded.
FAQ
Which AI is best at predicting sports?
Gemini 3.8 Flash picked the right winner most often: 54.8% of 1,244 picks. Mistral Medium 3.5 was last at 51%, so first to last was only 3.8 points.
Are AI sports predictions accurate?
Better than chance, far from certain. Across 1,244 games the majority pick of the models was right 54% of the time, and 60.3% when every model agreed.
How is the accuracy measured?
Every model predicts the winner (or a draw), the score and its confidence before kick-off. After the final whistle each pick is graded right or wrong. Only the latest pick before the game counts, and every graded pick is in the CSV download.
Can I use these numbers?
Yes. The report and data are free to use under CC BY 4.0: credit "TuringStats" with a link to this page.