LLM-SoccerArena - Brier Score: leaderboard
LLM-SoccerArena records timestamped model forecasts of football matches before the outcomes are known; this board covers all 104 matches of the 2026 FIFA World Cup, forecast 24 hours before kickoff under the probabilistic prompt. The score is the unscaled three-class Brier score of the home, draw and away probabilities against the 90-minute result, so it runs from zero to two and lower is better.
Metric: Brier score (0 to 2, lower is better; unscaled three-class Brier score of the home, draw and away probabilities against the 90-minute result, mean over the 104 matches of the 2026 FIFA World Cup; forecasts made 24 hours before kickoff under the probabilistic prompting strategy, averaged over the open-book and closed-book conditions). Source: arxiv.org. Saturation forecast: Around August 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 0.51 | #94 |
| 2 | GPT-5.5 | 0.52 | #49 |
| 3 | DeepSeek V4 Pro | 0.52 | #188 |
| 4 | Grok 4.3 | 0.52 | #333 |
| 5 | Qwen 3.7 Max | 0.53 | #145 |
| 6 | Claude Opus 4.8 | 0.53 | #56 |
| 7 | Mistral Large 3 | 0.55 | #817 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=llm-soccerarena-brier-score · How It Works · Data refreshed daily, snapshot 2026-11-13.