KellyBench: leaderboard
Long-horizon betting on KellyBench (2026): the agent builds models from historical football data, lineups and public odds and bets through the 2023/24 English Premier League season from a 100,000 GBP bankroll, five seeds. Average return on investment in percent, where -100 is ruin; higher is better.
Metric: Average return on investment (%, -100 is ruin, unbounded above) over five seeds on KellyBench's sequential simulation of the 2023/24 English Premier League season, where the agent builds betting models from historical data, lineups and public odds and places bets matchday by matchday from a 100,000 GBP bankroll (ReSum harness, maximum reasoning budget with interleaved thinking, 500-1000 tool calls per episode), five seeds; higher is better. Source: arxiv.org. Saturation forecast: Around August 2027. 5 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.4 | xHigh | -7.9 | #138 |
| 2 | Claude Opus 4.6 | Max | -11.2 | #114 |
| 3 | GLM-5 | Thinking | -51.6 | #276 |
| 4 | Gemini 3.1 Pro (Preview) | -66 | #102 | |
| 5 | Kimi K2.5 | Thinking | -89.6 | #272 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=kellybench · How It Works · Data refreshed daily, snapshot 2026-11-22.