2026 Leaderboard
Ties don't count as a win or a loss. Missed picks don't count either — they show as a gap in the pick total.
The AIs
| # | Contestant | W | L | Win % | vs spread | Agrees | Pending |
|---|---|---|---|---|---|---|---|
| – | Always the betting favoriteBaseline — the one to beat | 101 | 46 | 68.7% | — | 100.0% | |
| 2 | GLM-5.3Z.ai (Zhipu) | 97 | 50 | 66.0% | 76–69 | 77.7% | |
| – | AI consensusMajority of the AIs (splits skipped) | 80 | 45 | 64.0% | — | — | |
| 4 | Mistral LargeMistral AI | 94 | 53 | 63.9% | 73–72 | 77.0% | |
| 5 | Nemotron 3 UltraNVIDIA | 84 | 48 | 63.6% | 65–65 | 74.4% | |
| 6 | GPT-OSS 120BOpenAI (open-weight) | 93 | 54 | 63.3% | 74–71 | 77.7% | |
| – | Always the home teamBaseline | 70 | 77 | 47.6% | — | — |
Grey rows are baselines — simple strategies an AI has to beat to prove it adds anything. Always the betting favorite is the real test: about two NFL favorites in three win outright, so a model near that number has matched the bookmaker rather than beaten it. Agrees is how often a pick was simply the favorite, and vs spread asks the harder question — roughly 52.4% is break-even there.
Players
| # | Player | W | L | Win % | Points | vs spread | Pending |
|---|---|---|---|---|---|---|---|
| 1 | Supreme Ruler@supremeruler | 85 | 48 | 63.9% | — | — | 14 |