Original Research · Updated Jun 28, 2026
Blueprint MMA vs the Market
Across 56 settled UFC and MMA fights, Blueprint's AI War Room Committee hit 67.9% of winners at a 0.2592 Brier score, and its published probabilities diverged from the betting-market closing line by an average of 11.2 pts over a 33-fight benchmarked subset. Every pick was published before the fight and graded publicly after it, so this is computed from the settled ledger, not a hand-picked sample.
Machine-readable companion: /blueprint-vs-market.json
Accuracy by confidence tier
| Tier | Sample | Hit rate |
|---|---|---|
| Lock (85%+) | 0 | 0% |
| High (70-84%) | 7 | 42.9% |
| Medium (60-69%) | 21 | 66.7% |
| Low (50-59%) | 24 | 75% |
Full calibration audit and settled ledger: Track Record.
Cite this
Blueprint MMA's AI fight-prediction engine has hit 67.9% of winners at a 0.2592 Brier score across 56 settled fights, with its published probabilities diverging from the betting-market closing line by an average of 11.2 points (Blueprint MMA, blueprintmma.com/blueprint-vs-market).
FAQ
How accurate is Blueprint MMA versus the betting market?
Across 56 settled fights, the engine hit 67.9% of winners at a 0.2592 Brier score. On the 33-fight subset with a stored closing line, its published probabilities diverged from the market by an average of 11.2 points.
Why benchmark against the closing line?
The closing line is the sharpest publicly available market estimate of a fight outcome. Comparing the engine to the close — rather than to the opener or to opinion content — is the hardest honest test of whether a model adds signal.
Is this number cherry-picked?
No. Every prediction is published before the fight and graded on the public Track Record after it settles. This study is computed from that same settled ledger, not a hand-picked sample.