2,000+
Fighters Tracked
600+
Events Analyzed
72%
Prediction Accuracy
Upcoming Events
The Official Calendar
Fight Archives
Rewatch history
The Official Calendar
Rewatch history
The dedicated UFC portal for rosters, records, and every fight card we track. Curated intelligence, live data snapshots, and decades of fight history.
Two honest scoreboards: the backtest for the model version running right now — measured on recent fights it was never trained on — and the live record, where every public pick is graded against the real result. Wins and misses, no retroactive edits.
Current model — held-out backtest
v1.2 · live since Jul 2, 2026Winner accuracy
65.8%
393 held-out bouts, Oct 2025 – Jun 2026
On bouts with betting lines
71.5%
vs the market favorite's 69.7% over 170 such bouts — and better calibrated
Calibration error
2.3%
promotion gate requires < 8%
Backtest = the model graded on 393 recent bouts it never trained on, using the same features and model that run in production. On the 170 of those that had betting lines it holds its own with the market (71.5% vs 69.7% — the edge is within noise on a sample this size, but it prices those bouts with lower error); on the 223 bouts with no line it works from fight data alone and lands at 61.4%. Method (v1.0) and round (v1.0) are much harder calls, at 48.2% and 52.0%. How this works: methodology.
Live record
Winner
84.6%
13 graded bouts
Method
23.1%
13 graded
Round
—
3 graded
Bouts
13
each counted once
Small sample. Only 13 bouts have been graded since launch — a handful of events can swing these numbers hard in either direction. Until a few hundred bouts are in the books, the held-out backtest above is the more reliable estimate of the model's skill. Every graded bout so far was called by an earlier version; the current v1.2 model went live Jul 2 and hasn't had a graded event yet.
| Model | Bouts | Winner | Method | Graded |
|---|---|---|---|---|
| v1.0 | 13 | 85% | 23% | Jun 2026 |
Calibration
When we say a fighter is, say, 70% to win, do they actually win ~70% of the time? Closer = better. This table fills in as more bouts resolve.
| Confidence band | Predictions | We said | Actually won |
|---|---|---|---|
| 50–60% | 3 | 58% | 67% |
| 60–70% | 10 | 63% | 90% |
| 70–80% | — | — | — |
| 80%+ | — | — | — |
Want the model's pick, confidence, and method breakdown for every upcoming bout — plus an interactive AI analyst to talk it through?
Get AI Predictions — ChampionFor analysis & entertainment only. Predictions and model-edge figures are informational — they are not betting or financial advice, and past accuracy does not guarantee future results. We are not responsible for any losses from betting or gambling activity. You must be 18+ (21+ in some jurisdictions) to gamble. If gambling is a problem for you or someone you know, call 1-800-GAMBLER or visit ncpgambling.org. Full disclaimer
Curious how the model works? Read our methodology.