Sat, November 7 at 5:00 PM EST — Meta APEX · Las Vegas · NV · 📺 Paramount+. Every fight shows the WizLine (our model's win probability and its fair odds) directly above the books' line, so you can see where we agree and where we don't. Tap a fight for the tale of the tape behind the WizLine.
⚙ New model, graded in the open. Our fight model is calibrated (when it says 65%, that happens about 65% of the time) but it hasn't proven it can out-price the books — so nothing here is a pick. A ◆ lean just marks where we and the books disagree the most; we're tracking every card publicly to find out if those disagreements mean anything. Research, not betting advice — 21+.
How has the UFC fight board graded this season?
The UFC fight board is 65-47 between Sep 15, 2026 and Oct 31, 2026, a 58% hit rate over 112 graded calls.
That is 9 more wins than a coin flip would have produced over the same 112 decided calls.
A pick is whichever fighter our model gives more than 50% before the card, graded at the result; nothing is staked, so this is a hit rate and not a return.
That covers the last 12 cards, every fight the model published a number on, small-sample bouts included. On the out-of-sample backtest, 2,511 of 4,099 fights from 2019 to 2026 went the model's way (61.3%), and on the 2,718 with a closing line the books were the sharper forecaster — log loss 0.601 against our 0.653.
Graded through · 112 calls
Main Card
Heavyweight · 5 RDS≈ 11:20 PM ET◆ we like Walker +10.5% vs the books⚔ compare
fight lengthwe say 64% it goes past 2½ rounds · books say 45% (O +110 / U -140) ◆ we lean Over
how it endsKO/TKO 31% · Submission 11% · Decision 58%
WizLine = the odds our model's win probability implies. Best book = the longest price across the ~7 US books we track, with the books' percentage shown vig-free. The bar under each matchup runs from the red corner (left) to the blue corner (right), split at our win probability. Small sample = a fighter with almost no UFC tape, where the WizLine is a rough guess. Lines move — always confirm at the book. Not betting advice — 21+, play responsibly (1-800-GAMBLER).
Build a same-fight parlay — winner, method and round together, priced off one joint rather than three separate numbers.
Soon
How the fight model works
Every fighter carries a strength rating that updates after each bout and weights the quality of the opponent — beating a ranked contender moves it far more than beating a debutant. Two ratings produce a baseline win probability, which is then adjusted for the stylistic matchup: the striking exchange implied by output against absorption, and the grappling exchange implied by takedown rate against takedown defense. Age and layoff factor in, because both bite harder in this sport than most.
That probability is converted to fair odds and shown directly above the best price across the US books we track, with the books' number stripped of vig so the two are actually comparable. How the pieces are weighted is ours; the output and its record are public.
Why there are no picks here
Because the model hasn't earned them. It's calibrated — when it says 65%, that happens about 65% of the time — but calibration and profitability are different things. A well-calibrated model that agrees with the market everywhere it matters has no edge, and that's the honest description of where this one sits today.
So a ◆ lean marks where the WizLine and the market's diverge most, and nothing more. We're tracking every card in public to find out whether those disagreements mean anything. If they don't, this page will keep saying so.
That's the same conclusion we reached on several baseball prop markets — home runs, total bases, hits+runs+RBIs — where a full review found no edge worth acting on. The boards stay up because the projections are useful; they don't produce plays because the market prices them well. See how MatchWiz works for the fuller version of that argument.
Where the model is weakest
Short tape. A fighter with two UFC bouts has a rating built on almost nothing. Those rows are marked small sample, and the number attached to them is a rough guess wearing a decimal point.
Late changes. Fights fall out, opponents change on days' notice, and a fighter missing weight materially changes the matchup without changing any stat line the model reads.
Variance is enormous. One punch settles a fight that the better fighter was winning. Over a card, results tell you very little; over a season of cards, the record starts to mean something. That's the horizon everything here should be judged on.
Every fighter carries a strength rating that updates after each bout and weights the quality of the opponent, so beating a ranked contender moves it far more than beating a debutant. Two ratings give a baseline win probability, which is then adjusted for the stylistic matchup — striking output against absorption, takedown rate against takedown defense — plus age and layoff. That probability is converted to fair odds and shown above the best price across the US books we track, vig-free.
Does MatchWiz post UFC picks?
No. The model is calibrated — when it says 65%, that happens about 65% of the time — but calibration and profitability are different things, and it has not proven it can out-price the books. A lean marks where the WizLine and the market diverge most, nothing more. We track every card publicly to find out whether those disagreements mean anything.
Where is the UFC model weakest?
Fighters with short tape — two UFC bouts gives a rating built on almost nothing, and those rows are marked as small samples. Late opponent changes and missed weight also alter a matchup without changing any stat line the model reads. And variance is enormous: one punch settles a fight the better fighter was winning, so a single card tells you very little.
What do the fighter leaderboards measure?
Rate stats from per-fight lines rather than records — striking output and accuracy, strikes absorbed per minute, takedowns per 15 minutes, takedown defense, submission attempts, and finish rate. A fighter qualifies once he has fought in the last 24 months with at least five UFC bouts on our tape, because rate stats over one or two fights are noise.
Are these UFC betting picks?
No. Every number here is a research signal from a statistical model, published with its graded record. Not betting advice, and no outcome is guaranteed. 21+, please play responsibly (1-800-GAMBLER).
How has the UFC fight board graded this season?
The UFC fight board is 65-47 between Sep 15, 2026 and Oct 31, 2026, a 58% hit rate over 112 graded calls. That is 9 more wins than a coin flip would have produced over the same 112 decided calls. A pick is whichever fighter our model gives more than 50% before the card, graded at the result; nothing is staked, so this is a hit rate and not a return.
How is each UFC fight call graded?
A pick is whichever fighter our model gives more than 50% before the card, graded at the result; nothing is staked, so this is a hit rate and not a return
