Pick two fighters and see who's likely to win, how, and when,
powered by an Elo model fitted on every UFC bout in history.
VS
wins
wins
Tale of the tape
Method & round
Career stats per fight time
significant strikes per minute (landed / absorbed), accuracy,
takedowns & submission attempts per 15 minutes, knockdowns per 15 minutes
Betting view: fair odds
fair odds = the break-even price for each market. A real bet only has
value if a bookmaker offers higher odds than shown here. The model is fitted on
past bouts only; it knows nothing about camps, injuries or weight misses, and MMA is
the most upset-prone major sport. Not betting advice.
How this predictor works
Every UFC bout ever fought (8,700+ of them) feeds a per-fighter Elo rating:
beat someone strong and your rating jumps; lose to a nobody and it drops,
with knockouts and submissions moving ratings more than decisions. The win
probability comes from the Elo gap adjusted for ring rust and age:
long layoffs and being the older fighter both count against you, with the
size of those penalties fitted on thousands of past bouts and verified on
fights the model had never seen. On top
of that, each fighter carries a finish profile describing how their wins happen
(KO, submission, decision, and in which round) and a victim profile
describing how they tend to lose. Mixing the winner's finishing tendencies with the loser's
vulnerabilities (smoothed toward league-wide base rates so small samples don't
scream 100%) gives the method and round predictions. Career stats (strikes
landed and absorbed per minute, takedowns, submission attempts) are shown for
the eye test. Everything runs in your browser; honest caveat: MMA is the most
upset-prone major sport, one punch changes everything, and even the best models
hover around 60% accuracy. Treat these as informed estimates, not prophecy.
What the numbers mean
Elo rating
A skill score updated after every bout based on the opponent's strength and
the result. League average is 1500; champions sit around 1750–1850. With
the ring-rust and age adjustments the model picked the winner in about 61%
of recent bouts it had never seen, a strong result for a results-only model
in a sport this chaotic.
SLpM / SApM
Significant strikes landed / absorbed per minute of fight time.
High SLpM with low SApM is the striker's ideal; the reverse is a red flag.
TD & Sub per 15 min
Takedowns landed and submission attempts per 15 minutes, the grappling
pressure a fighter brings per full fight's worth of time.
Method & round
How the fight ends, probability-wise: each fighter's own finishing habits
blended with how the opponent tends to lose, smoothed toward league base rates.
Round probabilities are shown for finishes only; "goes the distance" covers
the rest.
⟳ Fetch latest
Downloads the newest community UFC dataset (~10 MB, the same source the
site is built from) and rebuilds every rating, record and profile in your
browser. The dataset is community-maintained and usually updates within days
of an event; fights more recent than its last update can't appear until it
refreshes. The win-probability coefficients stay from the verified backtest.
5-round fight
Main events and title fights run 5 rounds instead of 3; more time means
more finishes and different round math. Flip this on for main events.
Betting view: fair odds
Model probability converted to decimal odds (1 ÷ probability), the
break-even price. A bookmaker's offer is only mathematically good if it's
higher than the fair odds shown. Not betting advice.
♥ Support the project
SportMath is free, open and runs entirely in your browser. If it made the
fights a little more fun, you can chip in to keep it running and growing.
Any amount is hugely appreciated!