UFCPredictor

What Actually Predicts UFC Fight Outcomes

An effect-size ranking of metrics that historically separate winners from losers, derived from 7,000+ fights.

UFC fight prediction is full of folk wisdom: reach matters, southpaws are tricky, wrestlers beat strikers. We ran the numbers across the full UFC stats history. Some folk wisdom holds up; most of it doesn't. The list below ranks predictors by Cohen's d (standardized mean difference between winner and loser), not raw correlation.

Top predictors, ranked

#1
Head-strike accuracyLarge (Cohen's d > 0.8)

The percentage of significant strikes a fighter lands to the head is the single most predictive in-fight metric. Winners hit the head ~11pp more accurately than losers on average.

Betting implication: Fighters with high career head-strike accuracy are systematically undervalued. Check the Strike Breakdown card on each fight page.
#2
Ground striking accuracyLarge

Once a fight hits the mat, the fighter on top who can land accurately wins almost every time. Median winner accuracy from ground position is 20pp above the loser.

Betting implication: Bet wrestlers/grapplers who can land ground-and-pound, not pure BJJ artists who only threaten submissions.
#3
Time spent on the groundLarge

Winners spend ~7pp more of fight time on the ground than losers; losers spend ~8pp more at distance. Pure kickboxers and karate-style strikers underperform their cardio.

Betting implication: Avoid pure distance strikers (Karate, Tae Kwon Do backgrounds) unless they have exceptional defensive grappling.
#4
Overall significant-strike accuracyMedium-Large

Total sig-strike accuracy correlates strongly with winning. Hitting clean matters more than hitting often.

Betting implication: Look at SLpM × Str_Acc, not just SLpM. A 5.0 SLpM fighter at 38% lands the same volume as a 4.0 SLpM fighter at 47%, but with twice the energy expended.
#5
Knockdowns scoredMedium

Knockdowns are rare but decisive. Most fights have zero, but when they happen they correlate ~1:1 with winning the fight.

Betting implication: Less useful for prediction (rare event), more useful for live betting and live round scoring.
#6
Control timeMedium

Time spent controlling the opponent on the ground or against the cage strongly favors the controller. Judges reward it heavily on close cards.

Betting implication: Edge for grapplers in decision-heavy matchups. Check the Striking Compare card for TD_Avg.
#7
Takedown accuracyMedium

Landing takedowns at high accuracy — not just attempting them — predicts winning. Failed takedowns burn energy and concede position.

Betting implication: TD_Avg combined with TD_Acc beats raw takedown attempts. Wrestlers who shoot blindly often lose decisions.
#8
Strike-defense %Small-Medium

Winners absorb ~0.2 fewer strikes per minute on average. The effect is modest but consistent.

Betting implication: Look at SApM (strikes absorbed per minute) on every fighter page. Sub-3.0 SApM is elite defense.

What actually doesn't predict outcomes

Height advantage

Median height differential between winner and loser is zero. Tall fighters don't systematically beat short fighters.

Reach advantage

Reach matters in specific matchups (long strikers vs short brawlers) but doesn't predict outcomes on average.

Submission attempts

Frequent sub attempts without finishing often lose rounds. Attempting > scoring on judges' cards.

Reversals

Reversals in scrambles are rare and have negligible effect size — judges and fans both notice them less than expected.

Stance matchup win rates

Cross-stance fights from the historical record, minimum 5 sample size per matchup. Rare stances give a small but consistent edge through unfamiliarity.

MatchupEdge favorsWin rate
Switch vs SouthpawSwitch54.1%
Switch vs OrthodoxSwitch52.9%
Southpaw vs OrthodoxSouthpaw53.5%

The red corner bias

The red corner wins approximately 65% of UFC fights. This is NOT because the corner is lucky — it's because the UFC assigns red to the higher-ranked fighter, champions, and the “A-side” of every promotion. The corner color is downstream of who the matchmakers think will win.

When our model favors the blue corner, it's explicitly bucking the matchmaking prior — which is exactly when contrarian bets tend to have value. Watch the “Vegas disagreement” field on every prediction card for these spots.

Methodology

Cohen's d = (mean of winner−loser deltas) / (std of deltas). We compute it across every recorded UFC fight with complete statistics, decisive outcomes only (draws and no-contests excluded).

The model that powers our predictions weighs these factors plus 30+ engineered features including recency, layoff, ELO, and opponent quality.