Tell 01
Chin Decay
Does a first KO loss mean anything the odometer didn't already say?
53% more likely to be KOed again. Half of them get cut before they can prove it.
- Sample
- Every UFC bout since 2005, 8,799 of them
- Confidence
- 27% to 82% more likely
- The catch
- survivors only
Last run after UFC Fight Night: Gamrot vs. Salkilld, Aug 21, 2026.
What everyone believes
Junior Dos Santos went eleven fights in the UFC without being stopped. Then Cain Velasquez did it in October 2013, and six of the eleven that followed ended the same way.
Everyone watched and drew the obvious conclusion. The obvious conclusion has a problem: you remember Dos Santos because the bookings kept coming. About 610 fighters took a first KO loss and never got three more UFC fights, and nobody built a theory out of them.
Commentators say the chin never comes back, cornermen build gameplans on it, and the betting chat treats a first KO loss like a diagnosis. It's one of the few things everyone in the sport agrees on, which is usually a good reason to check it.
The folk version smuggles in two other facts. Fighters who've been KOed are older than fighters who haven't, and they've eaten more punches getting there. A chin that goes right when a fighter turns 35 with 400 strikes absorbed is just a career arriving on schedule. So the real question is narrower: does the first KO loss tell you anything the odometer didn't already?
The comparison
Share of fights that end in a KO or TKO loss from strikes.
Never KOed
11.3%
After a first KO
17.3%
53% more likely27% to 82%
- Age
- Never KOed31.9
- After a first KO31.9
- UFC fights
- Never KOed9.0
- After a first KO9.4
- Strikes eaten
- Never KOed331
- After a first KO326
- Fought around
- Never KOedAug 2018
- After a first KOJul 2018
Knockdowns absorbed, per 100 strikes eaten
Never KOed
0.47
After a first KO
0.63
33% more knockdowns absorbed per strike eaten. It shows up even in the fights they survive.
What we measured
Every UFC fight in the database, 8,799 bouts across 783 events, gated to 2005 onward. A chin event is a KO or TKO loss from strikes and nothing else. Slam KOs don't count. Cuts don't either.
Two outcomes. The headline is how often a fighter loses by KO again. The sensitive one is knockdowns absorbed per 100 significant strikes absorbed, which moves even when a fighter survives the round.
Does it compound?
Each row is a fighter's risk after that many KO losses, next to a matched fighter with none. The line at 1 is no difference. A bar that touches the line means we can't tell it apart from no difference at all.
after 1st KO
4,258 fights
after 2nd KO
1,423 fights
after 3rd+ KO
508 fights
Why it doesn't compound
The folk theory says each KO makes the next one easier. Only the first step is measurable, and before reading that as chins healing, look at who is left. A fighter still on the roster after three KO losses is either unusually durable or unusually good, because everyone else got cut. The sample after a third KO is 508 fights.
Who got cut
Out of every 100 fighters, how many never fought three more times in the UFC.
29Never KOed
46Any other loss
50First KO loss
Count the ones who left and the gap nearly doubles.
What this can't tell you
Whether the KO broke something or revealed it. A fighter may carry a fragile chin into the sport and the first KO is the day it shows. The data supports "a first KO loss marks a more fragile fighter than the odometer suggests." It stays silent on why.
For picking a fight, that gap matters less than it sounds like it should. The same goes for the attrition: some of that 21 point gap is just what losing does to a career, and a fighter who is about to get cut is still a fighter you'd think twice about backing. A marker earns its place by telling you what happens next. Whether it caused the thing or only saw it coming first is a better question and a much harder one, and you can pick a card without settling it.
Every headline number here is computed on the half that stayed, which is the tougher half. Run it without that filter and fighters come out 94% more likely rather than 53%. So 53% is the floor of the effect, and the UFC's matchmaking already removed the fighters it would have been largest for.
Since we published
Nobody new has fallen into this one yet. Check back after the next card.
The fine print
Sample
UFC fights 2005-01-01 to 2026-08-08. Main comparison: 4258 treated fighter-fights from 654 fighters vs 6816 matched control fighter-fights (Kish effective N 1506) across 776 strata; 1000 treated fighter-fights pruned off common support. Attrition cohort (24mo window): 1219 first-KO fighters.
Method
Risk-set coarsened exact matching: fighter-fights after the Nth KO-by-strikes loss vs fighter-fights of fighters with zero prior KO-strike losses as of that fight, matched exactly on age (2yr bins), UFC fight count bins, cumulative sig strikes absorbed quintiles, weight class group, and era; all covariates as-of pre-fight from dcc.asof_fighter_stats (leakage-tested upstream). Outcomes: KDs absorbed per 100 sig strikes absorbed (pooled ratio of sums, stratum-weighted) and KO/TKO-by-strikes loss rate per fight. Fighter-level cluster bootstrap, 2000 replicates with full rematching per replicate, percentile 95% CIs. 1-NN sensitivity check point-estimate only. Treated fighters gated to 3+ post-event fights; attrition reported separately with a 24-month censoring window. Runner: npm run analyze:chin-decay.
Assumptions
Chin event = KO/TKO loss by strikes only (injury TKOs, doctor/corner stoppages, retirements excluded; full audit table in scripts/analysis/chin-decay-method-map.json). The first KO marks rather than causes any change; matching on observables cannot separate accumulated damage from revealed fragility. Fighters are treated as independent sampling units although they fight each other. UFC bouts only; what happens outside the UFC is invisible to the data.
The commentators were right. They just never mentioned that half the proof gets cut before it can testify.