41%4,643 bouts
KO / TKO
Striking outcomes classified by recorded knockout, technical knockout, or recognized stoppage descriptions.
Original Fight Atlas research · professional career archiveHow fights end.
How fights end.
What the record can say.
A finish is a result, but it is not automatically evidence of a more aggressive division, a more effective style or a greater chance of a stoppage next weekend. Those interpretations require context that a method label alone cannot supply.
Here we count the outcome descriptions in compiled professional MMA fight histories. The method breakdown and the round breakdown have different denominators. Keeping them separate prevents an appealing chart from turning into a misleading prediction.
647compiled fighter careers
11,427deduplicated bouts
100%with a recorded date
7,514stoppages with usable rounds
Outcome descriptions
Decisions, submissions and stoppages in the archive
Every included bout receives one method category. Counts below describe this dataset, not the entire sport. Rounded percentages may not add up to exactly 100%.
25%2,883 bouts
Submission
Recorded submissions, including named chokes and joint locks.
33%3,731 bouts
Decision
Recorded decisions, regardless of unanimity or score margin.
1%104 bouts
DQ / technical
Disqualifications and the technical categories recognized by the aggregation rules.
1%66 bouts
Other / unknown
Results whose method text does not meet a defined classification rule.
Counting check: 11,427 classified records from 11,427 deduplicated bouts. The source-method grouping is described below so a missing or unusual method is not silently treated as a knockout.
Among classified categories, KO / TKO accounts for 4,643 records (41% of the full method sample). That does not establish that it is the most likely outcome for any particular matchup.
Timing of recorded stoppages
The first round has a different denominator from the fifth
Only source-classified KO/TKO and submission bouts with a recorded stoppage round from one through five appear here. We are counting when finishes were recorded, not estimating the chance of a finish after a fighter reaches a given round.
7,514 stoppages have a usable recorded round in this subset. The bars are scaled relative to the busiest round, not to a universal fight total.
The key interpretive trap
Why fewer late finishes do not mean late rounds are safer
Raw counts are not per-round hazards.
Every scheduled contest begins in round one, but not all scheduled contests have a fourth or fifth round. Others finish earlier. To calculate a round-specific finish probability, we would need the number of bouts that actually reached each round, with the scheduled maximum duration accounted for.
The graph above does not contain that exposure denominator. Dividing fifth-round stoppages by all fights in the dataset would understate the relevant risk for fighters actually contesting round five.
Round counts also tell us nothing by themselves about finishing efficiency, judging criteria, fighter fatigue or competitive quality. We do not present a result that the dataset cannot support.
Method and sources
What this study includes and excludes
Source population: compiled professional career histories available in the Fight Atlas dataset. They include pre-promotion appearances and are not a census of every professional MMA contest worldwide.
Deduplication: mirrored athlete records are collapsed using dated fighter identities, event, method, round and time. Unresolved or incomplete identity links can limit how many records are matched.
Classification: disqualifications, technical results and no contests are recognized before ordinary decisions; named submissions and striking stoppages are categorized separately. Ambiguous descriptions remain Other / unknown instead of being invented into a more precise outcome.
Round chart: decisions, disqualifications, no contests and unknown methods are excluded. Physical stoppages without a usable round from one to five are excluded from the timing chart but remain in the method analysis.
Not a forecast: no ranking, opponent-strength adjustment, rule-set adjustment, time-period stratification or per-round survival denominator is calculated here.
Data checked: September 8, 2026. These are archive observations from the available source material, not live event predictions.