"A fighter who has been knocked out is never the same again." Anyone who follows the sport has heard it many times. Having your lights put out by strikes leaves something behind, and the next time out you go down again. Is that true?

The previous article measured how well this site's rating predicts actual results. This one does the opposite. It looks for the places where the rating stops working.

All that is used is the rating the site carried into that bout. For each bout the two ratings give an expected win probability, and within each group the average expectation is set beside the actual win rate. A positive difference means the site was underrating those fighters; a negative one means it was overrating them.

One thing has to be said first. This rating already treats a loss by KO or TKO more heavily than a loss on the cards. A fighter who has been stopped has already had his rating cut, at that moment. What is measured here is not whether losing by knockout makes a fighter worse -- it is whether the cut was enough.

The subject is limited to bouts where both men had at least five prior fights on the rating.

36,516 fighter-views, 1980 to 2026

For a fighter coming off a loss on the cards, the rating is very nearly perfect: the difference is 0.1 of a point. But coming off a KO or TKO, the actual win rate falls 2.7 points below expectation. Being submitted sits between the two.

And nothing of the sort happens on the winning side. After a knockout win it is +1.1, after a submission win -0.4, after a decision win +0.5. All close to zero. This is not a story about momentum. It is a story about the fighter who was stopped.

The gap between coming off a knockout loss and coming off a decision loss is 2.6 points. One bout enters as two rows, one per corner, so counting them plainly understates the uncertainty. Resampling 2,000 times with the bout as the unit, and taking the conservative side, the 95% range is 0.7 to 4.5 points. It does not include zero.

Split by the gap since the previous bout (bouts in brackets)

Up to twelve months, the way the previous bout ended explains it. Only the fighters coming off a knockout sink; those coming off a decision do not. Past twelve months the order swaps, and coming off a decision is worse. What is at work in that band is not the knockout but a different problem: the rating thinks too highly of a fighter who has been away a long time.

So what this article can speak about is the year after being stopped, and nothing beyond it.

The 1990s row rests on few bouts and is there for reference only, but the difference between the figure after a KO/TKO loss and the figure after a decision loss runs the same way in all four decades (8.6 / 3.6 / 2.5 / 2.2 points, in order). Changes to the rules and to refereeing in recent years cannot account for that on their own.

Mirko Cro Cop against Cheick Kongo, 8 September 2007. Four months earlier he had been knocked out by Gabriel Gonzaga. The ratings were 2155 to 1650, an expected win of 91.1%. He lost on the cards.

Eddie Alvarez against Timofey Nastyukhin, 31 March 2019. Eight months earlier he had been stopped by Dustin Poirier. 2153 to 1642, an expected win of 91.4%. He lost by TKO.

Junior dos Santos against Yorgan De Castro, 20 May 2022. Seventeen months earlier he had been stopped by Ciryl Gane. 2044 to 1536, an expected win of 89.4%. He lost by TKO.

The other direction is worth showing too. Yushin Okami against Kim Jae-young, 2 December 2023. Twelve months earlier he had been stopped by Aung La Nsang -- and at 1750 to 880, an expected win of 93.3%, he took the decision. 2.7 points is a quantity of that size. It does not call individual bouts.

The cause cannot be separated. Either the rating is not cut far enough, or being stopped leaves something behind. Both produce the same number.

Matchmaking is not neutral. Somebody watched that knockout before deciding what came next. This is not about fighters who have been knocked out; it is about fighters who were given a bout after being knocked out.

Debuts are a different problem. A fighter with nothing on the rating comes in at an expected 33.5% against an actual 21.5% -- out by 12.0 points. That has nothing to do with knockouts; it is the cost of knowing nothing.

We hold no data on age, weight cuts, injuries or training environment. The cause cannot be laid at any individual fighter's door.

Seeing 2.7 points, one could punish a knockout loss another notch. We will not. While it is unclear whether the residue is an insufficient cut or the damage itself, moving a coefficient would break the parts that currently work along with the part that does not. As the previous article showed, this rating predicts well overall. Leaving what is not understood visibly unresolved is more accurate than papering over it.

What was learned goes into how the numbers should be read instead. For a fighter in the year after being stopped, discount this site's figure a little.

[Correction] 20 September 2026 — the table by decade had been computed on a different population from the rest of the article (all decisive bouts). It has been recomputed on the same population used everywhere else: both fighters with 5+ prior bouts. The conclusion — that the difference runs the same way in all four decades — is unchanged.

The figures in this article are based on data as of 23 September 2026. The data is updated weekly, so the last digit may differ from the current tally.