Promoted by Mike Fagan.
Here's your winner for round two, folks. Congrats to YCD, who wins bragging rights for the next month. Honorable mentions go out to last month's winner PistonHyundai and JCS_FM from Fight Matrix.
If there's a topic you want to see covered by the research contest, leave it in the comments.
The relationship between age and fight outcomes turns out to be surprisingly complex, with 'youth' and 'age' exerting somewhat different effects on outcomes. But let’s start with the basics.
On average, winners are just over a year younger than losers (380 days), and the younger fighter won 57% of the fights in the dataset. Both these trends are extremely unlikely to be due to chance (both p < .0001). When we subdivide by method, we see more variation. The age gap is largest for Kos/TKOs (~1.5 years, 61% of fights), right around the overall mean for submissions (~1 year, 57% of fights), and smaller for decisions (~½ a year, 54% of fights) for decisions. Even considered independently, each of these differences is statistically reliable, and the differences between them are also reliable.
Thus, younger fighters tend to win regardless of win method, but this difference is largest for KO/TKOs, intermediate for submissions, and smallest for decisions. These results are displayed in Figure 1, which shows the means and the 95% confidence intervals around the means (roughly speaking, this means that if we repeated this experiment with new fight data, there is a 95% chance that the average from that new experiment would fall within these intervals).
How do we interpret these differences? One thing that immediately becomes clear is that an analysis focusing on the age difference can’t identify whether the most important factor is the benefit of youth (athleticism, explosiveness) or the detriment of age (declining abilities, being shop-warn, chinnier, etc), or both. To investigate the joint impact of both fighters’ ages, we switch to the context of regression, in which we can estimate the impact of one variable when the effect of other variables are controlled.
In our analysis, we begin by predicting the probability of the younger fighter winning as a function of the age of the younger fighter and the age of the older fighter (for the curious, this is logistic regression, which predicts the probability of a categorical outcome, i.e. of winning a fight). This analysis reveals that both fighters’ ages matter, and matter at very similar levels. Basically, we can conclude that for every year of age gap, whether driven by the age of the younger fighter, the older fighter, or both, the younger fighter is about 1.1 times more likely to win (that is, has about a 10% greater chance of winning).
Next we can factor in the method, to see whether it is variation in older or younger fighters that accounts for differences in victory method. Doing so reveals something quite interesting: Variation by type of victory depends on the age of the older fighter, not on the age of the younger fighter (in statistical jargon, the effect of the age of the younger fighter is solely a main effect, while the effect of the age of the older fighter also interacts with method of victory, that is, it varies depending on the victory method). More precisely, we can say that as the older fighter’s age increases, his chance of getting knocked out increases substantially, his chance of getting submitted increases but somewhat less, but his chance of losing a decision stays relatively flat. For ease of visualization, Figure 2 summarizes these relationships by showing the probability of victory for the younger fighter as a function of the gap in age.
So what does all this tell us? Being young is good—younger fighters are more likely to win fights across the board. But the downside of aging is more interesting—most prominently, older fighters are considerably more likely to lose via KO/TKO, which could be interpreted as evidence of a declining chin or simply less ability to get out of bad situations. Interestingly, the one area in which the youth advantage is minimal is decision victories; over a longer fight, one can imagine that experience is able to counteract most of the advantage of youth.
One important limitation needs to be mentioned, though. I would argue that MMA is likely to have strong ‘cohort effects’; that is, the group of younger fighters have advantages that go beyond youth. Notably, by coming up during a more established period of MMA, they may (on average) come from better camps with more sophisticated and thorough training and more consistently high competition. If this is right, it means there is a confound between age and training quality, such that younger fighters may have received better training. To the extent that this is the actual causal factor driving the results reported above, what’s reported here may exaggerate the true effect of age.