Left-handedness is associated with greater fighting success in humans

Left-handedness is a costly, sexually dimorphic trait found at low frequencies in all human populations. How the handedness polymorphism is maintained is unclear. The fighting hypothesis argues that left-handed men have a negative frequency-dependent advantage in violent intrasexual competition giving them a selective advantage. In support of this, many studies have found that left-handed men are overrepresented among modern professional fighters, but studies typically find no difference in fighting success between left and right-handed fighters. We studied over 13,800 professional boxers and mixed martial artists of varying abilities in three of the largest samples to test this hypothesis to date, finding robust evidence that left-handed fighters have greater fighting success. This held for both male and female fighters, and for both percentage of fights won and an objective measure of fighting ability. We replicated previous results showing that left-handed fighters are strongly overrepresented in professional combat sports, but left-handed fighters did not show greater variance in fighting ability, a hypothesis suggested in previous studies. Overall we find strong evidence consistent with the fighting hypothesis.

advantage would lead to an increase in the frequency of left-handed fighters until the advantage exactly offsets the costs of being left-handed, which may be small in populations with access to modern healthcare 2 . Thus, detecting the effect of left-handedness on fighting success may require very large sample sizes. Likewise, top fighters by definition have little variance in fight success, making detecting relationships in these datasets difficult. Top fighters may also have encountered enough left-handed opponents that any advantages due to unfamiliarity would be diminished. Evidence for whether left-handed fighters perform better than right-handed fighters is thus inconclusive. The present studies tested whether left-handed fighters are better than right-handed fighters in 3 large samples consisting of professional fighters at a variety of ability levels. In particular, one of our samples comprised the majority of boxers professionally active at the time of writing.
Previous studies also used win percentage records, number of wins, or ranking from a single tournament as proxies of fighting ability. These may fail to capture long term fighting performance, particularly for fighters with 0 losses, (which gives a win percentage of 1 regardless of the number of fights). These metrics also do not weight wins by quality of opponent, and fail to include how fighters beat their opponent. For example, winning a boxing match by having a better judges' score after 10 rounds may indicate less formidability than a win by knockout in the first round. In our samples we excluded fighters who had few fights, and additionally compared left and right-handed boxers using their BoxRec score. The Boxrec score is a comprehensive measure of fighting ability, where the points awarded for winning fights are weighted by the swiftness and manner of victory (e.g. knockout or judges' decision) as well as the skill of the opponent. This addresses the concerns with use of win percentage as a measure of fighting ability. See http://boxrec.com/media/index.php/BoxRec_Ratings_Description for a description of how a BoxRec score is calculated.
The fighting hypothesis for the evolution of left-handedness is based on male-male contest competition, but there is no reason to expect the frequency-dependent advantage of left-handedness in combat to be confined to males. However, there have been almost no studies of the success of left-handed female fighters. To remedy this, one of our samples consisted exclusively of female professional boxers and our sample of MMA fighters included women as well as men. Additionally, comparison of the left-hand advantage in male and female fighters allows us to investigate negative frequency-dependence. If there are fewer left-handed female fighters than male ones, the fighting hypothesis might predict left-handed female fighters would have a larger advantage.
Lastly, a previous study by Dochtermann et al. 22 demonstrated that left-handed MMA fighters show greater variance in probability of winning a fight than right-handed fighters. They argue that this is because the advantage left-handed fighters possess increases the probability that they will reach professional level compared to right-handers even if they are less skilled. We attempted to replicate this finding in our samples.
In summary, we investigated representation and fighting success of left-handers in 3 of the largest samples tested thus far, consisting of professional male and female boxers and MMA fighters of varying abilities. For boxers, we also tested the difference between left and right-handers in BoxRec scores, a holistic measure of fighting ability. Our study provides one of the most powerful tests of the fighting hypothesis attempted to date.

Samples
Our first sample comprised every male professional boxer in the world listed as 'active' on www.boxrec.com at the time of writing (January 2019). BoxRec.com is a community-run boxing website that aims to document the careers of every professional boxer to have ever taken part in a recorded match. Boxers are listed as active if they have fought in an officially licensed bout in the past 12 months. Our second sample comprised all professional female boxers listed on www.boxrec.com for which stance data was available. For the female sample we included both active and retired boxers, as this ensured a large sample. Finally our third sample comprised all the MMA fighters, both male and female, listed on ufcstats.com at the time of writing. ufcstats.com is a comprehensive, respected MMA database that is the official statistics provider to the Ultimate Fighting Championship (UFC).
We excluded fighters with fewer than 5 fights, as their fight record is too preliminary to accurately reflect their fighting ability. We additionally excluded fighters with a win percentage of 20% of less. Many of these fighters are likely what are referred to in boxing slang as "tomato cans": uncompetitive fighters who take matches with opponents they have little chance of beating simply to earn money. They are often matched against young up-and-coming fighters in order to gain the fighter more wins on their record. For these reasons their win percentage and Boxrec score may not reflect their fighting ability, and as such they were excluded. The

Results
All statistics were run in R 35 , and all data and analysis code is available on the open science foundation (https:// osf.io/x3unr/). For all samples, the number of fights left-and right-handed fighters had participated in, fighter ages, win percentages and BoxRec scores were all non-normally distributed, so nonparametric statistics were used throughout.
A Mann-Whitney U test showed that left-handed male boxers did not differ in age (p = 0.88) from right-handed boxers. For female fighters, age was not analysed as some boxers were retired, deceased or not currently active. Age was not available for the MMA fighters. Mann-Whitney U tests found no significant differences in number of fights between left-and right-handed fighters among male boxers (p = 0.40) and female boxers (p = 0.69) though left-handed MMA fighters did have marginally more fights than right-handed fighters (p = 0.047). Additionally, t-tests showed that left-and right-handed MMA fighters did not differ in overall weight, height or arm length (also known as "reach") (all p > 0. 19). This data was not available for either sample of boxers. ( www.nature.com/scientificreports www.nature.com/scientificreports/

Are left-handers overrepresented among professional fighters?
To test whether left-handed fighters were overrepresented in our samples we ran three separate, one-tailed binomial tests against percentages of left-handers found in a large representative, western population 6 . We tested the percentage of left-handed male boxers against the percentage of left-handed men (12.6%) and female boxers against the percentage of left-handed women (9.9%) in the general population 6 . The MMA sample included both male and female fighters, so was tested against the percentage of left-handed men, as this was the most conservative test of our hypothesis. Table 1 shows that left-handed fighters were significantly overrepresented in all three samples (all p ≤ 0.001).

Do left-handed fighters show greater fighting ability than right-handed fighters? We compared
the fighting success of left-and right-handed fighters with one-tailed Mann-Whitney U tests. Each of the 3 samples was compared separately by win percentages, and the samples of male and female boxers were also compared by BoxRec scores. We calculated the measure of stochastic superiority 36,37 as an effect size for each comparison. The measure of stochastic superiority is the probability that a randomly selected left-handed fighter would have a higher win percentage/BoxRec score than a randomly selected right-handed fighter. The measures of stochastic superiority for each sample, along with 90% bootstrapped confidence intervals around them are given in Fig. 1.
Among male boxers, the probability that a randomly selected left-handed fighter would have a higher BoxRec score than a randomly selected right-handed fighter was 52.4%, which a Mann-Whitney test showed was significant (p = 0.00069). The measure of stochastic superiority for win percentage was also 52.4%, which was also significant (p = 0.0007). Thus left-handed male boxers have significantly higher BoxRec scores and win percentages than right-handed male boxers.
Among female boxers, the probability that randomly selected left-hander showed a higher BoxRec score was 53.9%, which a Mann-Whitney test showed was marginally significant (p = 0.053). The measure of stochastic superiority for win percentage was 54.5%, which was significant (p = 0.031). Thus left-handed female boxers showed significantly higher BoxRec scores but not win percentages.
Among MMA fighters, the probability that a randomly sampled left-handed fighter showed a higher win percentage than a randomly selected right-handed fighter was 53.5%, which was significant (p = 0.016). A Spearman's correlation between total number of fights and win percentage revealed a negative correlation (r = −0.14, p < 0.0001), indicating that this effect is not due to left-handed fighters having more fights. Thus left-handed MMA fighters showed significantly higher win percentages than right-handed MMA fighters.

Do left-handed fighters show greater variance than right-handed fighters?
We compared the variance in BoxRec scores and win percentages among left-and right-handers by bootstrapping differences in variance (10,000 samples), with bias correction and acceleration following 38 to obtain robust p-values. All p-values are one-tailed. Left-handed male boxers showed higher variance in BoxRec scores (p = 0.002) but not in win percentages (p = 0.74). Left-handed female fighters did not differ from right-handed female fighters in the  www.nature.com/scientificreports www.nature.com/scientificreports/ variance of their BoxRec scores (p = 0.80) or win percentages (p = 0.77). Likewise left-handed MMA fighters did not differ from right-handed MMA fighters in the variance of their win percentages (p = 0.54).

Does the left-hand advantage show negative frequency-dependence?
The prevalence of left-handedness in female boxers was much lower than in male boxers (17% vs 12.5%), while the magnitude of left-hand advantage in the BoxRec scores of female fighters was higher (54.5% vs 52.5%). If the advantage left-handed fighters have is negative frequency-dependent, then we might expect left-handed female boxers to have a relatively larger advantage than left-handed male boxers. To investigate this, we compared the measures of stochastic superiority in the BoxRec scores of male and female boxers, and we bootstrapped a confidence interval around the difference (10,000 samples). The difference in the advantage of left-handed female and male boxers was not significantly different from 0 (bias corrected, accelerated p-value = 0.29). Thus, we have no evidence that female boxers experience a greater left-hand advantage than male boxers.

Discussion
Across three samples, we found that left-handed boxers and MMA fighters are both overrepresented in their respective sports and are more successful fighters. In male boxers, these effects held for both win percentages and BoxRec scores. In female boxers we found that left-handed fighters showed higher BoxRec scores but not higher win percentages. Our results are consistent with the fighting hypothesis that left-handedness is maintained in populations because it provides an advantage in contest competition.
Our finding that left-handed fighters have better records than right-handed fighters in both male boxers and MMA fighters contrasts to most previous studies 18,20,21 , (but see 19 ). Two factors may have played a role. Firstly, the effect is small and may only be detectable in large samples such as ours. Second, it may not be detectable in datasets with low variance in fighting ability, such as when studies use samples of only elite fighters 18 . The fact that we find similar results in both win percentages and BoxRec scores, which are a more complete measure of fighting ability, lead us to believe our results are robust.
Our positive finding for MMA fighters may be surprising, as a similar study 21 did not find a significant advantage of left-handedness in a sample approximately 70% of the size of ours. The study collected data from the same website we did ~6 years earlier, so its data set likely overlaps somewhat with ours. The different results may be due to our larger sample size and to the choice of analyses in the two studies. Baker & Schorer 21 compared the winning percentages of left-handed and right-handed fighters using t-tests, which are underpowered relative to the Wilcoxon tests when error is non-normally distributed 39 , as is the case in our data. Moreover, t-tests assess the difference in the means of two groups, while Wilcoxon tests assess stochastic superiority. If fitness depends on the outcome of pairwise interactions, as is likely the case in male-male combat, then the Wilcoxon test will be the more appropriate test of the hypothesis. It is noteworthy that in Baker and Schorer, left-handed fighters had a non-significantly higher win percentage, so the trend reported is consistent with our results.
We found that left-handed female boxers had better BoxRec scores than right-handed female boxers. Left-handed fighters were less frequent in the female sample than in the male sample (12.6% to 17.3%), so we tested whether the left-handed advantage in female fighters was higher than in male fighters. Lower frequency paired with greater fighting success would be consistent with a negative frequency-dependent advantage, and thus with the fighting hypothesis. We did not find this pattern. Cross sex comparisons should be interpreted cautiously however, as there are established sex differences in sport-relevant cognitive and perceptual functions 40 . That the left-handed advantage in combat is negative frequency-dependent remains to be convincingly demonstrated, and is a crucial topic of future research. This might be investigated by comparing fighting leagues with varying levels of left-handers, or by testing whether increased contact with left-handed opponents over a fighter's career increases his/her probability of winning.
Unlike Dochtermann et al. 22 , overall we found little evidence that left-handed fighters showed higher variance in fighting ability. Across all samples, only male left-handed boxers showed significantly higher variance, and then only in BoxRec scores. The difference in results could be attributed to the fact that Dochtermann et al. tested variance in the probability of a fighter to win a single given fight, whereas we examined variance in fighting success as measured by a fighter's record over their career thus far. It is possible that coaches (many of whom may suspect the existence of a left-handed advantage) or the left-handed fighters themselves adapt their training to compensate for their fighter's lower skills. However we warn that cross sectional data, such as ours and that of Dochtermann et al., are limited in their ability to answer this question. Longitudinal work that tracks whether left-handed amateurs are more likely to reach professional level regardless of initial skill would be valuable, and shed more light on this interesting hypothesis.
The current study examined handedness as a binary trait, but people vary in the strength of preference for each hand, and this may also influence fight success. One recent population study found that participants who showed weak hand preference (ambidextrous) were higher in aggression and self-reported fight success than both strong left and right handers 5 . It is unclear how many of the fighters in our dataset are in fact ambidextrous, and we cannot distinguish between those who favour a given stance strongly or weakly. The fighting ability of ambidextrous professional fighters, who could presumably switch stance during a fight, could potentially be quite high, but is currently unknown. This would be a valuable topic of further study. A related issue is that a fighter's stance may not always map onto their overall laterality 41 for example Vasyl Lomachenko, regarded as one of the best active boxers at the time of writing, fights in a left-handed stance despite claiming to use his right hand for the majority of tasks. It must be noted however that the fighting hypothesis does not require that all fighters use a stance consistent with their preferred hand, only that they be correlated.