Fading red? No evidence that color of trunks influences outcomes in the ultimate fighting championship (UFC)

In fish, reptiles, birds, and primates it has been documented that the color red plays an important role in dominance and threat displays, with the more intimidating individual being associated with winning contests (Pryke, 2009 for review). A much debated paper by Hill and Barton (2005) sparked off research concerning the influence of red attire on combat sports outcomes in humans, arguing that red could also play a role in these dominance contests. Hill and Barton (2005) presented evidence from four combat sports during the 2004 Olympics showing significantly higher win-ratios for red as opposed to blue outfits, suggesting that red clothing could heighten the likelihood of winning. The findings were explained as psychological effects on either one or both of the contestants originating from evolutionary and cultural associations between red and dominance as well as between red and aggression. A study by Feltman and Elliot (2011), using a scenario methodology, found preliminary evidence for both wearing and viewing effects of red equipment on perceptions of dominance and threat. However, a replication attempt of Hill and Barton (2005) relying on the 2008 Olympics did not support a performance enhancing effect of red (Seife, n.d.). Moreover, the statistical assumptions for tests in Hill and Barton (2005) could have been violated (non-independence) and selection bias in allocation of red and blue is possible (Seife, n.d.). In addition, Rowe et al. (2005) proposed an alternative explanation for Hill and Barton’s (2005) findings. After finding a similar winning bias for blue over white in judo competitions Rowe et al. (2005) argued that visibility differences between contestants may be responsible for observed effects since red was not represented in these judo fights. Dijkstra and Preenen (2008) indicated that the Rowe et al. (2005) study was confounded and found no winning blue effect after controlling for potential biases. The “winning red” effect in combat sports was supported, however, by Hagemann et al. (2008), who argued that red can lead to a perceptual bias in referees. As predicted, referees were found to award significantly more points to taekwondo competitors in red rather than blue apparel. The present study aims to conceptually replicate the effects found by Hill and Barton (2005) by testing for a “winning red” effect based on the color of shorts (trunks) in the ultimate fighting championship (UFC) (www.ufc.com), a televised mixed martial arts competition. The UFC has very limited rules for combat and hence provides an opportunity to test whether the “winning red” effect generalizes to a less regulated combat sport. Finally, the effect of color on referees’ decisions and clear fight outcomes will be examined. For episodes 118–148 (210 matches) we coded the base color of the shorts worn by the fighters (for description of sample: Pollet et al., in press). Sometimes shorts have multiple colors, a primary and a secondary, but we coded the “base” color, typically also displayed at the bottom of the screen. The predominant colors were Black, White, Red, Blue, and “Other,” which included rarely chosen colors: Brown (n = 4), Camouflage (n = 3), Gold (n = 1), Gray (n = 5), Green (n = 6), Pink (n = 2), Purple (n = 3), and Yellow (n = 2). The UFC does not have colored gloves as in boxing but the black gloves do have either a small blue or red band. These were not coded, as the bands on the predominantly black gloves are substantially smaller than the trunks. Three draws were excluded from our analyses. Analyses were conducted in SPSS 20.0 and were two-tailed. Sensitivity power analysis via G*Power 3.1 (Faul et al., 2007) indicates sufficient power to detect a relatively weak effect (g = 0.18) in our sample (Binomial test, two-tailed, α = 0.05; Power = 0.8). A simple binomial test did not support a winning red effect (27 wins out of 61; Binomial test p = 0.443; Bayes factor = 4.245 in favor of 0.5 win rate; Rouder, n.d.). In a Generalized linear model (GZLM) with binomial link, color (5 categories) did not account for fight outcome (Win/Loss; Wald χ2 = 4.153, p = 0.386; Figure 1). Incidentally, we also tested a multilevel logit model taking into account a nested structure based on “fighter,” as the same fighter can appear multiple times and could have a preference for shorts. In this model, color also did not account for fight outcome, p = 0.387. Similarly, conducting a GZLM separately for jury decisions and “clear” outcomes (Technical Knockout, Knockout, submission or tapout) did not affect the effect of color on