Does Losing Lead to Winning? An Empirical Analysis for Four Sports
In our paper “Does Losing Lead to Winning? An Empirical Analysis for Four Sports”, we attempt to replicate the finding that being slightly behind halfway through a competition increases the chance of winning.
In an influential paper, Berger and Pope (BP hereafter) argue that lagging behind halfway through a competition may not necessarily imply a lower likelihood of winning, and that being slightly behind can actually increase the chance of coming out on top.
In particular, they argue that if winning is the goal, then the performance of the opponent will serve as a salient reference point. If people are loss averse, then those who are slightly behind will be more motivated to extend effort than those who are slightly ahead.
To test this hypothesis, BP analyze more than sixty thousand basketball matches. Their main analyses rely on NBA data, where they find that teams that are slightly behind are between 5.8 and 8.0 percentage points more likely to win the match than those that are slightly ahead.
BP is regarded as one of the first studies to show that loss aversion is not limited to inexperienced subjects in low-stakes settings, but also affects the behavior of experts in a high-stakes professional environment.
In our paper, we extend the analysis of BP to large samples of Australian football, American football, and rugby matches, and then revisit the analysis of basketball.
We use a state-of-the-art regression discontinuity design to estimate the effect of being slightly behind at half-time on the likelihood of winning the match.
For Australian football, we find no evidence of a performance-enhancing effect of trailing. In both leagues that we study, the effect is not significantly different from zero. For the AFL, the point estimate is even in the opposite direction than hypothesized.
For American football, the estimated effect sizes for the two leagues that we consider are sizeable but in opposite directions, and both are not statistically significantly different from zero.
For Rugby, we also find no convincing evidence that trailing at half-time discontinuously affects the chance of winning. The estimated effect ranges from a 2.9 percentage point decrease to a 6.5 percentage point increase. All are statistically insignificant.
We thus found no statistical support for the hypothesis that being behind improves the odds of winning in Australian football, American football, or rugby. We therefore turned our attention to basketball, the sport that was central in the original paper.
When we consider the period that BP considered for the NBA, our findings are very similar: trailing statistically significantly improves the odds of winning by about 8 percentage points (left: BPs results; right: ours).
However, we find no compelling evidence of such an effect for NBA matches outside this sample period, NCAA matches, or WNBA matches.
So overall, we only find the result for the original NBA period that BP considered, but nowhere else. However, as can be deduced from the wide confidence intervals: some of our tests have little power. Therefore, we turn to meta-analyses to synthesize our results.
We first conduct meta-analyses per sport. For basketball, we perform the analyses both with and without the original samples analyzed in BP. Again, all estimated effects of trailing at half-time are statistically insignificant.
Finally, we conduct meta-analyses for all sports combined. The estimated treatment effect is a statistically insignificant 1.2 percentage point if we include the original BP samples. If we exclude their samples, it is economically and statistically indistinguishable from zero.
So, in the end, the answer to the question we pose in this paper seems clear: losing does not lead to winning in these four sports.
Of course, our null results do not mean that trailing in a competition does not or cannot have a systematic positive, motivating effect. However, it does show that in these situations, where feedback (and players’ responses to this feedback) are continuous and where the players are highly experienced, trailing at a particular point in time does not seem to have a discontinuous effect on winning.
At the end of this post, I want to give credit and express gratitude to Jonah Berger and Devin Pope. This project started because we liked their original paper and thought it would be cool to show the effect in other sports. When we did not find the effect out-of-sample, we were a bit nervous about how they would respond. We are happy to say that they could not have been more supportive. Berger and Pope shared data and provided constructive criticism on earlier drafts of our paper. While the review process is anonymous, we are pretty confident that one of them acted as a positive reviewer. Their behavior provides a positive example for the field.
The paper is joint work with Bouke Klein Teeselink and Martijn van den Assem, and is available on SSRN .