By John Ezekowitz and Andrew Cohen
Last week, we analyzed the effect of tempo on NCAA Tournament upsets and found that contrary to Dean Oliver’s theory, a faster tempo (more possessions) was predictive of underdog success. We searched for potential explanations for this phenomenon and, after taking suggestions, researched into the two most popular explanations.
Many suggested that the increased tempo observed in upsets could be caused by the fact that upsets are often close games, leading to late game fouls that artificially increase the possessions. This theory certainly seemed plausible and worthy of study. Our results come after the jump.
Unfortunately, the data that would best analyze this theory, possessions after the first 35 minutes of the game rather than the full forty, was unavailable. As a proxy, we looked at margin of result (this is the absolute value).
The intuition that upsets were generally closer games than non-upsets was proved correct. After eliminating one outlier, Long Beach St’s 35-point drubbing at the hands of Tennessee, the average margin of result for upsets was 7.89, while for non-upsets it was 11.86. This difference was strongly statistically significant, with a P value<0.005.
The second part of the intuition, however, was not backed up by the data (brief stat geek interlude: because the margin of results distribution was right skewed, we used a square root transformation). There was no significant correlation between closer games (smaller margin of result) and a higher tempo (more possessions). The r2 was only 0.043. We caution that our sample size in this data set is relatively small (only 144), and that the effect may hold true for a wider swathe of games. Nonetheless, this seems to reject the theory that the fouling that is present at the end of closer games (ie upsets) led to an increased tempo.
Another potential explanation for our previous results was that despite successful underdogs playing faster than their unsuccessful compatriots, relative to their competition they were in fact playing slower. This would mean that upset victims’ s average tempos during the year. However, this theory too was not supported by the data. There was no significant difference between the average tempos of upset victims and the pace they played in their losing NCAA Tournament efforts.
Having examined the data we have and the theories so far put forward, we have little choice but to conclude that successful NCAA Tournament underdogs simply play at a faster tempo during their upsets. This, of course, does not preclude another explanation for the phenomenon, and if you have further thoughts, please chime in in the comments.
Please stay tuned to HSAC as we attempt to create a model that predicts successful underdogs in the NCAA Tournament using tempo-free stats. Perhaps we can help you successfully pick that 12 vs 5 upset this March.