Practice?! We talkin’ bout practice?!
An analysis of how regular season performance impacts national meet finishes.
Okay, I’ll admit it. The famous Allen Iverson quote doesn’t perfectly match here but it’s close enough. Cross country is already a somewhat ridiculous sport but trying to rank runners and teams only adds to that. There’s an entire debate about whether times even matter and comparing performances across courses and years is nuanced at best. Regardless, we try to do it every year with the idea we can predict what will happen on the national stage. That begs the question — do regular season performances even matter?
To try and answer that, data from 1580 Division 1 cross country runners was analyzed from 2012 to 2021. The goal was simple — determine if the regular season mattered in the postseason. And the answer is… yes, kind of.
Look, it doesn’t take a data scientist to figure out people who win a lot of races probably have a better chance at winning nationals. But how true is that trend when you start getting deeper into the field? Do people who finish 15th on average still do that well at NCAAs? And how does speed factor into all of this?
Let’s start with place because we know times have flaws. For simplicity, athletes who DNF’d (Did Not Finish) or DNS’d (Did Not Start) a race had those results excluded from the race. Those results can often be skewed by a variety of factors (injury is common) and so they were not included here. Normally, there a number of top runners who will DNF at NCAAs for one a number of reasons but you will not see them in any of the charts.
It’s abundantly clear that season performance and NCAA finish are tied together. Using the chart above, it’s easy to see the relationship between the two. The farther out you get, the more noisy the data gets, but even for runners finishing in the top 100, they generally need to average in the top 50 across their regular season. What’s noteworthy is that for both All-Americans (top 40) and fringe (41st to 60th), a handful of athletes have snuck into those groups despite underwhelming performances during the regular season.
To get further into the statistics, correlation tests for NCAA finish vs both average time and average place indicated significance — i.e they both matter. These tests were done separately on men and women due to the time differences between the two. In both cases, the correlation between average place and NCAA finish was > 0.8 (max is 1.0), indicating a very strong relationship. Average times were 0.5 for men and 0.79 for women. Men likely saw a lower mark as they run 10k at nationals but 8k during the season. Women compete over 6k for the full season.
What about time, the ever debated topic of cross country? As a quick precursor, folks love to debate whether cross country times actually matter. Courses change a bit year-to-year based on markings and such, weather is always different, and some courses use a loose definition for ‘8k’ and ‘6k’.
Based on the data, times matter but only to some degree. In fact, times are a lot like getting into a club. You need to have some baseline of social status to get in but once you’re in, you’re in.
For women, there’s a clear trend of time ranges that tend to land you in the All-American vs fringe vs top 100 categories. There’s always a bit of crossover but in general, you need to be this tall to ride the ride.
Men are less clear cut. This again comes back to the 8k vs 10k debacle, where some runners are less suited for that additional 2k. Generally, faster is better but when you look across the groups, the ranges aren’t nearly as defined as the women. That’s not to say time doens’t matter at all, but it’s certainly less of a factor and a worse predictor for the men.
The major takeaway here is runners should focus on winning and let the times come as they may. Sure, time can be a predictor, but your finishes during the season are more indicative of where you’ll land at NCAAs. To quote Al Davis, “Just win baby.”
All graphics and analysis were done using R and the ggplot package. Visuals use the theme_fivethirtyeight() option from the ggthemes package.
