By Kurt Bullard
With the leaked bracket having been officially released by the Selection Committee, millions of Americans across the country will begin to fill out their own brackets, trying to prove to friends and loved ones that they know more about college basketball. While there were some surprises announced in the field—most notably Tulsa—many familiar coaches have returned to the tournament. Jim Boeheim, Coach K, Roy Williams, Tom Izzo and Bill Self will all coach come Thursday. Combined, the five have coached in 122 tournaments over the past four decades. But has their performance in March Madness improved with experience, or do previous NCAA Tournament appearances for coaches have a negligible effect?
To look at this question, I examined the performances of coaches over the course of their careers from 2002-2014, stemming from a dataset that I used in posts last year. I then measured how these coaches performed each year relative to their seed. In order to set this benchmark, I calculated the average amount of wins for each seed over these years to set an expectation for the team. Having this, one could measure over- and underperformance by seeing how many wins more or less the team notched compared to how many one would expect them to get.
I used only data for the top 5 seeds, because I wanted the threshold for outperforming to be above getting just one win, which is only true for the top 5 seeds. The expectation for the top 5 seeds is as follows:
Then, I measured the relative performance of each team for these 13 years by subtracting the benchmark from each team’s tournament wins that given year. The plot below shows the wins above expectation measured against the coaches’ experience in terms of March Madness appearances.
There doesn’t appear to be a strong association between experience and relative performance in the tournament. If a coach’s performance in the Tournament improved over time, one would expect a positive correlation between the two variables. But just to confirm, I ran a linear regression of out-performance and years of experience in the tournament.
As is shown here, there isn’t a significant relationship between relative performance and experience in March, meaning that years of experience doesn’t give these long-tenured coaches an additional bump in the Big Dance.
This is not to say that experience in coaching itself doesn’t matter in general. Rather, it is saying that there might not be an additional bump in March from extra experience playing in the NCAA Tournament, and that the coaches’ skill displayed throughout the regular season is already factored into the teams’ seedings. In other words, coaches don’t suddenly get better come Mid-March—they exhibit their influence during the year, and that is factored into seeding.
For example, Bill Self’s performance over the years has not seemed to exhibit any increasing ability to outperform his seed. In fact, the Kansas coach has under-performed in eight of his last ten NCAA Tournaments.
So as tempting as it may be to think that Coach K or Jim Boeheim can help draw up the X’s and O’s necessary to push their teams ahead in March, don’t think that they rise to the occasion more so in the NCAA Tournament than they do during the regular season.
Not sure if my original comment came through.
Great analysis, Kurt. Appreciate the effort. I found the chart about Bill Self particularly relevant and enlightening. Confirmed what my instincts told me. Earlier today I circulated an email to the 150+ participants in the pool I manage, noting the historical under performance of Bill Self and Kansas relative to expectations. Called that section “Self-doubts”. Your column is quite timely. Will likely include a link to it with my next update.
Thanks and I’d love to see a similar review of Jim Boeheim’s performance. I suspect it’s much of the same.
NCAA Tournament: Harvard Minds Pick the Upsets (Harvard …