The NCAA championships start up tomorrow. We’ve had the psych sheet for a while now, but if history tells us one thing, it’s that psych sheet rankings are an imperfect predictor of nationals results. Swimmer’s final times vary dramatically from their seed times. To measure how much, I compared swimmer’s seed times to their best time in an event at nationals for each of the last 6 years. Men’s D1 swimmers added an average of .2% to their seed times with a standard deviation of 1.3% on 3509 data points. The data were pretty normally distributed:
Time changes distribution
negative is faster, positive is slower
There’s more going on with this data than pure random noise. There are systemic reasons some swimmers have big meets and other swimmers add a bunch of time.
One major factor is how much rest swimmers have gotten in their events at meets earlier in the season. For example, Caeleb Dressel is 49th on the psych sheet in the 100 fly, an event he is favored to win. There is a noticeable negative correlation between fast seed times and performance vs seed. Before we dig into that, what does it mean to have a “fast seed time”? A “fast seed time” for Caeleb Dressel and someone else are different things. Dressel’s 50 free seed time is slower than last year, but would be very fast for anyone else. For him 18.66 is a slow seed time. To measure how fast seed times were I compared a swimmer’s seed time to their best time in a previous season at their conference meet or nationals. By comparing swimmers to themselves in the past we are capturing if a seed time is fast for a particular swimmer. Also, limiting the previous season best times to conference or nationals best times allows us to know with near certainty that their previous time was swum tapered, so it’s a fair metric for their actual ability in an event. This method leaves out freshmen and swimmers doing new events this year, but I was still able to get 433 data points from last year’s meet. That’s a pretty good sample size.
With the definitions out of the way, the results: For every 1% a swimmer’s seed time is better than their best time the previous season, they added .36% to their seed time at nationals. That is to say, swimmers with fast seed times got slower vs their seed times and swimmers with slow seed times got faster on average.
This isn’t to say that it’s better to have as slow of seed times as possible headed into the meet. In general, it’s better to have seed times that are as fast as possible headed into the meet. While fast seed times are associated with going slower than those seed times, the expected slow down is smaller than the associated overall improvement. Put another way, a 1% improvement in a swimmer’s seed time vs their previous season’s best time is associated with a .50% improvement in their nationals swim vs their previous season’s best time. Swimmers with slow seed times beat their seed times on average, but they don’t beat their best time from a previous season on average. Here’s a plot of the 2017 men’s data. That giant outlier in the top right corner is Joseph Schooling’s 200 fly.
It appears there are two competing effects here. First, some swimmers overall skill level has improved/regressed. Their seed times are faster because they are better this year. Or their seed times are worse because they are not as good. This is reflected in their seed times and their performance at nationals. Second swimmers had different amounts of rest when they achieved their seed times. Or they hit/missed their first tapers to varying degrees. The swimmers who already had a full rest and/or nailed their tapers drop off at nationals, and the previously less rested swimmers drop time. Both of these effects affect all swimmers to some degree and it’s difficult to tell how much of a change in performance is due to one, the other, or both.
So with all that said, here is a list of who has the slowest seed times that we would expect to improve the most this weekend. Caeleb Dressel’s 100 fly leads the way at 7.6% slower than his time last year. In case any one believed he was fully rested when he put up those crazy times at conference, the one nationals event he did swim was far enough off his best time that it still made the top 20 list of “slowest” seed times.
|Name||Event||School||Collegiate Conf/Nats Best Time||Seed Time||Difference|
|1||Dressel, Caeleb||100 Butterfly||Florida||43.58||46.89||7.60%|
|2||Shebat, John||200 Backstroke||Texas||1:37.24||1:41.52||4.40%|
|3||Reid, Christopher||100 Backstroke||Alabama||46.17||47.64||3.18%|
|4||Shebat, John||100 Backstroke||Texas||44.35||45.76||3.18%|
|5||DeVine, Abrahm||400 IM||Stanford||3:37.73||3:44.41||3.07%|
|6||Schooling, Joseph||50 Freestyle||Texas||18.76||19.33||3.04%|
|7||Schooling, Joseph||200 Butterfly||Texas||1:37.97||1:40.72||2.81%|
|8||Perry, Sam||50 Freestyle||Stanford||19.13||19.66||2.77%|
|9||Barone, Jack||200 Breastroke||Ohio St||1:54.61||1:57.62||2.63%|
|10||Coetzee, Ryan||100 Freestyle||Tennessee||43.09||44.19||2.55%|
|11||Roberts, Jonathan||400 IM||Texas||3:38.18||3:43.72||2.54%|
|12||Dressel, Caeleb||50 Freestyle||Florida||18.2||18.66||2.53%|
|13||Wich-Glasen, Nils||200 IM||South Carolina||1:45.03||1:47.65||2.49%|
|14||Carter, Dylan||100 Freestyle||Southern Cali||41.73||42.74||2.42%|
|15||Carter, Dylan||200 Freestyle||Southern Cali||1:30.95||1:33.15||2.42%|
|16||Vissering, Carsten||200 Breastroke||Southern Cali||1:54.09||1:56.81||2.38%|
|17||Schooling, Joseph||100 Butterfly||Texas||43.75||44.78||2.35%|
|18||Kaliszak, Luke||50 Freestyle||Alabama||19.53||19.97||2.25%|
|19||Haas, Townley||500 Freestyle||Texas||4:08.92||4:14.43||2.21%|
|20||Dressel, Caeleb||100 Freestyle||Florida||40||40.87||2.17%|