Squash Analytics: Men’s College Rankings
by David Keating

November 15, 2019

Today’s match between UVA and GWU crucial for both teams. 

Should you put any stock in the College Squash Association (CSA) ranking of teams at the beginning of the season?

That ranking is done by polling all the varsity coaches and certainly the coaches are aware of who got what new recruits. All of them know how to evaluate talent.

But with the incredible depth now seen among the top teams, a simple look at the rankings might lead you to believe there are more differences between the teams than is actually the case. This article will present another way to rank the teams. And it will offer a more informed look at the magnitude of the differences between the top teams.

According to the CSA, “total votes determin[e] the order of the preseason rankings.” The CSA reported that “Similar to last season, the men’s poll exhibited more movement [compared to the  2018-2019 final rankings after the national championships] than the women’s rankings. The headliner is Princeton’s jump into the top 4 after finishing 8th in last year’s tournament, all thanks to a strong recruiting class and top returning players. George Washington (No. 8) and Virginia (No. 9) will be right on the bubble all year long, hoping to stave off Dartmouth and Drexel, who both dropped relative to last season.” University of Virginia also jumped up three-spots.

There’s another way to rank the top teams. I’m not claiming it’s better, it’s just another method. One can compare the teams using the skill level ratings computed by US Squash for each player on each of the teams. Using this data, I built a computer model that simulated each of the teams playing the other right now. When I tested the model during the national team championships earlier this year, it did a better job than the seeding in predicting the final outcome.

Still, this method is hardly foolproof. There are several flaws I’ve identified and there are probably more. The most obvious is that the rosters listed online might not be accurate for each team. I didn’t ask the coaches about that. If the player is listed on the CSA roster, I assumed he will play for the team this season.

First, the skill ratings for many players are often not based on many meaningful college matches. Often matches are lopsided (e.g., Harvard playing Tufts) and that doesn’t help the rating algorithm as much as closely contested matches. And for international players who are freshmen, the skill ratings are often based on even less data. However, this flaw will become less important as the season progresses. This year many, if not most, teams are posting their challenge ladder matches online in the US Squash software, a commendable development that will lead to much more accurate skill ratings and help discourage stacking.

Second, I’m skeptical that junior skill ratings are directly comparable to college player skill ratings. Many juniors only play juniors, and few play adults on a regular basis. Fewer still play any college age players in matches that post to the US Squash system – there simply are very few opportunities for that to happen. I’m guessing that junior ratings are overstated compared to college ratings. That could well introduce a large source of error among American freshmen.

Third, no one knows how much differences in skill predict the outcome of the match. For example, if the difference in skill rating is 0.1, does that mean the higher rated player has a 60% chance of winning, or some other chance? My model makes informed assumptions for that, but no one has yet analyzed the data.

Finally, the model is fairly rudimentary in predicting a team’s odds of winning, which is actually a tough statistical analysis that I don’t have the knowledge yet to make. But my rough estimates have proven highly predictive in the past.

With all these caveats out of the way, I compared the CSA ranking of the top 10 teams with my model’s ranking.

Here are the top 10 teams according to the CSA:

#1        Harvard University

#2        Trinity College

#3        University of Pennsylvania

#4        Princeton University

#5        Yale University

#6        University of Rochester

#7        Columbia University

#8        George Washington University

#9        University of Virginia

#10      Dartmouth College

My computer model ranks these same teams as follows:

#1        Harvard University

#2        University of Pennsylvania

#3        Trinity College

#4        Princeton University

#5        Yale University

#6        University of Rochester

#7        University of Virginia

#8        George Washington University

#9        Columbia University

#10      Dartmouth College

As you can see the two methods produce remarkably similar results. Under the computer model, Penn moves up to #2 and Trinity moves down to #3. UVA trades places with Columbia to jump from #9 to #7.

Under either ranking system, the UVA match at GWU today is shaping up to be a crucial match for both teams. A loss by one won’t ruin their chances at the top division, but it would make it somewhat more difficult.

One thing is clear from my computer model, and probably to all the coaches of the top 10 teams – Harvard is clearly going to be the team to beat. Six of its nine starters sport skill ratings above 6.0. No other team has more than three. The #9 Harvard player would be at ladder position #5, or better, for any of the other top teams. My model predicts Harvard would win each match against the other top 10 teams by at least a 6-3 margin and nearly all the teams would be lucky to get two individual wins. An upset is possible, most likely if Harvard is playing an away match and/or down a player, though the team has incredible depth beyond the #9. My model predicts 65 individual match wins of a possible 81 against the other top 10 teams. No other team would win more than 49. By the end of the season, it may well become the best collegiate team ever.

Penn, Trinity, and Princeton are all closely matched. Each of those dual matches would be near toss ups, with the home team likely holding a tiny edge. (An earlier analysis I did for Daily Squash Report appears to indicate home court is worth on average about 0.66 extra individual match wins.) Yale appears to be a slight step behind those three, but certainly capable of pulling an upset, especially at home.

Yale in turn appears to be a slight step above Rochester, UVA, GWU and Columbia. The #6 to #9 teams appear to be closely matched, with Rochester perhaps just a bit stronger and Columbia just a bit weaker than UVA and GWU. All four of those teams behind Yale have a decent chance of pulling an upset against the Bulldogs, especially if Yale plays an away match.

Other factors beyond team control such as injuries, illness or home court advantage might well determine dual match results or the end-of-season rankings, which are built on head-to-head match results over the season.

But there are other things in the team’s control. How hard do the players work? Do they eat and sleep well? How good is the coaching staff? How good is the between game coaching and are the strategies for each opponent?

Lurking just below the top 10 teams are some highly capable teams that could easily break into the top 10, but I didn’t have time to analyze the other teams in the rankings.

The season is just beginning, but it promises to showcase the best college squash in history. There will be many exciting matches this season to watch. The early season rankings are just that, early. It will be interesting to see how the season progresses and how it ends.

Coming soon, a look at the top 10 women’s college teams.

The views expressed by the author are his alone, based on his experiences and observations. They do not represent the views or positions of any organizations with which he is affiliated. All information used for the article is available to the general public.