Hatfield, MA — This season, the College Squash Association has been operating under a new ranking methodology.
For years, the CSA utilized a manual ranking process that included a number of tie-breaking scenarios. This method worked fine for years; however, over the last few seasons, the parity between teams has increased. This parity led to challenging ranking situations, which in turn led many coaches to desire a new method to rank teams.
The CSA studied many ranking methodologies during the offseason. In early June, college coaches gathered at Yale for the CSA’s annual strategic review. During this two-day meeting, multiple ranking systems were presented to and discussed by the coaches.
In early September, the men’s and women’s Executive Committees voted to enact a change to the ranking methodology. For team rankings, each season has three phases:
- The preseason team rankings are determined by a poll of varsity coaches;
- After the preseason rankings are published, the men’s and women’s ranking chairs and the Executive Administrator rank the teams;
- Beginning with the ranking cycle that includes January 25th, a computerized ranking system will be enabled.
The third phase of the season’s ranking cycle is now upon us. The computerized ranking system the CSA will utilize is self-consistent Elo ranking, a modification of standard Elo to remove any dependence on match order. All matches are taken into account up to the ranking date, treating a match played on November 1st as having the same importance as a match played on the final day of the regular season.
Elo methodology is used by numerous organizations to calculate the relative skill levels of teams in competitor-versus-competitor situations. The method determines ranking points for each team based on match outcome. The distribution of ranking points among teams is such that the difference in points between any two teams at a given time is a predictor for the outcome of a future match between those two teams. The larger the difference in ranking points between the two teams, the larger is the expectation that the team with the greater number of points would beat the team with the smaller number of points.
Every team starts the season with the same number of points. In the CSA’s case, this number is 1,000.
When two teams with equal points play each other, the system assumes that each team has a 50% probability of winning. The winning team has exceeded expectations, and the losing team has failed to meet expectations. Elo exchanges points between the two teams, the winning team increasing its points and the losing team decreasing its points by the same amount. The adjusted points reflect an expectation of the result of a potential rematch between the teams.
When a higher-ranked team defeats a lower-ranked team, and the ranking points differential between the two teams is large, the ranking points do not change greatly for either team since the outcome was expected. However, if the lower-ranked team defeats the higher-ranked team there is a substantial exchange of ranking points between the two teams, the lower ranked team gaining points roughly in proportion to the point differential between the two teams prior to the match.
The Elo ranking methodology was tested with data from the past few collegiate seasons. The rankings were very accurate, with only a few minor variations from the historical rankings.
In the weekly rankings, the point totals for each team will be listed. Since the Elo ranking methodology is based on points, it is possible to have ties. The CSA implemented tie-breaking rules if this scenario occurs:
- If the tied teams faced each other during the current season, the winner of the last meeting is placed ahead of the losing team;
- If the head-to-head tiebreaker does not determine a winner, the team with the victory over the highest ranked opponent not involved in the tie is deemed the winner of the tie.
Until the ranking cycle immediately prior to the first Team Championship weekend, only varsity schools will be ranked. Prior to the first Team Championship weekend, non-varsity teams will be included in the computerized ranking system to determine seedings for the tournament.
Per the CSA’s rules, if needed, the Ranking Chair may adjust a non-varsity team’s ELO ranking position. This may be warranted if a club team did not play varsity squads and is undefeated at the end of the regular season. In this example, the Elo ranking system may inadvertently boost the club team’s ranking based on points. In running past seasons through the Elo ranking system, there were a few non-varsity teams that required adjustment due to their limited schedules.
For individual rankings, the Elo method will also be utilized. These rankings will seed the individual tournament. The CSA will utilize the final top-9 ladder positions played by each team at Team Championships to assist in the ranking process. For example, the #1 player on each team is assumed to be stronger at that moment in time than the #2 player, while the #2 player is assumed to be stronger at the moment in time than the #3 player. Using intra-team positioning and play against opposing schools will create a much more accurate individual ranking system.
The CSA is very excited to use the Elo ranking methodology.