BY THE NUMBERS: Women’s hockey projections

By Yale Undergraduate Sports Analytics Group | Posted on, 3:24 p.m.

The Yale Undergraduate Sports Analytics Group and media coordinator Michael Bogaty ’19 are excited to launch a new series featuring season projections for different Yale sports. This is the second post in the series, covering women’s hockey.

About the model:      

YUSAG’s prediction model utilizes the ELO ranking system, most notable for its use in chess. ELO has been applied to a variety of sports by FiveThirtyEight and other analytics platforms. When a team wins a game, its rating increases and the rating of its opponent decreases, taking into account the initial ratings of the teams. If a very good team defeats a very poor team, the result will not likely do much to change the system’s opinion of the relative strength of the teams and thus the ratings change will be small. In an upset, however, the underdog’s rating will go up by a much larger amount.

Our model also takes into account two factors that the traditional ELO system does not. First, it takes into account the home-field advantage in each sport. Second, it takes into account the score differential, or magnitude of victory. Research has consistently shown, in just about every sport, that considering a team’s point differential is more predictive than just considering wins and losses.

As of this writing, the Yale women’s hockey team is right on the playoff bubble, sitting eighth in the ECAC standings. In part due to a goal differential better than their record would suggest, our model (which ranks the Bulldogs seventh best in the conference) is relatively optimistic about the Elis’ chances of reaching the ECAC tournament, putting their odds at over 86 percent. Below are the chances of each ECAC team earning the top seed, making the tournament, and winning the tournament. Additionally, for Ivy League teams, our model calculated each squad’s probability of having the best record in Ivy League games. All probabilities are based on 10,000 simulations of the remainder of the season.

Screen Shot 2017-01-26 at 11.11.00 PM

Note: Due to possibility of Ivy League co-champion, percentages for winning the Ivy League sum to over 100.

We also computed win probabilities for the Bulldogs in each of their remaining games, starting with their upcoming home-and-home against Brown.

Screen Shot 2017-01-26 at 11.09.31 PM

Note: Ties are weighted as half a win.