Elo 2017: Mid-Season Recap
We have now reached the halfway point in the season for Ivy League games. Every team has played four Ivy games and, as of right now, Princeton is leading the Elo race. However, the actual standings look more interesting. Yale, Columbia and our Cornell Big Red all sit tied for first in the Ivy League with 3-1 conference records, while Princeton is tied for second at 2-2 with Dartmouth and Harvard.
Looking back to last weekend, there were again four Ivy League games with many interesting results. In the first game of the weekend, Harvard defeated Dartmouth by three points in Cambridge. It was a good showing by Harvard after losing its two Ivy games (Princeton and Cornell). On the Elo side of things, Harvard was given a 64% shot of winning the game and increased 10 Elo points after the victory.
Also last weekend, University of Pennsylvania played Brown in Providence, RI. Penn secured a comfortable 17-7 victory, which was not surprising considering the model gave the team a 65% chance of victory. Brown lost 15 Elo points with that loss and dropped to dead last in Elo rating.
The Columbia at Yale matchup was the highest stakes game of the week. Yale only had one loss so far in the season and Columbia was undefeated. Columbia was and still is having an unbelievable season after years of being one of the worst in the Ancient Eight. The game, however, was not all that close. The model had given Yale a 67% chance of victory and it turned out to be correct as Yale secured a 23-6 victory over Columbia. The win gave Columbia its first loss and made Yale and Columbia tied atop the Ivy League.
Even with all of the other matches and story lines, Cornell was the highlight of this week. Cornell was set to play at Princeton. The team was coming off a comfortable Homecoming win against Brown and was hoping to carry its momentum and pull off another upset this season. Opposite them, Princeton was playing in top form. The team had only dropped one Ivy League game and had scored more than 50 points against its last two opponents in massive blow-out wins. Last year, Cornell had lost at home to Princeton 56-7 and the Big Red saw this Saturday as its chance for revenge. The model gave Cornell a slim 16% chance of victory, but after a nail biter of a game that came down to two long field goals in the final minute, Cornell squeezed out a 29-28 win. The game put the Big Red into a first place position and increased its Elo rating by 13 points. Even though the team has the second worst Ivy League Elo rating, Cornell has won its past three Ivy games and is on track for a monumental season.
So far this season, Elo has been less than perfect and has predicted fewer than half of the games correctly. It will soon be coming up on the end of the season and we will reevaluate the model to see what improvements we can make. However, as the season progresses and Elo gathers more data on the teams, I would expect the model to improve in accuracy. It is also important to notice how unusual this season is. Two of the worst teams of the past years – Columbia and Cornell – are winning games in huge upsets. The average person can adjust quickly to these season-to-season changes, but Elo takes time to adjust and the project will not immediately be corrected for a sudden improvement in a team.
Below are the updated Elo ratings for teams.
The Elo model sits at 7-9 in game predictions so far. Check back later this week for the weekly Elo predictions.
CORRECTION: Last week’s Elo predictions incorrectly displayed Brown as having a 57% chance of beating Cornell at Homecoming. Instead, the correct values had Cornell showing a 57% of victory instead of defeat. The Elo calculations were adjusted correctly above.