Each week, BRSN sits down for a Q & A session with...
Ivy League ELO
The Big Red Sports Network will be releasing a rating system for Ivy League football. We will be using a relatively simple structure called ELO Ratings. The system was developed by Professor Arpad E. Elo in the 1950’s as a way to rank chess players around the world. Even though the classification was originally intended for a different activity, it can be adapted and used for a variety of multiplayer competitions. ELO has previously been used to rate national soccer teams, NFL teams, MLB teams, and was used briefly to help select the teams for the College Football Playoff.
ELO has a strong history in sports, but why did we use it over another rating system? The most important reason is that it’s a very simple system. ELO only uses the final scores of games to adjust team’s ratings, which makes it easy to calculate many years of ratings and is helpful for eras with poor record keeping. It can also show win probabilities and point spread for games based solely on the teams’ ratings. Another benefit of ELO is that it always increases a team’s rating when they win and lowers it when they lose.
The way ELO displays a team’s strength is with a score. All teams start out with a rating of 1,500, which represents an average team in the league. As they play games, each squad will gain or lose points. The nice thing about ELO scores is that if one team gains points, its opponent loses that many points and vice versa. Therefore, there is a finite amount of ELO points that Ivy League teams are fighting for.
So how do we actually calculate these scores? The formula used and the variables needed are shown below:
The formula is relatively simple – it takes the ratings of the teams before the game and adjusts them according to how they were expected to perform. For example, if Cornell has a rating of 1,650 and wins at home against a team rated at 1,350, then Cornell’s rating will increase, but only by a couple ELO points. However, if the results were flipped in that scenario, the other team would gain 20+ ELO points because they were underdogs.
The other factor to consider is the score differential at the end of the game. The “Margin of Victory Multiplier” is a messy equation, but it just takes into account how much a team wins or loses by. If a team pulled out a one-point victory, it would not gain the same amount of ELO points as a team that won by 30. However, the multiplier only rewards scoring to a certain extent. For instance, a team scoring more in a blowout doesn’t really matter, and a win of 25 points compared to a win of 40 points will not increase the team’s rating by much more.
The last piece of ELO calculation is the end of season soft reset. After each year, the teams’ ratings are brought back to the mean of 1,500 by ¼. For example, at the end of a season, a team with a rating of 1,700 will begin the next season at 1,650. This calculation is done in order to adjust for turnover in teams and to give them a fresh start for each season, while still retaining some of the program’s historical rating.
Let’s look at a more concrete example – the ratings for the 1956 and 1957 Ivy League seasons. These are the first two years that we will be rating because they are the first two years where all Ancient Eight teams played each other and the first two years that recognized an Ivy League champion. Thus, 1956 will act as the starting point for these ratings. Another important note is that only games amongst Ivy League teams will count toward the system.
The graph is straightforward with teams’ ratings plotted in comparison to the weeks of the season. In addition, the team (or teams in the case of a tie) that win the Ivy League title has a yellow marker at that end of their championship season. The 1956 season’s ratings are also not completely accurate because every team starts from a rating of 1500. It takes a couple of seasons before the ratings can accumulate historical data and become more precise.
We will be continuing to calculate ELO ratings up to present day. During the 2017 football season, we hope to have weekly updates to our ratings along with win probabilities for all conference games. ELO ratings will give us a new and interesting perspective on Ivy League football.