Contributors

This website is run by Matt Trenary and George Janke, along with others involved with University of Michigan Rugby. The site's code and and its update history are public in order to eliminate any fear of favoring Michigan. Find our GitHub here.

RugbyHawk’s Elo rating system is based on a master’s research project by Northwestern University’s Marcus Thuillier as well as FiveThirtyEight’s National Football League Elo implementation.

While RugbyHawk’s Pairwise Ranking system is tailored to college rugby, it is inspired by USCHO’s version for college hockey.

Elo Rating System

The Elo rating system was created for chess and has been adopted for a variety of sports, including rugby. This site and this site help explain our implementation for Elo, and our code is on GitHub.

The primary features of an Elo system are:

• Teams gain rating points by winning and lose rating points by losing.
• The gained and lost rating points for each game are equal; it's an exchange of points.
• Teams gain more when beating higher rated teams and gain fewer when beating lower rated teams (vice versa for losing).
• Margin of victory impacts the points gained or lost, but the impact tapers off as the margin increases.
• Home victories earn fewer rating points than away victories (vice versa for losing).

All teams start at 1500, which is considered the average rating.

Pairwise Ranking System (PWR)

RugbyHawk’s Pairwise Rankings emphasize head-to-head victories in order to determine which teams are “most deserving” of playoff selection. In the Pairwise, every team is compared against every other team, and the better team in each comparison gets a point. Teams are ranked by the total number of comparison points that they have.

Each comparison uses three criteria:

1. Head-to-Head Winning Percentage
2. Winning Percentage Versus Common Opponents
3. Elo Rating

The criteria are evaluated in order, and when either team wins a criterion, they are considered to have won the full comparison.

Eligible Teams

It’s only possible to give teams an accurate rating/ranking if there’s enough data of their games. More specifically, teams need:

• Enough Games: If a team has less than 5 games played, we don’t have a large enough sample.
• Connectivity to the Broader Network: We need to avoid bubbles of teams who have played enough games, but have only played each other.

If a team doesn’t meet our eligibility criteria, we still give them an Elo rating, but their starting rating is docked 200 points to 1300, since teams who don’t meet the criteria tend to be worse than teams who do. Additionally, ineligible teams aren’t included in pairwise comparisons, so they have pairwise scores of 0.

If your team isn’t eligible, you can help by digging up more game scores and reporting them here.

If your team doesn’t appear on this website at all, it’s because our records say they haven’t played any Division 1 games since Fall 2021.

Expected Updates

• Optimization for Mobile: Our results tables require horizontal scrolling on mobile. We need to make your viewing experience better on mobile devices.
• Algorithm Explanation: While RugbyHawk’s Elo algorithm is already public on GitHub, we should provide the formula on this page in order to make it easier to understand. For now, fill out any questions you have about either the Elo or the Pairwise in the form below, and we'd love to explain.

Have feedback? Please fill out this form to let us know.

Submit