What is Elo?
The rating system behind every RPI ranking, adapted from chess to rowing.
Origins in Chess
The Elo rating system was developed by Hungarian-American physicist Arpad Elo in the 1960s as a method for ranking chess players. It was officially adopted by the US Chess Federation in 1960 and later by FIDE (the World Chess Federation) in 1970. The key insight: instead of just counting wins and losses, Elo accounts for who you beat and who you lose to.
How It Works
Every competitor starts with a base rating (typically 1500). After each contest, ratings adjust based on two factors:
K is the sensitivity factor (how much a single result can change your rating). Actual is whether you won (1) or lost (0). Expected is the probability of winning given the rating difference between competitors.
If a highly-rated team beats a lower-rated team, ratings barely move (the result was expected). But if the underdog wins, both ratings shift significantly. This self-correcting mechanism means ratings converge on true ability over time.
Expected Win Probability
The expected score between two competitors is calculated using a logistic function:
A 200-point rating difference translates to roughly a 75% expected win rate for the higher-rated competitor. A 400-point gap means about 91%. This gives an intuitive meaning to rating differences.
Beyond Chess
The elegance of Elo is its generality. It has been adapted for virtually every competitive domain: FIFA uses a variant for international football rankings, the NBA and NFL use Elo-based models for game predictions (popularized by FiveThirtyEight), online gaming platforms like League of Legends and chess.com use it for matchmaking, and academic research applies it to everything from animal dominance hierarchies to machine learning model evaluation.
Why Elo for Rowing?
Rowing presents unique challenges that make Elo especially well-suited:
- Multi-team races
Rowing events often feature 4-6+ crews per race. We decompose each race into pairwise comparisons, updating ratings for every pair of crews that raced together.
- Infrequent matchups
Teams from different regions rarely race each other directly. Elo propagates strength through chains of competition: if Team A beats Team B, and Team B beats Team C, we learn something about A vs. C.
- Time-based margins
Unlike chess, rowing provides margin of victory in seconds. We use this to make larger rating adjustments when a win is dominant versus narrow.
- Variable conditions
Wind, current, and course conditions affect results. Our modified Elo dampens rating changes in poor conditions where results are less reliable.