Why the Colley Rating Method Beats Simple Win‑Rate Rankings

Main Idea of the Colley Method

The Colley method modifies the traditional win‑rate formula by incorporating schedule strength, i.e., the quality of a team’s opponents, using a linear system derived from Laplace's rule of succession. All teams start the season with the same initial rating, and as games are played the ratings adjust based on both wins and the strength of opponents.

This approach avoids ties common with simple win‑rate rankings, rewards victories over stronger teams, and eliminates unfairness where beating weak teams yields the same benefit as beating strong ones.

Example

A small example using a table of match results demonstrates how to construct the Colley linear system. The system can be expressed as Cy = b , where C is the symmetric positive‑definite Colley matrix, y is the vector of team ratings, and b is a right‑hand vector derived from win‑loss records.

Because C is symmetric positive‑definite, it has a unique solution and can be efficiently solved via Cholesky decomposition or standard numerical methods such as Gaussian elimination or Krylov subspace methods.

Key Points of the Colley Rating Method

The Colley matrix is a real symmetric positive‑definite matrix.

It uses the total number of games each team has played.

It incorporates the number of games between each pair of teams.

The right‑hand vector captures each team’s accumulated wins.

The method also accounts for accumulated losses.

The solution vector provides a generalized rating for all teams in the league.

The matrix dimension equals the number of teams in the league.

The algorithm solves the linear system to obtain the Colley rating vector for each team.