SportsEngine's algorithm is a sophisticated, formula-based process that's grounded in the Elo rating system. Originally created for ranking chess players, the Elo rating system can be used for any head-to-head competition.
At its core, the Elo rating system calculates a team's score and the likelihood of them beating their opponent. Based on the outcome of the game, the team's rating is updated. If a team wins unexpectedly, their rating will go up more than if they win as expected. This helps keep scores fair and prevents teams with long winning streaks from having a score that's too high to beat. Over time, their rating may level out or even decrease if they're no longer winning convincingly.
The Elo rating system formula is: New Rating = Old Rating + (Weight * Margin of Victory * (Actual Outcome - Predicted Outcome))
Weight is based on the timing of the game within the season, with higher values given to important games like the finals and grand final. This balances the impact of regular season games and important games on a team's score.
The Margin of Victory multiplier takes into account how a team won, giving more weight to underdogs and limiting the impact of blowout wins.
Actual Outcome and Predicted Outcome are calculated using a team's rating before the game, the score, and which team won, respectively.
A variable is included for the overall strength of competition in each state compared to other states. This means that if a state's teams struggle against teams from another state, it could affect the rankings of teams within that league. This allows for ranking teams across state lines.
At the start of each season, the ratings are "softly" reset to account for changes in team rosters. This is because players come and go, and resetting the ratings helps show these changes. Once the competition starts, the ratings stabilize quickly and reflect the impact of new players.
SportsEngine's algorithm factors in the Elo rating system and many other variables. It's always improving, with new modifications made regularly to make it even more accurate. The SportsEngine team is currently working on formulas to include factors such as home-field advantage, rivalries, and individual team performance in both Offense and Defense to provide the most comprehensive tool for comparing teams across Australia.