Follow predictions by ACE on our Twitter feed.
The Athlete Contribution Estimator (ACE) is a statistical model that quantifies how the performance of individual rugby union players affects match results.
To predict a match outcome ACE calculates the expected contribution of each player to their team’s overall score, based on their past performances and recent form. These contributions are combined with ‘team dynamics’ and home field advantage terms to produce a predicted margin, defined as the score of the home team minus the score of the away team.
Not all attributes that contribute to scoring and conceding points are captured by available player data – there’s obviously more to rugby than just individual players throwing passes and making tackles. The team dynamics term attempts to capture broader team attributes such as the quality of their set-piece, effectiveness of tactics and other coaching factors.
The advantage of this player-based approach is that the model can immediately adjust for changes to the match-day lineups due to injuries and suspensions etc. while still accounting for the overall strength of the teams, as well as highlighting interesting match ups between players.
For each position, ACE can also rank players from different teams according to their contributions to match outcomes and ultimately assist national team selections.
An example ACE match prediction from round 6 of the 2018 Super Rugby season is shown below. The starting players are listed head-to-head along with the net differences between their expected points contributions. Positive numbers indicate that the home team has an advantage, while negative numbers indicate the advantage lies with the away team.
At the first-five position, ACE estimated that Beauden Barrett gave the Hurricanes a 2.0 point advantage over the Highlanders, who fielded fellow All Black Lima Sopoaga. Comparing fullbacks Jordie Barrett and Ben Smith, ACE estimated that Smith gave the Highlanders an advantage of 1.4 points. Elsewhere, flankers Brad Shields and Elliot Dixon’s contributions were expected to cancel each other out.
The sum of the net score contributions over all players favoured the Hurricanes by 3.4 points, and their superior team dynamics gave them an additional 6.1 point advantage. Adding the home field advantage value (estimated as 4.5 points for Super Rugby) to these terms resulted in a predicted net score of 3.4 + 6.1 + 4.5 = 14 points. The actual result of the match? A 17-point victory to the Hurricanes.
A beta version of ACE operated for selected Super Rugby games in 2018 and early 2019. The first complete version of ACE began operating in round 2 of the 2019 Super Rugby competition. Development and testing of ACE will continue throughout 2019 and beyond.
ACE is developed and maintained by Niven Winchester and William Beard. Winchester is an economist at the Massachusetts Institute of Technology and founder of RugbyVision.com. Beard is a research engineer with a background in statistical modelling and analysis.
To find out more about ACE or to enquire about further applications of our research, please contact ace@rugbyvision.com.