A player’s individual performance is intrinsically linked to how they fit the style and dynamics of a team. That’s why it’s essential, when recruiting a new player, to assess how likely they are to gel with other players in the squad. Our Player Chemistry model uses real-world chemistry ratings to predict how well the playing characteristics of a new recruit will complement those of his prospective teammates to produce positive results on the pitch.
By analysing millions of passages of play, our Player Chemistry model uses a neural network to identify the most prominent player characteristics that yield positive match outcomes. These characteristics include on-field playing style metrics, as well as off-field human factors.
Identifying the characteristics that result in positive match outcomes also allows us to use real-world event data to value the chemistry between teammates. Our Player Chemistry model analyses how often players are involved in different passages of play together, building networks for different outcomes, such as goals, scoring opportunities, transitions or losses of possession. Weighting each outcome type enables us to define what makes a successful pairing and evaluate this over a number of seasons to produce a chemistry rating for each pair of players.
Our Player Chemistry model has been extensively researched and published at a world-leading AI conference. Click here to read the published paper.
In football, there’s always a sense that some players bring the best out of each other. Our Player Chemistry model takes a holistic view of each player pairing, analysing a range of interactions that lead to positive outcomes on the pitch. The below graphics show our ratings for the most effective attacking pairings in the Premier League over the last three seasons (correct as of date posted, 19th November 2020).
You would generally expect the most effective pairings to play in the most successful teams, however, this wasn’t the case during the 2018/19 Premier League season. Whilst Bournemouth finished in a respectable 14th place, Ryan Fraser and Callum Wilson were rated as the most effective attacking pairing. Demonstrating their chemistry on the pitch, they combined for 12 goals – only Alan Shearer and Chris Sutton (13 in 1994-95) have combined for more in a single Premier League season.
Fraser and Wilson signed for Newcastle United in the summer following Bournemouth’s relegation, with their proven chemistry clearly influencing the decision to recruit both players. It also made financial sense, with Wilson signed for £20m and Fraser snapped up on a free transfer. The pair have talked openly about their instinctive understanding of how each other play but they are yet to have the same impact on Tyneside. Fraser’s assist for Wilson’s second goal in the recent win over Everton was perhaps a sign they were starting to rekindle their partnership. However, subsequent hamstring injuries for both players means The Magpies will have to wait to see if one of the Premier League’s most effective attacking pairings can get back to their best in a black and white shirt.
Signing two players with proven chemistry is rare, and usually comes with a hefty price tag. At the same time, predicting chemistry between players who have never played together can be difficult and often based on subjective judgements. That’s why we’ve developed our Player Chemistry model to identify trends in player characteristics and apply chemistry ratings from real-world pairings to predict how prospective pairs of players will perform together.
The AiA Simulator, available via our online portal, predicts how a potential signing will perform with other players in a prospective team using the Player Chemistry Index. This metric analyses chemistry at a team level, based on an average rating that is weighted by minutes played across the past season. Combining our Player Chemistry Index with predictive indicators of Playing Style Suitability and Cost-Benefit, the AiA Simulator is invaluable for football clubs and player representatives, providing an objective view of a proposed transfer by predicting likely success and value for money.