Goalimpact 1x1
How Goalimpact rates Players
A big thank you also to our Editor Jean Alper Heidemann (@filmnobelpreis)
The Story
In 2004, Goalimpact (or its predecessor) was established. The algorithm's founder, Jörg Seidel, accepted an invitation to wager on the 2004 European Championships from his old coworkers who worked as investment bankers. Jörg was not deeply interested in soccer (at the time; he now supports Werder Bremen), but he was skilled at creating models thanks in large part to his training in finance and physics as well as his proficiency with computer programming. To fill up his betting sheet, he ultimately developed an algorithm for the tournament that determined the likely outcomes of each game.
The algorithm's initial foundation was determining team strength—a soccer equivalent of the Elo Chess Rating. Nobody anticipated the model to rate Greece as strongly as it did. Naturally, Greece went on to win that tournament, and Jörg was successful in his bet. He continued to use the approach when making bets on soccer games as a result of his success. After the summer transfer window closed, he realized that his methodology wasn't as effective anymore, primarily because roster changes had taken place and a team rating didn't account for them. As a result, he modified the algorithm to rate players rather than teams.
This method worked much better because it can still gauge a team's power even after members switch teams. In consequence, Jörg has profited on practically all of his other wagers on sporting events. This was the beginning of Goalimpact.
How Goalimpact rates Players
The system calculates a player's impact on their team's goal difference per minute. Instead of emphasizing descriptive player stats, it only concentrates on their scores. This top-down method was chosen since soccer is a complex game. How do we interpret that? Every action on the field is related in complex networks, resulting in a kind of sophisticated web. Because everything that happens on the field is interconnected, it is challenging to identify the specific actionsthat ultimately help a team succeed.
A soccer match, however, has a clear objective: you play to win. Johan Cruyff once said, "To win, you have to score one more goal than your opponent." A soccer game aims to score goals while preventing goals from being conceded. In essence, having a favorable score is everything in soccer.
As a result, it is possible to define a good player in the following way:
A player is good if their squad score more but concedes less goal while he is on the field. No matter why and how.
The method calculates the impact of a single player on goal difference. Goalimpact's methodology is thus supported by an objective definition of a "good player."
The key distinction between the concept and the +/- score in basketball is that the former does not take into consideration the opponent's strength. Therefore, a high +/- score does not necessarily indicate that a player is good; rather, it may indicate that they have only faced weak opposition. Goalimpact overcomes this difficulty by considering opponent strength,which emerges from the individual player Goalimpacts. Therefore Goalimpact is not an absolute score, but a relative rating and hence allows comparison of all players across the world.
The algorithm just needs match data as inputs, such as the starting lineup, goal minutes, player subs, and birthdates of the players. Starting with 1,000 minutes of playing time, the computer can determine player quality using these data. Red cards, fatigue levels, and home-field advantage are also factored into the formula. Goalimpact has more than 500.000 players in their database from all around the world because such information is readily available. This allows also to calculate a forecast of the player's future potential using their date of birth, and the huge database. The highest expected Goalimpact is called Peak and is around the age of 26.
How does Goalimpact learn?
Consider a matchup between Manchester City and Manchester United. At minute 30, City scores to take a 1-0 lead. Donny van de Beek replaces Casemiro for United at the break. At the 60-minute mark, United scores a goal, tying the game at 1-1. In this simplified example, assuming equal team strength, just two players' ratings would be modified because the final result was 1-1, meaning that most players had a goal difference of zero. The algorithm would raise van den Beek's rating and lower Casemiro's because their individual scores were 1-0 and 0-1, respectively.
One thing to keep in mind in this example is that all players are expected to perform equally as it is assumed that this is the first game played ever in the algorithm. However, since the Goalimpact algorithm will take historic player strength into consideration in all ensuing matches, the Algorithm adjusts for team strengths; for example, if Manchester City plays Plymouth Argyle, we would anticipate City to win with a goal difference of perhaps three. The players from Plymouth would receive higher ratings even though they lost because they lost by fewer goals than anticipated if City only wins 1-0. Hence Goalimpact is not just a naive +/- score but rather a relative score.
One game, however, is insufficient to confidently calculate the rating because, after all, one game can be pretty random. With more information, the Goalimpact algorithm can change, leaving only player quality.
But enough of the dull theory; how about an illustration?
Example: Alphonso Davies
Alphonso Davies' rating was one of Goalimpact's success stories. His market value was less than €500,000 when Goalimpact started to rate him as having world-class potential. He played in Canada.
The blue line details his career average Goalimpact rating over time, while the red-dotted line is the career Goalimpact predicted rating based on average player development. A Goalimpact rating of 140 or higher typically implies top-five European League potential. The top athletes globally have Goalimpact ratings of above 170 as a point of comparison. Thomas Müller, Sadio Mané, and Dani Carvajal are the players with the highest Goalimpact ratings.
Goalimpacts radical proposition
The fun part of Goalimpact is that it identifies a lot of players that are over or underrated by the football market. Football is highly emotional and therefore prone to a lot of cognitive biases. These can and probably do cost managers and clubs a hell lot of money. Our proposition is that you can identify market inefficiency and benefit from it.
Our algorithm finds players from low leagues across the world who have the capability to play in the top leagues.
This is what motivates us and what we offer to the current football circus. It might be a radical thought, but so was the initial Moneyball project.