Does a “clutch” goal scorer exist in the EPL? A player who seems to always score goals when the team needs them most. A player who, when the game is on the line, steps up and finishes. This is, of course, a tough question to answer. How do we even define a “clutch” moment? Is a game winning goal in the 90th minute more clutch than a tying goal in the last game of the season? Is a goal to go ahead by 3 in the 70th minute more clutch than a goal to cut the opponent’s lead to 2 in the 35th minute?
To start to quantify these questions, I have created an outcome probability calculator on my blog, Soccer Statistically. This calculator allows you to input the venue, minute, and goal differential of the game. Based on these inputs, you the team’s chance of winning, drawing and losing are given to you. These percentages are calculated based on historical EPL data. Basically, I aggregated EPL game data from previous seasons, and calculated a percent chance of the team winning, losing and drawing in every game situation. Overall there are 90 x 2 x 7 = 1260 of these situations in a game. In some situations, the percentages stray from the nice trend which is exhibited for the rest of the game. To solve this problem, I regressed over the data to create nice trend lines which extrapolate in instances when there is not enough data. This allows us to have an outcome probability for every situation. If you want to try out the calculator yourself, here is a link.
So what does this have to do with clutch goal scoring? Using these outcome probabilities, I am able to quantify the true value of each goal. The problem with ranking goal scorers just based on the number of goals they have scored is that it values each goal equally. Why should the game winning goal in the 90th minute be valued the same as the 5th goal in a 5-0 win? They clearly do not have the same value to the team. To solve the problem I created a metric called Expected Points Added, which weights each goal based on the probability it adds to the team winning. To simplify matters, I multiplied each probability by 3 to instead get the points added, which is a little easier to understand in my opinion. This way, the go ahead goal, at home, in the 34th minute is worth .87 expected points because it increases the team’s expected points by .87. Similarly, the tying goal, away, in the 65th minute is also worth .87 expected points. Each of these goals, although in very different scenarios, increase a team’s expected points by the same amount.
Now that I had a way to measure the true value of each goal a player scores, I was able to do this for every goal of the entire season and then re-rank the top goal scorers. This goal scorers list is ranked based on Expected Points Added, not goals. My list is similar to the actual top goal scorers list. However, there are many interesting changes. Some players, based on their goals, either underperform or over-perform when the weight of each of their goals is taken in to account. To measure this, I also created another metric called Average Goal Weight, which is just the player’s Expected Points Added divided by the number of goals they have scored. This measures how valuable, on average, the player’s goals have been this season. Players with a low Average Goal Weight could be seen as “overrated” based on just their number of goals scored, while players with a high Average Goal Weight could be seen as “underrated” based on their goal total.
Below is the table for the 25 players with the highest Expected Points Added this season: