If you’re a g+ expert and just looking for data, check out our interactive tables.

If you’re looking for a more detailed explanation of g+, read our article on the Deep Dive Methodology of g+.

AMERICAN SOCCER ANALYSIS’S NEW GOALS ADDED MODEL VALUES EVERY ON-BALL ACTION IN TERMS OF GOALS. HERE, IT USES CONTEXT TO ESTIMATE THE VALUE OF SEATTLE’S POSSESSION BEFORE NICOLAS LODEIRO’S SWITCH AS (1.5% CHANCE OF SCORING) - (1.3% CHANCE OF CONCEDING) = 0.002 GOALS AND AFTERWARD AS 3.1% - 1.0% = 0.021. THE DIFFERENCE, +0.019 GOALS ADDED, IS SHARED BETWEEN PASSER AND RECEIVER.

 By John Muller (@johnspacemuller)

Goals added (g+) measures a player’s total on-ball contribution in attack and defense. It does this by calculating how much each touch changes their team’s chances of scoring and conceding across two possessions.

For example, at the moment a player receives the ball at midfield, their team might have a 1.5% chance of scoring on that possession but also a 1% chance of conceding on the next possession. That situation isn’t very valuable. But if they play a throughball from there into the final third, their team is now in a much better situation and might have a 6% chance of scoring and only a 0.5% chance of conceding. The pass would be worth the difference in their team’s situation before and after it, or (0.060 - 0.005) - (0.015 - 0.010) = +0.050 goals added.

How do we estimate a team’s chances of scoring and conceding? By feeding data about the possession into a machine learning algorithm. The model knows where the ball is on the field and some context about how it got there, including what the last action was, how many actions have happened since a stoppage of play, and how fast the possession is moving up the field and across it. The model compares this possession to similar possessions in the past to see how likely they were to lead to goals for either team.

One important feature of goals added is that it splits the value of passes between the passer and the receiver. On easy passes like backpasses, which might be worth negative goals added, most of the value goes to the passer. On harder, more valuable passes like the throughball from our example, the value will be split more evenly, so that around half of the +0.050 goals added goes to the passer for playing an accurate ball and half goes to the receiver for making a well-timed run and getting on the end of the pass.

Goals added only looks at the likelihood of goals. It doesn't give players any credit for actually scoring. So a striker will be rewarded for finding space to receive a pass in a good position and may earn some shot value for turning a possession into a shot on target, but that value won’t change depending on whether the goalkeeper saves the ball. This keeps the model from assigning too much importance to finishing, which statistically is almost random, and instead rewards actions that consistently lead to goals.

Just like goals added rewards players for improving their team’s situation in possession, it also rewards defensive actions that break up opposing possessions. A defender who turns a dangerous situation for the opponent into a neutral situation for their team by winning a tackle will earn goals added by the exact same method as our passer.

Because goals added values every touch in the same way, it can compare different kinds of actions. Instead of counting shots, passes, receptions, carries, take-ons, tackles, interceptions, blocks, clearances, recoveries, headers, and fouls as separate stats, we can add them all up using goals added to calculate the overall impact a player had on the ball. And because that impact is measured in goals, it gives us an idea of how much the player’s touches helped their team’s chances of winning.

--

Action Categories
When summarizing or visualizing goals added, we assign every touch to one of six categories.

  • Shooting: Shots

  • Receiving: Receptions

  • Passing: Passes

  • Dribbling: Carries, Take-Ons, Miscontrols, Dispossessions

  • Interrupting: Tackles, Interceptions, Blocks, Clearances, Recoveries, Contested Headers

  • Fouling: Fouls Committed, Fouls Received

—

Additional Reading on Goals Added