Expected Goals (xG)
The ‘value’ of a shot or probability that a shot results in a goal depending on the distance and angle of the shot. For a better model and explanation check out the links in Michael Caley’s pinned tweet.
My method only takes into account the angle and distance of the shot, then a logistic regression is run giving the formula for any given angle and distance.
Expected Assists (xA)
I’m not too sure how this is generally calculated. For my values I took Key Passes and ran a logistic regression where the dependent variables are the start and end location of the key pass, whether or not it is a through ball and whether or not it is a long ball. Then did the same for crosses but using just the start and end location.
I’m fairly happy with the results but both this and expected goals but they also still need a lot of improving.
Expected Passing (xP)
This is taken from @FootballFactMan on Twitter explained in his post here. Like expected goals it just assigns a value to each pass based on the probability that is is completed. For this I split the pitch into 100 (10 x 10) zones and just did number of passes starting in one zone and ending in another divided by the number completed.
Expected Goals Added (xG Added)
Taken from @NilsMackay and explained much better by him here xG Added takes the xG value at the start location and end location of a pass then subtracts the start from the end to give the xG that was added from the pass.
Possession Adjusted Stats
These come from StatsBomb explained in a piece here. It’s a way to even out tackles and interceptions as these stats are much easier to rack-up for a team that spends the majority of the match without possession of the ball. I use the formula at the bottom of the article by @statlurker.