18/01/16 | SciSports
Expected Goals (ExG) has recently become one of the most recurring concepts in football data analysis. In short, Expected Goals is the amount of goals a player or team is expected to score, based on the amount and location of the shots they have taken.
For example, a shot in the six-yard box is more likely to be converted to a goal than a shot from 40 yards out. Therefore, the former shot will have a higher ExG-rating than the latter. The rating system assigns each goal scoring attempt a rating between 0 and 1.
The concept has become increasingly popular and can show whether or not a team over- or underperformed in a match (or season), based on their expected goals scored and conceded.
For example, consider a hypothetical match between Manchester City and Everton. If the expected goal result is 1.32 for City and 1.05 for Everton, but the actual result is 3-1, you can say, based on the model, that Manchester City has overperformed in this particular match.
Several data analysts have created different models to calculate the Expected Goals and therefore use different factors and different magnitudes of these factors. For example, data blogger 11tegen11 uses the following factors in his model:
- Shot location
- Shot type
- Big chance or not (assigned by human coders)
- Start of possession
- One pass after a through ball or not
- Cross or not
- Vertical speed
- Number of touches before shot
- Game state
The ratings assigned to these factors are usually based on data obtained through analysises of match situations and leagues in recent history. Not all analysts use models that are as comprehensive as 11tegen11, however. Most ExG-models usually consist of at least shot location and shot type and additionally might use a few other factors.
With the rise of Expected Goals, several derivatives of the model have surfaced, such as Expected Assist, Expected Points and Expected PDO.
If you want more information about Expected Goals, check out the following links: