The Expected Goals Philosophy: A Game-Changing Way of Analysing Football

Product Information

Many have described this as the worst individual performance by a striker in World Cup history…and it’s hard to argue when you see the chances he missed. However, I would've liked a more detailed description of how xG is actually calculated rather than just the categories used (e.g., location of the shot, strong vs. weak foot, cross vs. entry pass, air vs. ground, etc.). Plus, there are a bunch of grammatical errors, and that drives me nuts.

These can include which players are over-performing, how adept they are at getting into the right positions and who is benefiting from (or being hindered by) the quality of their teammates. The book uses a surprising amount of anecdotes to talk about a subject which is about large aggregated data. The scouting chapter focuses almost entirely on one club that uses xG; the player xG chapter focuses on certain players like Ronaldo, Hazard, Mane, and a few others; the first half of the book seems to revisit the same Arsenal-Manchester City match a thousand times. Expected goals is a statistic based on aggregated averages, so why not use more examples that incorporate that? Ironically, the author seems to bring up more examples that contradict the power of xG as a predictive tool (like the aforementioned Arsenal-Manchester City match), so, since the author never shows sufficient empirical support for xG as a predictive tool, the anecdotes included in the book only made me more suspicious of xG. It really is laughable. This then shows you that whilst, yes xG can be a good indicator of performance, it doesn’t always tell the full story of a team’s play, and there is no true replacement for reviewing the incidents themselves, rather than as statistics. This can for example be seen in the Premier League at the difference between xPoints based on xG and xPoints based on PSxG. xPoints are expected points scored in the league based on either xG or PSxG.Much of his manifesto feels jarring at first but it is hard to disagree how, far too often, ‘in football, the result dictates the narrative.’ We as “fanalysts” are often consumed by fandom over analysis. A team’s xP for a match is calculated from their xG in relation to their opponents and therefore gives a stronger indication of a team’s form entering a match then their recent results. Nevertheless, by including even more context as we do in our Bayesian model, you get even higher correlations. The context that we add is the following data from Wyscout. Please note that we do not include xG or any other expected something stat. We don’t know exactly when it happened, but some time over the past 10 years the phrase “expected goals” (or “xG”) became a staple of the football dictionary. To this day, we’re not quite sure what it means.

Brighton fans will be well aware of xG, as their team have repeatedly struggled to live up to expectations in the Premier League this year. In fact for the top 30, minutes played has a 37% correlation with goals scored the next season. To be clear, that is still way less than you need for decision making, but it puts to bed the idea that if you want top scoring players, you should look at xG. In fact our proprietary stat scores a 53% correlation for the top 30 top scorers of the Eredivisie, but again this is not enough for decision making. Tippet explains how XG is calculated and why this statistic is important in analysing football results, teams and players. So someone striking the ball into an empty net having rounded the keeper from 10 yards out, would be worth the same as if there keeper were present for a one versus one chance. With 10.79xG created, Brentford have been the most proficient team in the league so far this season at creating goalscoring opportunities," said James Tippett, the founder of the xG Philosophy account and author of the book The Expected Goals Philosophy.

So what\u2019s the book about?

And with at least 15 goal s scored the correlation drops to 53%. Now take a look again at the graphs that show how little information there is for a 50% correlation: