The expected goals philosophy - book review
What is xG and why is everyone so interested in it?
I first saw this on Twitter with an “xG” scoreline being posted for a Manchester United game I was watching. That day, I gave it a follow so that I could brag about how I was following it before it was cool, without never really knowing what it meant!
A year or so went by and it ended up slowly fighting its way into the public eye after a Jeff Stelling rant and a Mikel Arteta press conference. It was at this point I found myself thinking ‘I don’t really know what it is!’ So I eventually took matters into my own hands a month later and ordered the book.
I was not particularly for or against this method of analysing football matches, but I was intrigued as to how it is calculated. I will be splitting my review into two parts: "what is xG?" and "how useful is xG?".
So, answering the first big question - "what is xG?". The name stands for expected goals, and this is equivalent to the chance of the "average" player scoring a shot taken in a game. This is calculated using thousands of past shots across many different leagues and it will give a numerical probability of an average player scoring that chance.
My biggest issue with this at the start was how the premier league is far from a league full of average players, so what would be the use in calculating the expected goals of a team like Manchester City where a player like De Bruyne is far more likely to score from outside the box than your average player? Despite my point still standing, it turns out player xG really doesn't differ too much (as seen in chapter 6). Of course some players will be more effective at longer range shooting and some players (such as Michu's glorious season at Swansea under Michael Laudrup) will go through spells where they outperform their xG simply due to a little more confidence in front of goal. Despite all this, the fact stands, a player does not tend to consistently, vastly outperform their expected goals. The example the book uses is Cristiano Ronaldo, undoubtedly one of the best finishers in the world, who in his last three seasons at Real Madrid only scored a combined two more goals across three seasons than he was expected to. This shocked me with Cristiano being among the top scorers in all competitions he was involved in during those three seasons, suggesting the validity of the cliché "the best strikers are always there to take the chances". Of course the only exception of this is Lionel Messi who consistently outperforms his xG, due to the abundance of long range beauties he has consistently scored over the duration of his career.
So to sum up, xG is the expected goals of shot going in if the ultimate average player were to take the shot. Each shot is given a numerical value of xG regardless of how big or small this may be, giving an overall total of xG over the duration of a game. This can also be split into individual players and it can be a useful tool to measure whether a player is outperforming his expected goals or maybe lacks a little confidence and is underperforming, rather than just looking at his goals scored and shots taken to make the conclusion. The most amazing thing about expected goals is that if taken over the course of a season, it is very rare that a player scores a significant amount more than his expected goals total.
Next we address my biggest issue, the way we currently use xG. Despite my praise for the way its calculated and how useful it could be, I believe it is wasted as a method of football analysis in the media. I'm sure the scouting teams like Brentford and Arsenal who invested in their own football analysis companies that are said to be the pioneers in xG use this in a completely different way to the rest of us telly clappers.
My first question to you is when have you seen xG? My best guess would be seeing the expected goals score line of the game on match of the day or twitter. I am a firm advocate of using this in isolation as it then becomes just as useless as the statistics like shots on target that it is aiming to replace. Two teams can have similar xG despite one team dominating. This means that a team that takes lots of speculative shots can eventually accumulate a similar xG to a team that has one big chance and dominates the majority of the game. Because of this, I am going to suggest three different ways to look at xG available for free that will give you a far better "story" and reflection of the game.
The first is xG maps. These use proportional symbols to show the location of each team's shot to show you which team has the "better chances" to score, regardless of the simple xG scoreline. The app I now use for this is free to download on the apple app store and is called "InfoGol". This provides an xG map of each team so you can see where the xG total has come from and form a fair opinion on which team had the better chances to score. I am amazed we don't see these maps on sky sports or BT as they are a far better reflection of the game than an xG score line.
One of the more obscure ways of presenting xG is the expect goals timeline of a game. These can be found on twitter posted by @experimental361. These simple yet intuitive graphs show how each team's xG develops across the 90 minutes. This is a brilliant way of highlighting how long each team may dominate a game for. For example, should a team score 3 early goals and accumulate a very high xG during the first 20 minutes but then sit back and defend for the next 70, the xG score line may not be reflective of which team dominated for the longest period of time.
Finally there is xP. This is known as expected points and uses an algorithm along with expected goals to predict the expected points of a team if the game was to be played thousands of times. Returning back to the idea of teams having a high xG due to many speculative shots but never really looking like winning the game, this will show how the xG scoreline does not necessarily mean the team with the highest xG was the most likely to win. I would recommend looking a bit more extensively into xP and justice tables, but if you want a quick idea of how xP works, you can enter the individual xG of each shot from each team into this website (Match Expected Goals Simulator (danny.page)) and find an xP value for who was most likely to win the game and compare this to the xG scoreline.
All in all, the xG score line doesn't really allow any further football analysis than the original score line itself. To some extent, I agree that the xG score line is a useless stat, but only because of the way we currently use it. The use of xG maps, xG timelines and xP can really utilise the xG data and tell a much better story of the game, making it a very useful method of football analysis.
Using one word to sum up the metric of xG, I would call it "promising". I am sure it is far more than this in the current professional football analysis world, however for the casual viewers and the media, I believe it has the potential to be a very effective method of comprehending the performance of individual strikers to the chance creation of a whole team over the course of the season. As interest grows, I'm sure the research into the methodology will grow making it more and more useful as time goes on. For the mean time, I encourage you to think about your position on the philosophical spectrum of whether football is an art or a science and reconsider the way your formulate your footballing opinions.
I would highly recommend the book, with this article giving a little insight to the conclusions you may make. Reading this will armour you with case studies (such as the Brentford revolution), insight into the betting syndicates and information companies (such as Opta and Smartodds) and challenge the way you currently rank players, teams and divisions.
Please don't take my word for it, read it yourself! The amazon link is below.
Honourable mentions & further research
- The Brentford revolution & Matthew Benham
- Arsene Wenger and his public support of the xG methodology
- Mikel Arteta and the validity of his concusions on Arsenal's start to the 2020/2021 season
- @experimental361 on twitter
- The InfoGol app
- James Tippett (obviously!) - Author of the xG book
- Opta and the faults in their current xG methodology
- Why xG is still prone to human error
Comments
Post a Comment