Written by Adam Rae Voge | Soccer Writer and Data Analyst
Evaluation is central to the sports fan experience. Results matter, sure, but arguments between fans often get lost in the fine margins between good and great, or between poor and atrocious. It makes sense, then, that the pursuit of understanding remains central to sports fandom, and perhaps grows by the year.
Soccer is no exception. Since its invention, really, soccer lovers have looked for ways to evaluate player performance. It used to be simple: goals, assists, tackles, saves. But over the past several years, the experience of reading the stats from a soccer game has become increasingly complex, with passing networks, field tilt, and “expected” goals measured and analyzed.
Expected goals, also known as xG, are typically at the forefront. On any given day you may find some of the more mathematically inclined corners of soccer fandom debating a player or team’s xG and what it means for their outlook.
But just what is xG? How is it measured, and how much does it matter when you’re trying to judge a team’s performance? Let’s break that down a bit below:
The simplest way to think about xG is as a measure of probability. At the instant that player shot the ball toward goal, how likely was it to end up in the net? xG is calculated in slightly different ways by different outlets, but it aims to calculate that likelihood, taking into account such factors as the distance to the goal, the angle, the body part used to shoot, and more.
Let’s use the recent MLS Cup semi-finals to gather some recent examples. In the 88th minute of Philadelphia Union-New York City FC, Talles Magno received a cross from midfielder Gudmundur Thórarinsson and fired a left-footed shot into the back of the net. The goal gave NYCFC the win and sent them to the MLS Cup final.
Using xG, we can determine that the exact same scenario would have netted a goal quite often. MLS’ internal xG rating put that shot at a 68% chance of being a goal, or 0.68 xG. Another way of looking at it is that mathematically, that situation generates goals about 2/3 of the time. Had Magno missed it, then, it likely would have been a hard pill to swallow for NYCFC fans.
How about the other semi-final? The Portland Timbers largely handled business against Real Salt Lake to earn their own trip to the cup final, but traditional statistics such as possession (51% to 49%), pass completion percentage (79.6% to 79.5%), and duels won (54 to 57) don’t really tell that story.
Where the difference occurred was in the number and quality of chances created by either side. Portland’s defense limited Salt Lake to just 0.4 xG, meaning that mathematically the sum of the chances generated most of the time would not be likely to result in a goal. Salt Lake’s single best chance came in minute 33 when Damir Kreilach’s header from the edge of the six-yard box was saved. Despite a beautiful cross to set up the chance and a very good header, Kreilach’s chance was rated at just over 10% odds of becoming a goal, or 0.1 xG. The good finish didn’t change that.
To contrast, Portland’s Santiago Moreno scored an absolute world-beater of a goal from five or so yards outside the penalty area in the 61st minute, which made the score 2-0 and effectively ended the game. But mathematically speaking, Moreno’s shot was not a good chance. It was rated at just 0.03 xG, a very against-the-odds chance that really highlights the quality of the finish. Had he missed, though, Moreno likely would have been criticized for shooting in the first place.
What’s the point of all these numbers anyway? After all, isn’t the final score the only thing that matters? Well, yes and no.
Strong xG believers will tell you that teams in good standing who have poor xG and xG allowed are headed for a crash. After all, if you’re generating fewer chances and giving up more, you must be getting lucky.
The converse is true as well – a team with poor standing but a good xG difference could expect that things will turn around soon. You’re clearly generating quality chances and limiting your opponents, which is what the game is all about.
xG can also be used to measure how reliable a player is at scoring. A player who routinely misses xG chances over 0.2 or so will be judged as a “poor finisher,” since they’re failing to score when many players would. A player who has a lot of goals that come with very low xG could be judged as going through a “purple patch,” or an especially lucky period. Or, depending on which side of things you fall on, the player could just be a really good shooter.
As with any statistic, xG watchers should be wary of anomalies, small sample sizes, and other odd events that may affect how repeatable results are. For example, a team may have 1.0 xG from a single play if a player’s high-probability chance is saved and their teammate fluffs the rebound.
With your new xG knowledge in hand, may you go forth and evaluate your teams expertly. At the very least, you’ll have a new metric to bring up in arguments.
