Analysis: Minnesota United demonstrates there’s more to the statistical picture than expected goals

In the world of sports statistics, there are only a few numbers that are concerned what “should have” happened, not what actually happened. One is errors in baseball, a stat that dates back to the 19th century. Another is a soccer (and sometimes hockey) statistic, one that has only been popularized in the 21st century but has come to dominate much of the talk about the game: expected goals.

Minnesota United is third in the MLS Western Conference standings with 29 points going into Saturday night’s game at Seattle. The Sounders are 11 points behind Minnesota. Yet a glance at the expected goals numbers shows the Sounders have a positive expected goal differential for the season, ahead of the Loons.

What xG, as it’s usually abbreviated, represents is an attempt to assign a number to every shot based on how likely it is to become a goal. It’s become far and away the most-quoted “advanced” statistic in soccer. Coaches will reference it in postgame news conferences, minutes after the final whistle.

The issue is that, just like baseball errors, xG calculations represent subjective judgments — no matter how many computers are used in the calculation. To illustrate this, let’s look at two moments from the Loons’ 2-2 home draw with the LA Galaxy on May 15.

The first came in the first minute of the second half. Loons wingback DJ Taylor crossed the ball for Tani Oluwaseyi, but the ball was slightly behind the striker, 6 yards in front of the Galaxy net. Oluwaseyi, who had a defender on his shoulder, attempted to reach out his right foot behind himself and redirect the ball behind himself and towards the far post — but ultimately the ball rolled 2 yards wide of the goal.

On the other end 15 minutes later, L.A. winger Gabriel Pec made his own pass across the face of Minnesota’s goal, behind the Loons center backs — where he found a wide-open Dejan Joveljíc for what amounted to a tap-in goal for the Galaxy forward.

Now, a subjective observer sees those two plays and knows that Oluwaseyi’s attempt would usually be described as “audacious” — a moment of invention, with little probability of finding the net. Meanwhile, the observer would give Joveljíc — behind the defense, 6 yards out, no defenders around him, just needing to put the ball either side of the keeper — nearly a 100% chance of scoring.

That’s not what xG “sees,” though. It sees two strikers, both attempting a shot with their right foot from the center of the penalty area, 6 yards from goal — among other factors, of course. So if you look at the numbers from Sportec, the league’s data provider, you see that Joveljíc was assigned 0.55 xG and Oluwaseyi 0.58 xG.

Moreover, on FBRef.com — which has advanced data from Opta, the dominant numbers provider in world soccer, which analyzes the game in even more detail — Joveljíc’s shot is given 0.57 xG, but Oluwaseyi’s only 0.37 xG.

So not only do these numbers not tell the whole story, they vary from place to place.

In general, spikes like these tend to even themselves out only over the long run, which is part of the reason most analysts get annoyed when anyone tries to use xG to prove anything about a single game, or even a short run of games.

Loons coach Eric Ramsay is aware of this. “I’m very conscious of the fact that no one gets the xG trophy come the end of the year,” he said. “I think over the course of a season, those types of things tend to point to whether you’re performing well or not performing well, and generally if you’re performing well, you’re going to be where you want to be.”

This is not to say that Ramsay and his staff don’t use statistics, but that xG is used as “shorthand” rather than a definitive number.

“We really do pick apart performances from a statistical perspective,” he said. “I think it’s really important that if you’ve got a very clear way of playing, you should also have a very clear way of analyzing the effectiveness of that.”

Those numbers aren’t the traditional numbers, not even xG, but what Ramsay calls “bespoke” numbers. They measure the things that he focuses on with his team, not just chance creation. For example, he mentioned metrics that cover how compact the team is on defense, how well they press in certain areas of the field, and the effectiveness — not the quantity — of the team’s attacks.

“The important thing is that the players feel that coherence and consistency day to day and week to week — so I’m not getting drawn in too much to the more traditional ways of evaluating performance, I suppose,” he said.

In other words, the xG numbers might tell you something, but not everything, about Saturday. They will tell you that there’s evidence that the Sounders aren’t as bad as their record. But just like how errors don’t tell you everything about how good a fielder a baseball player is, no matter what xG tells you, there’s a lot more going on.