clock menu more-arrow no yes mobile

Filed under:

Goals will come? The story of expected goals in Crew SC's loss to Houston

Expected goals may be a new stat in soccer, but there is certainly information to be gleaned from how shots are turned into goals.

Thomas B. Shea-USA TODAY Sports

The preseason predictions have been positive, and league watchers a picking Crew SC as one of the best team's in the Eastern Conference, but Saturday was a 1-0 loss after the team should have scored several times. Columbus fired off 18 shots, Federico Higuain hit the post, Kei Kamara's headers forced excellent saves, and Emmanuel Pogatetz even got a few chances for the visitors to get on the board. Tyler Deric earned team of the week honors with his performance. "Lucky" Houston only had 8 shots, but made them count.

That's not the entire truth. Crew SC looked more dominant than the Dynamo, but that speaks more to volume than quality of shots. Here are the 18 shots Columbus took on Saturday.

ColumbusvHouston Shot Chart

Many are from distance, are from a wide angle, or headed. The math (which we'll get to in a few minutes) shows these are much lower percentage shots. Rather than scoring one in five attempts, these shots might go in every 12 attempts.

Compare that to Houston's shots:

HoustonvColumbus Shot Chart
Again, many of these are lower percentage shots except one, Giles Barnes winner.

Now for (some of) the math. Traditionally, statistics in soccer have been limited to goals, and more recently assists. Within the last decade, there has been more data scraped and analysts are now starting to develop models to better rate and rank events on the soccer field. One of the more interesting developments is around expected goals (xG). It's always been true that some shots are better than others, but now that's starting to get quantified.

Michael Caley of Cartilage Free Captain, SBNation Spurs blog, and the Washington Post has refined a model to rate each type of shot. The most effective shot is the penalty kick, it's uncontested, but even that is only finished at a 78% clip (a .78 xG). A shot from inside the 18 yard box but from a tight angle drops off precipitously, perhaps as low as .05 xG. Haley's three keys to expected goals are: where the shot was taken, shot type and the pass that lead to the shot, and what kind of move it came from.

A shot taken from a wide angle after a slow buildup and off of a lobbed pass will be less likely to result in a goal compared to a shot taken from about the same angle after catching the defense on the counterattack where the player dribbled into position. The result of this analysis looks a little like this (courtesy of Michael Caley):

xG map vHouston 3

According to Mr. Caley's xG modeling, Crew SC's shots should have resulted in at least one goal, but few of the attempts were exceptional. Houston had far fewer shots, but two of them were very good attempts. Barnes goal came after broken defense where he took a cutback pass on the ground and shot with his foot. It was shot with the highest xG shot of either team.

It's far to early in the season to draw too many conclusions and I'll have more on expected goals as the season continues. I'll certainly be watching how quickly Crew SC moves into attack and where they eventually turn that possession into shots. Plenty of Crew SC's shots "weren't troubling the keeper" and the stats back that up.