When Expected Goals Don't Matter
When Expected Goals Don't Matter
New York City FC recorded a season high xG in an embarrassing 5-1 loss against the Philadelphia Union, but those numbers don’t mean what you think they do.
Last Wednesday, the air was somber as manager Nick Cushing dejectedly entered the press conference room in the crypt of Yankee Stadium. As he sat down, he fidgeted with the energy drink product placement on the table, adjusted his microphone, and prepared himself for the unpleasant task of addressing the media after a 5-1 loss to the Philadelphia Union, one of the worst defeats in New York City Football Club history.
Any home loss to division rivals like the Union would sting, but the magnitude of this defeat, combined with the major playoff implications that this late-season fixture had, made the loss especially excruciating. The Union were on the outside of the playoffs looking in, while New York City, winless in MLS regular-season play since July 3rd, was trying to halt its slide down the table.
The game had been a nightmare for the Pigeons on both sides of the ball. Defensively, a series of puzzling individual errors in the first half allowed Philly to go up three goals within a 16-minute span.
In the 15th minute, Tayvon Gray became too enamored with Jakob Glesnes during a corner kick, failed to push forward along with the rest of the defensive line, and kept Tai Baribo onside for the go-ahead goal.
Moments later, an inopportune first touch from Thiago Martins allowed the ball to reach Mikael Uhre via Baribo for the second. Minutes later, the normally reliable Matt Freese, who had been ranked in the upper third of starting goalkeepers in G+ handling, fumbled a ball on the goal line to a raring Dániel Gazdag for the third goal of the game.
The offensive play was equally futile, but not for a lack of trying. Alonso Martinez snuck behind the Philadelphia backline in the final seconds of the first half to close the gap to two. Cushing subbed on Maxi Moralez at halftime and sent just about everyone up the pitch in an ultimately fruitless attempt to change the game.
The numbers sent forward created chances—really, really good chances—but they did not result in goals. Keaton Parks conjured his inner Gordon Bombay, gliding his gangly frame into the box for a couple of 1v1 chances that he couldn't put past Union boy-lkeeper Andrew Rick.
Parks was not the only Pigeon stifled by a standout performance from Rick. Mounsef Bakrar made a great play on a long ball to chip Rick. The ball appeared destined for the net, only for Rick to get a fingertip on it and deflect it just wide off the post. Even substituting on a second striker, Jovan Mijatović, proved to be an insufficient tactical wrinkle, as Rick finished the game saving nearly an entire goal more than expected.
When all was said and done, according to Opta, New York City finished with 26 shots, nine on target, five big chances, but only Martinez’s goal to show for it.
While the inefficacious offensive output did not affect the outcome of the game, it had the curious effect of seeing New York City win the xG battle by nearly a goal and a half—3.1 to 1.7—despite a four-goal defeat. The 3.1 xG was New York City's highest this season, something that Cushing highlighted in his response to The Outfield’s Chris Campbell during the press conference:
Very, very strange goals tonight in the first three. Disappointing set play on the fourth. But, I think the positive part, that we also have to look at is we create some huge chances tonight. If you look at the expected goals, I'm sure it'll be about three, three point something, which is something that we have to continue to do. You tighten up on one end and you tighten up on the other end. We get back to winning ways really, really quickly.
No other statistic has done more to revolutionize the sport of soccer than Expected Goals (xG). Time and time again, xG has been found to be a better predictor of future results than nearly every statistic. It’s superior to shots on target, total shots, goals scored, and points per game in predicting future points per game at virtually any point in a season.
So, even in defeat, seeing the Pigeons accumulate a season-high xG must be a positive sign and a good omen for what lies ahead…right? Not quite.
To get the best predictive use of xG1, American Soccer Analysis’ Eliot McKinley says you need at least seven to ten games. Single-game xG, as Cushing used it in the press conference, can serve as a descriptive tool for discussing chance creation. However, an important factor to consider when examining New York City’s extremely high xG in this game is the gamestate.
The gamestate matters because most of New York City’s xG came after they were down three goals. Earlier this year, in his blog post “Expected Goals, Gamestate, and Predictiveness,” analyst Tony ElHabr concluded that in the Big Five European leagues, “xG accumulated in lopsided games (either by the trailing or leading team) is not all that meaningful for projecting how either team will perform in the future.”
That analytical conclusion makes sense, considering the type of play typically seen in lopsided games. Sometimes, the trailing team simply gives up. In other cases, when the trailing team fights, as New York City did against Philadelphia, you see teams hysterically push numbers forward to attempt to get back in the game.
Since Philadelphia’s lead came so early, they had the luxury of sitting back on defense and waiting for the game to conclude. Defense didn’t really matter because Taty Castellanos could have descended from the heavens, scored a hat trick, and the Union still would have won the game by a goal.
So what Cushing said was probably half-true. The goals Philadelphia scored were indeed weird, but the expected goals accumulated by New York City in a blowout are largely meaningless for analyzing their offensive performance.
Now compare that to the expected goal numbers for the Miami game, which finished in a 1-1 tie on Saturday. The numbers, in part due to the much closer gamestate, make a lot more sense with what was seen on the field.
New York City struggled to create open-play chances against Miami and was perhaps a little lucky to score the Sands header, finishing with 1.1 xG- the lowest single-game total at home all season.
On the other hand, the Pigeons' defense was excellent with the singular exception of Leo Campana’s goal. The backline was able to frustrate Luis Suarez, who didn't get a shot on target, while Sands and the midfield were able to corral Lionel Messi, who finished with only a single touch in the box, out of dangerous areas. At 0.8 xG, for only the eighth time all season, Miami’s vaunted offense finished the match under 1 xG—and 0.6 of that came from the Campana goal. For New York City, it was perhaps a little unlucky that Miami’s only serious chance resulted in a goal.
The comparison between the Sands goal and the 2021 Gudi Thórarinsson free-kick goal that helped spark the 2021 MLS Cup run seems a little far-fetched. Season xGD points to NYCFC as the fifth-best team in the Eastern Conference, whereas the 2021 team had been underachieving with a historically great xGD total.
So where does that leave New York City FC with four games left in the regular season? With many questions that remain unanswered. ❧
Image: Akseli Gallen-Kallela, Lemminkäinen's Mother
Worth noting that using xG is less reliable to predict future results in MLS than in the Big 5 European leagues. MLS gonna MLS.