Stats aren’t everything. But modern soccer stats do give us some interesting insights into the overall quality of team play, and we can speak to style through more traditional stats. Today will be about the team, with individuals to follow.
For me, the most interesting thing about taking a deep dive into the San Jose Earthquakes’ 2019 season, with Matías Almeyda’s fascinating scheme, was that the distinct “acts” of the grand play were perhaps a bit different than fans might expect. My read is that they came in four phases, which I’ll analyze in sequence. But the most important takeaway is that something unusual happened in the last 12 games, attributable either to luck or tactics, and the answer to that question will define how optimistic Quakes fans should be going into the 2020 season.
Act I (March 3-May 11): Here we go again
Goals For | Goals Against | Differential | |
Expected | 1.26 | 1.93 | -0.67 |
Actual | 1.27 | 1.91 | -0.64 |
Results: 3 wins, 2 draws, 6 losses (11 points)
Things started out badly, and not just in the first four games. Fans understandably focus on the NYRB and LAFC matches, but the truth is the Almeyda system took at least another half dozen games to fully settle in.
The underlying stats didn’t begin to improve until after this 11 game set, and even 2 of the 3 wins the Quakes managed in this period were rated as “lucky” by the expected goals model. The offense was producing similarly to Dom Kinnear’s teams (not a compliment), while the defense was in full apocalypse mode, giving up a staggering 1.93 xG per game.
In general, though, the team was “getting what they deserve” according to the advanced stats, with expected and actuals almost identical. If that xGD and points accumulation continued over the entire season, this team would have ended up a strong contender for a second straight Wooden Spoon.
Act II (May 18-July 27): A stroke of genius
Goals For | Goals Against | Differential | |
Expected | 1.89 | 1.07 | +0.82 |
Actual | 2.08 | 0.92 | +1.17 |
Results: 8 wins, 2 draws, 1 loss (26 points)
May 18th, 2019, home against Chicago. That’s the date we can look back on when Almeyda first managed to fully express himself on the pitch, and gave the first indications in San Jose of the genius that made him so revered in Mexico. The hosts blasted the visitors 4-1 on the day, and kicked off a run that was genuinely championship-level.
The performances were strong in every dimension. Both the expected goals for and against would have rated 2nd best in MLS (behind LAFC) if held up over the season. That meant a goal differential that would’ve left every other team in the league in its dust. All of that of course matched up to the eye test, which indicated a team that was back in love with the game and suffocating its opponents with its unorthodox system. They outperformed the underlying numbers by a bit, but not by so much you would be worried about its sustainability.
It was not ridiculous to suggest at this point that the team could create an upset in MLS cup and send the squad home with some silverware.
Act III (August): Luck Runs Dry
Goals For | Goals Against | Differential | |
Expected | 2.36 | 1.20 | +1.17 |
Actual | 1.60 | 1.80 | -0.20 |
Results: 2 wins, 1 draw, 3 losses (7 points)
Here’s where things get weird. The performance levels, at least according to the advanced stats, stay incredibly strong through this period. In fact, they’re actually a bit better than the purple patch that preceded it.
What happened was that the very stable relationship between xG and actual goals that held up for the first 22 games in the season (and generally holds up across all of world soccer) suddenly evaporated. There was a staggering difference of 1.37 between expected and actual goal differential. Essentially, the team became astoundingly unlucky, all of a sudden. Or at least that’s how it appears (more on that later).
The Quakes did manage to pound two terrible teams (Orlando and Vancouver) but the damage was done in 3 frustrating matches against Columbus, Colorado, and Kansas City, where they outplayed each one, but came away with just one point. The 4-0 loss to LAFC was not nearly as bad as it appeared on the surface (just -0.62), without even taking the missed penalty review into account. No shame in losing to the best team in MLS history on the road.
A brief poor run of luck, but such things happen. There was nothing obvious going on through this period that should’ve indicated a team that was about to crash out of the playoffs.
Act IV (September 12-October 6th): The Bottom Falls Out
Goals For | Goals Against | Differential | |
Expected | 1.39 | 1.21 | +0.18 |
Actual | 0.50 | 2.00 | -1.50 |
Results: 0 wins, 0 draws, 6 losses (0 points)
Somehow, the Quakes luck got worse, with the gap between expected and actual goals going even farther out of their favor. But if that was all that happened, they still probably would’ve been a playoff team, collecting a handful of points, and ending up comfortably above the playoff red line.
What happened instead is that for the first time since the beginning of the season, the performances also fell off dramatically. Part of that no doubt was a product of an unusually difficult run-in against the three best Eastern teams, two of the three best Western teams, and another playoff team, with 4 of those 6 opponents on the road. But still, the numbers went from “nearly LAFC” to “middle of the pack” in this period. The once-explosive offense became pedestrian again. This drop-off, combined with the continued terrible luck, was fatal to the Quakes’ playoff chances.
But let’s talk about luck for a second. If xG was a perfect metric, “luck” is all it would be, and we could rest comfortably with the knowledge that it’s unlikely Almeyda will be quite so unlucky in 2020, and we will likely have a playoff team on our hands.
But it’s not a perfect metric. One of the lingering issues with it is that it doesn’t do a particularly great job of accounting for game states (i.e. what the score is) and how coaches like to set up, tactically, in each one. That’s because xG is based primarily on location of shot and service to set it up, and cannot account for every other variable that might affect its likelihood to go in, including tactical setup. As one small example, a player is more likely to take a shot from a particular location that is less favorable (given the millions of variables that xG can’t account for) when he’s chasing a game rather than protecting a lead. And he’s likely to face a defense that’s sitting much deeper and is better organized. Both of those things would lead to xG underperformance. That could be the reason, rather than luck, that the expectations and reality diverged so strongly in the last third of the season.
It’s no secret that I was highly critical of Almeyda down the stretch for his game management. I felt that extremely aggressive, offensively-oriented substitutions frequently gave away points that could’ve been earned through draws. And the manner in which this collapse occurred (dropped 11 points from winning positions, blew 5 second half draws) indicates that the team was not prepared to grind out results late on. That could be a tactical issue, such as an inability to switch approach and solidify a defense, or a preparation issue, such as exhaustion that sets in after a long season of a high-intensity system with limited rotation. It could even be that teams were beginning to catch on to the unorthodox system, which can lead to unusually high-danger chances on the defensive end once broken. Regardless, these causes of xG underperformance would point to coaching failures rather than luck, and be warning signs going into 2020. Simply put, the team did not manage its game states well, either protecting or chasing a game.
I’m not definitively saying it’s luck or tactics. Likelihood is that, like most things, it’s a mixture of the two. My instinct is that the luck will even out a bit next year, indicating a likely playoff team, but with serious limitations if he doesn’t learn how to manage games late on, especially if, as expected, the squad does not change dramatically.
Random team-level style notes
- Quakes were tops of the league in possession (55.8%) and second in passing percentage (84.3%) while being dead last in aerial duels won (10.6). None of those things are intrinsically bad or intrinsically good: it just indicates a playing style with high pressing and a build-from-the-back, keep-it-on-the-ground-and-retain playing style. It is, however, night and day from recent Quakes regimes.
- The difference in set pieces under Almeyda was notable, with delicately choreographed routines becoming the norm from almost every dead ball situation. 10 set play goals, good for 5th in the league, would indicate that the time spent on the training ground was well spent, after very poor returns in that department in recent years.
- SJ was definitely a bit rough and tumble in terms of fouls (17.7/g) and cards (70 yellows and 5 reds) but they weren’t actually at the very front of the pack on either, rating 4th and 5th, respectively.
- They only bagged one penalty all year, by far last in the league. No doubt, the bizarrely un-reviewed LAFC incident and the blown review against the Union would have improved that. But penalties tend to vary significantly year on year, so I wouldn’t be surprised by a bit of reversion to the mean in 2020.