Sports analytics don’t tell the whole story, particularly in soccer, a game that is notoriously hard to quantify, with 11 players on each team, low sample size of successes (goals), and about a million confounding variables.
I do, however, like to dive into the data, particularly advanced metrics of the sort maintained by the excellent American Soccer Analysis website, to evaluate the narratives that are out there about the Quakes and see if the data suggests we’re either on to something, or we’re wrong. I’ve based many of my observations here on the “expected goals” concept, which ASA explains here, but essentially attempts to quantify how much a team “should” be scoring given the overall quality of their play, independent of luck.
Here are a few narratives I took out for a spin, with the caveat, of course, that we’re just 9 games into the season:
“Quakes are a disaster this year”
They aren’t necessarily all that good, but they’re definitely not a tire fire. The overall expected goal difference per game (xGD/g) is -0.29, which is not great, but is the 7th best mark in the West overall. That implies San Jose is just one spot out of a playoff place, and the 6th best team (Seattle) isn’t far ahead at -0.17. As such, the underlying metrics suggest a team that could easily be playoff-quality if it took a modest step forward, either through player acquisitions, tactical changes, or improvement as a group.
One thing I’d note, however, is that the bulk of the “positive” xGD has come from the two games against Minnesota United, who are 19th out of 23 MLS teams in xGD. Take those two games out, and the number falls all the way to -0.60 xGD/g, which is…bad. That would put the Quakes much closer to DC United and RSL than the playoff redline.
“Well at least it looks like they’re getting better”
Answering this one depends on how you slice the data. From my read, there are three distinct phases:
- The first three games, which included quality opponents SKC and NYCFC, where the Quakes averaged +0.61xGD/g
- The next four games, where the Quakes were simply dreadful, registering an apocalyptic -1.44xGD/g
- The most recent two games, against mediocre opposition, where the Quakes turned it around to the tune of a +0.69xGD/g
If they’re “improving,” that improvement certainly isn’t linear, since the best phase of the season was followed by an alarming regression. However, it’s reasonable to say that they’ve pulled out of the tailspin in the last two games. Now the trick will be proving it against a quality opponent, and keeping it up over time.
“Mikael Stahre Is Not the Answer”
This #narrative is one I have been pushing harder than just about anyone online, so it’s the one I should scrutinize the most closely. It was more powerful prior to the Minnesota game, when the numbers were twice as bad, but things still don’t look all that great. Here are the Quakes three most recent coaches, just accounting for their performances from the beginning of 2017 on, to keep the comparisons as close as possible.
| Coach | Expected Goal Difference per game (xGD/g) | Expected Goals Scored per game (xGF/g) | Points Per Game |
| Kinnear | +0.15 | 1.28 | 1.35 |
| Leitch | -0.18 | 1.45 | 1.35 |
| Stahre | -0.29 | 1.37 | 0.89 |
I, for one, am concerned about this. While the goalscoring has marginally improved under Stahre when compared to Kinnear, the expected goal differential has gone from +0.15 (better than Atlanta Utd in 2017) to -0.29 (significantly negative and in the bottom third of the league). Points-per-game is not an “advanced” metric in that it has not been adjusted for predictive quality, but may well speak to a coach’s ability to play the game situation effectively to get points. In that category, we have moved from the EPL equivalent of Everton under Kinnear and Leitch to the EPL equivalent of Stoke.
The reason I find this downward trend particularly disturbing is that the underlying player quality has gone in the opposite direction: Stahre has the highest-paid player in club history (Vako) and second most expensive purchase (Eriksson) to work with, whereas Leitch only had Vako and Kinnear had neither. And it’s not like San Jose lost any real contributors in that period. To me, that points to coaching as the likeliest cause of decline, and I’m concerned that the trend will not turn around.
“This is the year Wondo finally lost a step”
This one has an element of truth to it, but is wrong on the specifics. Wondo is actually at a level almost identical to last year, with 0.34 xG per game in 2017, and 0.32 xG/g in 2018. The drop-off in performance, it appears, actually occurred prior to 2017, since in the 2014-2016 seasons Wondo averaged 0.50, 0.54, and 0.51 xG/g respectively.
Note: Throughout this piece, if I say “per game” for a player, what I really mean is “per 96 minutes of playing time,” since that is the average game length in MLS, and allows us to account for substitutions.
One thing to note is that since expected goals (xG) are based on location and mode of setup, they reflect the chances a player gets, and don’t speak to finishing. However, those chances are based both on service (getting the ball to the right places) and the attacker attempting to score (who has to, of course, make the right runs to get into the right places). So it’s not 100% clear that it’s entirely Wondo’s fault, but it clearly signals a problem.
“Anibal Godoy is mailing it in to save himself for the World Cup”
I’m guilty of this one. In the past, Godoy has seemed imperious, and when he’s on form, he can completely run a game. We haven’t seen that this year, which has been a big disappointment. So I thought I would compare the three primary central midfielders this year, with all stats on a per-game basis:
| Player | Florian Jungwirth | Aníbal Godoy | Jackson Yueill |
| Tackles | 3.7 | 3.0 | 2.0 |
| Interceptions | 3.3 | 2.9 | 0.3 |
| Fouls | 1.0 | 2.9 | 0.5 |
| Dribbled Past | 1.5 | 2.2 | 2.6 |
| Shots | 1.0 | 0.7 | 0.5 |
| Key Passes | 1.1 | 0.3 | 1.6 |
| Dribbles | 0.1 | 1.9 | 0.5 |
| Fouled | 2.0 | 3.1 | 0.8 |
| Dispossessed | 0.7 | 1.1 | 1.2 |
| Unsuccessful Touches | 0.4 | 1.3 | 1.1 |
| Passing % | 84.6% | 89.5% | 84.0% |
| Difference b/w Pass% and xPass%* | +4.3% | +4.6% | +0.4% |
| Avg yards advanced per pass in the middle 3rd | 4.21 | 3.75 | 4.02 |
| xGChain** | 0.47 | 0.3 | 0.36 |
*Measures how much more accurate of a passer the player is than the model expects given the difficulty of passes attempted
**adds up any actions that led to an expected goal, from deep in the buildup (such as a tackle leading to a counter-attack) all the way up to the goal attempt itself.
First off, Flo is very clearly the best option as a defensive midfielder, since he has by far the best defensive numbers, takes the best care of the ball, and despite his defensive reputation, appears to actually be the most involved in chance creation and buildup. Godoy does stand out for his ability to beat men on the dribble and win fouls, but he’s a step behind Flo everywhere else. Yueill is behind the established starters in most categories, but has promising “key passes” numbers and xGChain (buildup) metrics, signalling potential as a deep-lying playmaker if he can round out his game in other areas.
The conclusion, then, is that Godoy has indeed been off his game, not contributing nearly as much as he typically does in creating goals, and being a defensive liability when it comes to fouling or being blown past on the dribble. But he’s still the second best central midfielder on the team, and in many categories isn’t far off of Flo.
“Shea Salinas shouldn’t start in MLS. Nick Lima, however, is the Chosen One”
Another one I 100% have propagated in recent months. In fact, I took it for granted so much that I almost didn’t put it to the statistical test. Boy was I surprised. Here are the two primary fullbacks for the Quakes compared on a per-game basis:
| Player | Shea Salinas | Nick Lima |
| Tackles | 3.0 | 2.6 |
| Interceptions | 0.5 | 1.9 |
| Fouls Committed | 1.2 | 0.4 |
| Clearances | 2.0 | 3.4 |
| Dribbled Past | 2.5 | 0.4 |
| Dribbles | 3.0 | 1.4 |
| Fouled | 0.5 | 1.3 |
| Dispossessed | 0.9 | 1.3 |
| Unsuccessful Touches | 0.4 | 1.5 |
| Crosses | 0.5 | 0.7 |
| xGChain (buildup rating) | 0.2 | 0.24 |
| Key Passes | 0.5 | 0.66 |
There’s no doubt you would prefer Lima to Salinas as a defender with those numbers (he has clear advantages in Interceptions, Fouls Committed, and Dribbled Past), but the difference is perhaps a bit less stark than I had imagined. Where I was really surprised was on the other side of the ball: Salinas is not notably behind Lima in any meaningful offensive/buildup category, and in fact comes out well ahead in dribbles and turnovers.
Not saying I’ve changed my mind that LB could be upgraded, just some food for thought.
“[Insert CB] is trash”
I’ve heard different people pick different punching bags at center-back, and seem to blame them for the lion’s share of problems at the back. Jungwirth has only started two games at CB, so I did not include him in this analysis, but I thought it would be interesting to look at the numbers to see if one crew of CB-haters had bragging rights over the others:
| Player | Yefferson Quintana | Harold Cummings | Francois Affolter |
| Tackles | 0.9 | 2.3 | 3.3 |
| Interceptions | 2.8 | 2.8 | 3.0 |
| Fouls | 1.8 | 0.8 | 1.9 |
| Clearances | 5.8 | 4.8 | 2.5 |
| Dribbled Past | 2.2 | 1.0 | 1.4 |
| Blocks | 1.4 | 0.8 | 0.8 |
| Long Balls | 3.9 | 2.8 | 2.2 |
| Passes | 47.3 | 40.1 | 47.3 |
| Passing % | 74.6% | 82.9% | 84.5% |
| Difference w/ Expected passing % | -1.1% | +0.9% | +0.4% |
| Avg yards advanced per pass | 7.13 | 5.07 | 5.5 |
Conclusion? Unclear. I think Cummings probably has the best overall picture of defensive actions, but he’s the most limited passer in the group. Quintana has a weirdly low figure for tackles, and gets dribbled far too much, but otherwise looks decent as a defender. As a passer, he clearly plays differently than the other two, with a more vertical/direct approach, and it’s hard to tell to what extent that’s tactical/circumstantial and to what extent it speaks to a limited passer. Affolter unsurprisingly appears to be the most comfortable passer in the group, playing it out of the back at the highest rate, but actually grades out as an extremely active defender, which surprised me a tad.
Essentially, I think I would probably have sympathy for the crowd that thinks Quintana is the weak point. But none of the three are producing at high levels, and it’s a position where the play could be seriously improved, either through tactics, shifting Flo back permanently, or spending yet more money on another center-back.
