PHOTO: ISI Photos
Introduction
This past weekend, @NBCSQuakes reminded us the 2018 Earthquakes have three players in or within striking distance of double-digit goals with seven games to play.
Who will finish as the @SJEarthquakes top scorer this season? pic.twitter.com/2lsKeK1htG
— Earthquakes on NBCS (@NBCSQuakes) September 9, 2018
At 41 goals, San Jose have equaled their highest goal total since 2012 (41 goals in 2015), where they scored 71 goals. This is not due to more shots — as currently they are behind every other season by a good margin — so it must be due to higher shot quality and better opportunities. As you can see below, on a pure goals-scored basis, the Quakes are 8th out of 13, with Portland (38), Minnesota (37), Seattle (32) and Colorado (29) behind them. On an Expected Goals (xG) basis, the Earthquakes would also be 8th in the Western Conference.
Screenshot from American Soccer Analysis (@AnalysisEvolved) Team Expected Goals tables: https://www.americansocceranalysis.com/asa-xgoals/
If San Jose had a middling defense, the team would at least be challenging for a playoff spot. Today we’ll examine advanced data metrics, which should help us understand why the likely Wooden Spoon winners are surprisingly more offensively efficient than in the last six seasons.
As we covered earlier in the season, xG tells us the number of goals a team is likely to score given the number and quality of their chances. The higher the xG of the shot (effectively the percent chance, based on historical data, that a goal will be scored) the better the opportunity, generally speaking. Probabilities tell us two 25% shots have a greater likelihood one goal will be scored, while five 10% shots have a better likelihood more than one goal will be scored but also no goals will be scored (check it out for yourself). In non-statistical language, it is the difference between Vako taking two shots in the box vs. taking five shots outside the box. On a particular night he might get lucky and score a couple outside the box but not repeatedly. Expected Assists (xA) has the xG value of the shot given to the player who made the pass (called a key pass) to the shooter, if any. Given Danny Hoesen, Vako and Chris Wondolowski are scoring most of the goals for the Quakes, we can expect them to have the higher xG values on the team, and given Magnus Eriksson has the most key passes, we can expect him to have the highest xA value on the team. Let’s take a look at shots and key passes as well as xG and xA since 2015 for San Jose.
The first visualization shows shots per 96 minutes against key passes per 96 minutes, while the second shows Expected Goals (xG) per 96 minutes against Expected Assists (xA) per 96 minutes. The simplest way to think about the differences between these two visualizations is volume (shots and key passes) versus quality (xG and xA). For example, while Vako in 2018 has the highest volume of shots, Wondolowski in 2015/2016 has the highest quality shots. The reason we are not comparing goals against assists is that in statistical terms both are subject to a higher amount of random factors, particularly assists.
What is clear here is the best Quakes player at serving the final ball was Matias Perez Garcia in 2015. One reason for this is that he took the corners and free kicks. If we took out corners and free kicks, we may come to a different conclusion. In 2018, we see the goal scorers are clustered together for shots and xG with really only Eriksson standing out for key passes and xA. The reason we look at xA in addition to key passes can be seen with Tommy Thompson in 2017: he’s somewhere in the pack in key passes, but we see him stand out in xA, showing he had a better than average final ball compared to his peers. We also see Eriksson getting closer to Hyka 2017 on the xG vs. xA visualization, likely showing his average xA per key pass is higher in 2018 than Hyka’s was in 2017.
But this data in this format doesn’t help us identify what’s changed this season to create more goals. To get a closer look where the creation of goals are coming from, we turn to the Expected Goal Chains (xGC) metric and its relative the Expected Buildup (xB) metric. xGC awards the xG of the shot to all the players who participated in the buildup of the shot; i.e., made a successful pass. xB awards the shot xG to everyone except for the shooter (who can be measured by xG) and the key passer (who can be measured by xA). For the most part, xG + xA + xB = xGC.
Recently American Soccer Analysis introduced xGC and xB on its interactive tables. We can use this data combined with xG and xA data to look at which players are contributing to the buildup of shots.
Looking at xG+xA against xB we can see which players are participating in the shooting phase vs. strictly in the buildup phase. There are no surprises here in terms of where names appear on the graph, but what is clear is almost every player is higher in xB in 2017 than in 2018, with Wondo and Jungwirth as exceptions. This gives us a clear indication that there is less buildup in 2018 than in 2017. It is also an indicator that the buildup is failing in 2018 or is just more direct in style. It seems looking at individual players may not lead us to any conclusions, so let’s look at some positional data.
As you would expect, most of the buildup comes from the midfielders, this includes the central midfielders and outside midfielders (those not considered an “attacking midfielder”). In this view, Vako is considered a regular midfielder as the data shows there haven’t been any consistent attacking midfielders for San Jose since 2015 (Perez Garcia). The data in the visualization above is summed for all players but is shown on a per 96 minutes basis, given that 2018 still has several games to be played. It is clear there has been a massive drop-off in the participation of the outside backs in the buildup that has resulted in shots, despite having two offensive-minded outside backs. One could easily use this data to argue Stahre is not as efficient with his use of the outside backs as Kinnear and Leitch were. Regardless, these are lower numbers across the board. One reason is that the Quakes are playing a more direct style under Stahre, which would generate less xGC and xB.
In a general sense, there is not any evidence that the buildup has improved in 2018; in fact, it may have taken a step backwards. This is clearly not the reason for the improved offensive output. The question is if there are any metrics where efficiency has improved, even if these top-line offensive metrics have not.
The above metrics seem to reveal the answer (note the red rectangles). In the available data since 2015, the Quakes have mostly under-performed (negative values on the visualization) in Goals minus Expected Goals (G-xG) per shot and Assists minus Expected Assists (A-xA) per pass in key positions. In every year starting in 2015, the Quakes have under-performed goals per shot relative to xG in every position, until the forward position this year. This year the forward position has performed right at xG, scoring the goals they should have in the end. Also, the midfielder position has not faltered as badly shooting as in previous years, likely due to goals from Vako, Eriksson and Hyka, and the same can be said for the few shots taken by the outside backs (Nick Lima’s goal against Vancouver is the lone highlight). But that’s not all, the quality of the shot is higher, thanks to better passing in the final third. According to American Soccer Analysis xPassing, San Jose is over-performing Expected Passing (xPass) in the final third by 2% and is making a better quality final pass, leading to a better shot. The forward and midfielder positions are over-performing against xA together for the first time since 2015, when this data first started to be tracked for MLS.
Conclusion
There are concerns here. While the Earthquakes are reaping some positives of keeping a core set of players together like Hoesen, Wondo, Vako and Hyka for another season making for better scoring chances, there remain significant issues in the buildup phase from the central midfield and regression at the outside back and center back positions when looking at the totality of the season. There is little evidence from this data that the pairing of Anibal Godoy with Luis Felipe Fernandez is paying off in the buildup phase, and, according to the Expected Buildup (xB) metric, it seems replacing either Godoy or Felipe with Jackson Yueill (or finding another way to get Yueill on the pitch) would produce more linkup which would lead to shots. Such a move, however, could further put the defense into disarray. Coming into 2018, the Quakes would likely benefit from the signing of a true #10, who could carry such a responsibility that Vako does not appear suited for the distribution duties of such a role. By doing so, however, they must either trust Luis Felipe with the sole #6 duties or find someone who can take them. In 2019, regression to the offensive mean could spell disaster for a club already skirting with MLS irrelevance.