On May 20th, the New England Revolution fell apart in Philadelphia. They managed just 8 shots to Philly’s 17 and lost several players to injury during the match, eventually losing 3-0 to the Union.
It was an abysmal performance.  The kind of performance that can break a team’s spirit.

Photo: —

Fortunately, that’s not what happened with the Revs.  In the 10 matches since their road loss at Philadelphia, the Revolution have gone 5-1-4 (W-L-T) in league play.  That’s a blistering 1.90 ppg pace.  In that span, they’ve outscored opponents 23-14.

What’s strange about these past 1o games is that despite the scoreline dominance, the Revs’ underlying numbers have lagged.  The vaunted ‘expected’ metrics don’t seem to paint quite as rosy a picture of the Revolution. Per Opta, opponents have scored 14 goals against New England (not far off from the predicted 13.8). New England, however, is outperforming their own 11.8 xG by nearly double.


I’ve often said that advanced metrics are great at showing us ‘what’ is happening but not always ‘why’ it’s happening.

With that in mind, let’s delve into the ‘what’ and ‘why’ of New England’s advanced metrics and see if we can make sense of this overperformance.


Wait, What’s Going On?

So, I’m back on the expected goals thing again.

Wait! Hear me out.

There’s been a lot of talk about the merits of advanced or ‘expected’ metrics used to analyze soccer.  Proponents will argue that they are a statistically reliable way to analyze team performance, whereas critics take issue with it being over complicated and ignoring the most important stat (the scoreline).

I fall somewhere in the middle.  We can tend to get hung up on what individual data points say.  For example, ‘Team X lost the game but won the xG battle and therefore were the better team!’  Expected goals are calculated using massive datasets, and therefore in small sample-size cases, there are going to be some disparities.

We can, however, use those individual data points (game by game) to establish a trend and notice any deviations from that trend.


That brings me to the point of this article.

Photo: —

This is a match-by-match chart of New England’s expected goal differential (xGD).


That’s to say, the number of goals they were expected to score minus the number of goals they were expected to concede. Positive numbers should, in theory, indicate good performances, and negative numbers should indicate poor performances.

I’ve circled the aforementioned 10-game stretch since losing 3-0 at Philadelphia. I’ve done this because it represents something of a puzzling shift for New England.

In the 13 games up to and including the loss at Philadelphia, New England went 7-3-3 (24 points, 1.85 ppg).  In the following 10, New England went 5-1-4 (19 points, 1.90ppg).  Seemingly, as far as the table is concerned, it has been smooth sailing for New England all season.

Their xG differential, however, tells a much different tale.  Before May 21st, the Revs outperformed their opponent’s xG 8 times out of 13 (61.5%).  Since then, they’ve managed the same feat on just 4/10 attempts.  Furthermore, the margin of positive xGD diminished significantly.  In games where the Revs ‘win the xG battle,’ the margin of victory is cut in half, from 0.99xG to 0.45xG after the Philadelphia game.


OK, So WHY is That Happening?

There are many factors that could have contributed to this shift in xGD, but I’ll hone in on the 3 most likely explanations.  Keep in mind the Revs are doing well right now.

1: There has been a dip in defensive performance, leading to a higher number of expected goals against.

2: There has been a decrease in the quality/quantity of chances leading to lower expected goals for.

3: Game states are interfering with the data.  For example: Are the Revs racing out to a big lead on a few chances and spending the remainder of the game absorbing pressure and conceding shots?


Defense?

The first option to explain the dip in xGD in recent weeks could be an increase in the number or quality of chances that the Revolution defense is conceding to their opponents.  If the offense continues at their normal clip, but the defense allows more chances, that could lead to the Revs losing the xG battle more often.

I want to make it clear just because you concede more chances, that does NOT mean you concede more goals.  Not necessarily, anyway.

The Revs are conceding about 1.2 goals per match overall this season.  If the defense were getting worse in an ‘actual goals‘ sense, we’d expect to see an increase in goals conceded per match over the games in question.


Meanwhile, if New England were merely conceding more chances, we might see an increase in xGA that outpaces the actual ‘goals-against.’

Photo: —

In the end, we sort of see…neither?  The red bars above indicate actual goals conceded by the Revs’ defense.  Black bars show the expected goals conceded.


In the first 13 matches of the season, the Revs conceded 14 goals on 17.6 expected goals.  They overperformed their xGA over that span by 3.4 goals.  Since then, New England conceded 14 goals on 13.8 expected goals, mostly on target.

Their pre-May-20th ‘goals against per match’ was 1.08, and it has since risen to 1.38.  ‘Expected goals allowed per match’, however, has stayed pretty stable, going from 1.35xGA/90 to 1.38xGA/90.


So what does this tell us?


The takeaway is that the Revs are conceding chances, on a macro-scale, at about the same rate that they have been all season.  They went from conceding fewer chances than expected to concede at about the expected rate.  This could be simple regression to the mean or a dip in goalkeeping, or simply that the Revs are conceding better (but fewer) chances.

At the end of the day, however, actual goals don’t have an impact on xGD, and with a stable xGA, it’s unlikely that a dip in defensive performance is contributing to the phenomenon above.

Offense?


Below is a similar chart to the one we just saw, except this one covers goals scored vs expected goals by match.

Photo: —

This is where things start to get interesting.


There does, visually, seem to be a bigger delta between goals and expected goals over the past 10 weeks.  And, in fact, that’s what the data tells us.

From match 1-13, the Revs scored 19 goals on 21.1 expected goals, a slight underperformance.  Since then, however, the Revs have nearly doubled their expected goals rate.  They scored 23 goals over the past 10 matches on just 11.8 worth of expected goals.


That’s a big shift.

So why is that happening?


Either the Revs are creating fewer chances than they used to, but those chances are better ~or~ they’re scoring on low-percentage chances at a higher-than-expected rate.

Photo: —

This chart shows the expected goals value for every Revolution goal scored in 2023, separating out this 10-game span in orange.  This will tell us if the Revs are simply creating (a low number of) great goalscoring opportunities to account for the g/xG overperformance.


At first glance, that doesn’t appear to be the case.  In fact, the Revs average ‘converted chance xG’ only rose slightly over the period in question from 0.265 xG-per-goal to 0.268 xG-per-goal.  That would seem to eliminate the ‘fewer but better chances’ explanation for the diminished xGD.

Things get interesting, however, when we sort those goals into categories by their xG value.  Assuming anything with an xG value above 0.5 (can also be thought of as above 50% expected success rate) qualifies as a BIG chance, anything with an xG value below 0.1 (less than 10% expected success rate) qualifies as a LOW chance. Anything in between (0.1 – 0.5 xG) as a MEDIUM chance, we can plot those goals like this.


Photo: —

The blue bars represent the percentage of Revs’ goals (in each category) from before the loss at Philadelphia, whereas the orange bars represent all the games that came after.


What we can gather from this is that a higher percentage of “BIG chances” may be slightly lifting that “pre-Philly” xG-per-goal I mentioned earlier.  We can also conclude that the Revs have seemingly been at least a little better at finding higher percentage shots.

There’s been a shift over the past 10 weeks from LOW quality to MEDIUM quality goals.  It doesn’t fully account for the ‘converting at twice the expected rate’ phenomenon we see, but it’s a step in the right direction.

Game States?

One argument that could explain a downward shift in expected goal differential while also explaining an upward shift in actual goal differential is game-state.  Games don’t play out the way they do on paper.  Teams that trail in the scoreline tend to play more aggressively to try to gain ground on their opponent.  The opposite is true for teams in a winning position, who are more likely to bunker in and absorb pressure to hold their lead.

If the latter case were true, that would go a long way toward answering our question.


We can test this by determining what percentage of each game’s xGA comes while New England holds a lead.

Photo: —

Now we’re getting somewhere.


New England held the lead in 7 of 10 matches.  In 5 of those 7, they would concede over 50% of the xGA from a game state in which they held the lead.

I think it’s reasonable to conclude that New England was content to sit on a lead and absorb pressure.  That could explain how New England could outperform their xGD by such an extreme, at least in games where they lead.


Conclusion

If you’re still with me, thanks for plodding through all the charts and stats.


The question at hand was: How did New England outscore opponents 23-14 over a 10-game-span while still losing the expected goals battle in 6 of those matches?


Poring over the defensive xGA stats, we can see that while New England is doing worse at keeping teams from converting their chances, they’re conceding chances at roughly the same rate as usual.  This indicates that the defense isn’t to blame for the diminished xGD over the past few months.

This leaves offense and game states.

Offensively, the Revs are generating less xG per match than they did earlier in the year.  Over the past 10 games, they average 1.18 xG/90.  That’s down from 1.62 xG/90 earlier in the season.

They are, however, scoring at a higher rate than earlier in the year.  Of those chances, they appear to be creating slightly better looks in recent weeks.  Goals from “low percentage shots” dipped from 42.1% to 26.1% over the past 10 matches.

Furthermore, when they score, they’re absorbing pressure to hold on to leads.  New England scored first in 5 of those 10 matches, holding onto the win in 4 of those 5.

That combination of shooting less, scoring early, and sitting back to hold onto leads seems to be a reasonable explanation.


So there you have it.  Mystery semi-solved.  How… satisfying?
Tune in next week when I count all the passes the Revs have made all season.

Photo: —

Load More Related Articles
Load More In Revolution
Comments are closed.

Check Also

Did We Learn Anything From the Revs 3-1 Season Opening Loss in D.C.?

There’s an axiom within journalism, called Betteridge’s Law. It states “Any headline that …