Norwich are currently the last-place club in the Premier League. Since the start of the season, they have been favorites to be relegated, if not finish in that spot. That's what happens when you don't have the resources of other clubs in one of the top soccer leagues in the world.
Norwich also have a lot of intriguing young players: Fullbacks Max Aarons and Jamal Lewis, center back Ben Godfrey, and midfielders Todd Cantwell and Emi Buendía are each 23 years old or younger and making major contributions to a Premier League team.
The Canaries have struggled perhaps in part because of that youth. Their defense is one of the Premier League's worst; Norwich have allowed the most goals in the league. Likely the best member of their back four, center back Cristoph Zimmermann, is 27 years old. At least anecdotally, it seems like the club's inexperience and defensive struggles are linked. I wondered if this was the case and decided to test this hypothesis against available data from the Premier League this season.
I started by looking into the relationship between players' ages and how often they foul opposing players. I first thought to investigate this topic after seeing Cantwell commit a foul trying to tackle an opposing dribbler. Seldom do officials call fouls off the ball, so foul frequency might be a good indicator of shoddy tackling.
At least in the three leagues I examined (this season's Premier League, Bundesliga, and La Liga) this does not appear to be true. I found practically no correlation between their ages and their propensity to foul. This went for both Fouls per 90 Minutes and Fouls per Dribble Faced.
I thought perhaps that the foul data might look significantly different if I grouped players by their positions — defenders are hopefully better tacklers than non-defenders, after all. Instead, I saw the same lack of any strong pattern.
So younger players, in general, do not foul more often on their tackle attempts than older players do. But are their tackle attempts successful less often?
Probably not. The whole data set wholly rejects the idea that younger players are worse tacklers; if anything, players' tackling might worsen as they age. But even that idea is difficult to prove, based on these stats. The Premier League and Bundesliga data suggest there might be a slight negative correlation between age and tackle success rate among defenders, but the La Liga data are nowhere significant enough to support that notion.
Interestingly, neither in the fouling data nor in these do we see anything resembling a normal distribution. If players' tackling ability peaks around a certain age, it happens on an individual basis, relative to their own career stats. In the macro, there's no obvious clustering of good tacklers at a particular age.
There's more to defending than tackling, however. The difficulty is measuring the other elements of defending in a way we can use to compare players.
Interceptions are valuable, but with the data available, we can't turn interceptions into a useful rate-based stat. FBRef.com offers interception totals, which are only a start. Let's say we divide players' interception totals to see how many they make per 90 minutes. That favors players who have more opportunities to make interceptions. Those opportunities might be more plentiful because certain teams face more passes as a result of seldom possessing the ball; because a player's teammates don't consistently win back possession, forcing him to step up; or because a player takes a role in the team that doesn't always put him in the right spaces to make interceptions. To properly measure players' intercepting abilities, we'd need to know how many opportunities they get to make interceptions, which — as far as I know — we don't publicly have.
Sometimes, a player is a good defender because of how he defends space rather than how he disrupts how opponents function with the ball. That's incredibly difficult to quantify on an individual basis.
It's not ideal, but we can try looking at overall defensive performance on a team level, comparing the average age of a team against how many goals and Expected Goals it has allowed. For the sake of consistency, as well as excluding goalkeepers, I've chosen to reference the same set of players I referenced earlier in this post.
First, here are the correlations using goals allowed.
It's fair to say we can't draw any pattern from this table. The correlations are dramatically different from league to league. Perhaps there's something intrinsic to the Premier League that makes younger teams better defensively than in the Bundesliga, but that seems unlikely, especially after looking at the tackle data above.
That said, goals are rare events. Good process (i.e., taking a lot of shots from advantageous positions) isn't always rewarded, and suboptimal process (i.e., barely seeing the opposing penalty area) can still lead to goals. Expected Goals models came about to account for this fact. Since Expected Goals is a predictive measure; maybe it will tell us more than simple goal totals.
These correlations are certainly stronger than those we saw before. The question is their usefulness. The age of a team's non-defenders, first off, appears totally useless in predicting how many Expected Goals it will allow.
Defenders' ages look somewhat predictive, meanwhile, but it's hard to say how predictive they actually are. Expected Goals are supposed to be more predictive than goals, but I wouldn't think their predictive power should create a shift this strong.
The only league with a negative correlation between defender age and goals allowed, the Premier League, suddenly has the strongest positive correlation between defender age and Expected Goals allowed. The reason seems to be that the four youngest groups of defenders in the Premier League (Chelsea, Manchester United, Southampton, and Norwich) have each allowed at least 5.2 more goals than Expected Goals says they should have.
That doesn't necessarily mean that younger defenders underperform their Expected Goals allowed numbers, however. Anecdotally, the fifth-youngest defender group in the Premier League (Liverpool) has allowed 8.4 goals more than expected, and the two most underperforming groups each of the Bundesliga (Werder Bremen and Borrusia Mönchengladbach) La Liga (Espanyol and Leganés) are older than the league average.
Perhaps younger back lines fare better than older ones, but this sample doesn't tell us enough to confirm that. It's just not large enough (just 58 teams in three leagues). Additionally, using just the average ages of the defenders who have played for each team doesn't account for how often individual defenders have played; some players have only seen a match or two's worth of minutes, while others have started every one of their club's matches. And our r-squared values in the Expected Goals test range from .04 to .08, which isn't statistically significant. There's not enough evidence to say that a younger defender group should do better than an older one without deeper examination.
Based on the what I've discussed here, I'm unconvinced that you can rely on any age-based rule of thumb when considering the broad player population's defensive abilities. Is a center back likely to be better at 23 years old than 18? There's a chance — a study of age curves would better answer that question — but if you need immediate help, you shouldn't sign a 23-year-old center back over an 18-year-old just because of the age difference.
Premier League data are as of Jan. 24, La Liga data are as of Jan. 26, and Bundesliga data are as of Jan. 28. They all come via FBRef.com.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.