31 March 2009

A Look at Fouls and Defense

Perhaps I am incorrect to do, but I sometimes get the impression that Tom Soehn believes you can evaluate the toughness of a team by how willing they are to foul. Now, I am somewhat skeptical of this idea, but my skepticism should not be taken as Truth. So I'd like to examine this question in a little more detail.

Let's call this the Theory of Hack'em. I'm thinking the rational is that the closer your mark and the more frequently you challenge an attacker. As a result of this, you may have more free kicks conceded to the opposing team, but the overall effect is to blunt your opponent's attacks. When worded this way, it doesn't sound as ludicrous as a surface reading might suppose.

Now, as I said, I'm not sure I believe this theory at all. To me, fouls are far more likely to be a indication that a players is beating you on the ball, and so fouls indicate a weak defense, not a strong one. But this is a hypothesis that we should be able to test. I don't think I do so here, but I wanted to at least provide the overview before we went further, to give people an opportunity to say "I think you're barking up the wrong tree here."

Let's look at 2008. We have a full season's worth of statistics, so let's compare Goals Allowed per game against Fouls Committed per game. If we plot that, we get something that looks like this:

United, for your reference, committed 12.63 fouls per game (7th most in the league) and allowed 1.57 goals per game (2nd most in the league). Chivas led the league with over 16 fouls a game, and LA of course let in the most goals. Now, from this look, there doesn't seem to be any correlation between fouls committed and goals allowed. If anything, there's a weakly positive correlation.

So on a league wide basis, at least confined to the 2008 season, this seems to be bunk. But here's how I propose to really look at this (and it will take me some time, so if you think this is a fool's errand, speak now): This could be something we can only see in a season. Let's take a look at United's games only. We know that, on arithmetic average, they committed 12.63 fouls per game. So on games where they did more than that, did they allow fewer goals? This allows us to control for just the Tom Soehn system, plus we should be able to pull out 30 or so good data points for 2008. Make sense to you? Any takers on what we will find? Because honestly, part of me hopes I am wrong about thinking this entire theory is bunk.



At 31 March, 2009 16:21, Anonymous Grunthos said...

Thinking out loud...

I think trying to pursue controlling for confounding influences by reducing the data set isn't going to help you, primarily because you can't control against enough factors here, statistically. The number of fouls committed will vary by home/away, scoreline, personnel on the field, number of cards already shown, quality of cards already shown, game temperature... we could profitably ask what the correlation is between FC and GA across MLS history, or all professional leagues above a certain income level, as you have done for MLS 2008; but establishing causation with multivariate controls is going to take more time than it's worth, IMO.

Perhaps a better tack would be to rephrase the question with improved theorizing. We know FC can be "good" (fouling in your attacking zone to deny your opponent creative time) and "bad" (fouling just outside your box when you already outnumber the oncoming attackers). If you went over the data and sorted fouls into three baskets (good, neutral, bad) based on field position, time elapsed, scoreline, and balance of possession/shots/corners, and then asked whether each basket was correlated with GA in some way... or whatever, pick your method, but separating the "good" and "bad" effects of FC looks to be in order. Assuming the good FC depress GA (as we would intuitively expect) and the bad FC increase GA, then this Soehnist theory would be supported if the neutral FC also had a significantly negative effect on GA.

At 31 March, 2009 16:26, Blogger jrnail23 said...

I think Grunthos is probably right, and the one anecdotal case I can think of that demonstrates a disconnect between fouls and good defense is Michael Parkhurst.
The guy hardly ever fouls, yet he's been universally regarded as one of the top defenders in MLS (although now he's in Europe).
On one hand, fouls can demonstrate defensive pressure and hustle... on the other hand, it can signal bad positioning, poor tackling technique, and even laziness (grabbing a jersey instead of moving your feet to stay with your man).

At 31 March, 2009 16:32, Blogger D said...

I agree with both of you, but here's why I still think there's merit in this.

First, I don't think we can describe good/bad fouls necessarily.

But we can look specifically at how a Tom Soehn team uses its fouls. In short, are they being used to reduce goals, in aggregate, or not? That seems a testable hypothesis. If we look at all 30 games last year, then we should see in games where United has 13+ fouls a GAA below the 1.5 they were posting. Theoretically.
It will also be interesting to see if there is a normal distribution of fouls in United's 30 games...

At 31 March, 2009 18:01, Anonymous Bootsy said...

D writes: But we can look specifically at how a Tom Soehn team uses its fouls. In short, are they being used to reduce goals, in aggregate, or not? That seems a testable hypothesis. If we look at all 30 games last year, then we should see in games where United has 13+ fouls a GAA below the 1.5 they were posting. Theoretically.

That's only true if all other factors wash out over the course of a season. Perhaps, for instance, the matches where DCU played tougher opponents (and gave up more goals) also tended to be matches where the center ref was one who's historically a bit more whistle-happy (thus calling more fouls on DCU than average). That would appear at first glance to contradict the hypothesis you're testing; but it doesn't mean what it seems. And that's just one factor. Grunthos suggested a bunch of other factors involved; to have all of this wash out, you have to have an extremely large sample size, and my gut feeling is that two years of Soehn aren't enough to give you that large a sample.

But I'm looking forward to what you find. Off on a trip, back on Friday evening.


Post a Comment

<< Return to The DCenters Main Page (HOME)