It’s well known that having a good goal difference is strongly correlated with your position on the league table. And that makes a lot of sense. If you’re scoring a lot more goals than you’re conceding, then you’ll probably be winning more often than not, and the amount of points you're getting out of games should go. Fine and dandy. However, there are really two things that go into goal difference: goals scored and goals allowed. And so we might want to ask the question, what’s better? Being really good at scoring goals or being really good at keeping them out? Is a goal in their net worth two (not) in your own? I’m going to attempt to use data analysis techniques that I’ve learned as an economist to answer this question, mostly because I’m bored on a Sunday night and statistics are fun for me.
But before I continue, I cannot stress enough that the results will only tell us what teams can expect to happen ON AVERAGE and not in any way indicative of what will actually happen in reality in every instance of a thing. So if we find, for example, that the best strategy appears to be that teams should park the bus when in an away game, then pointing out all the examples of when this strategy failed is missing the point. Further, if, for example, we find that attacking is way more important than defense, pointing out that, "well, yeah, maybe, but the best team in the EPL era was 2005 Chelsea, which was super good at defense but not as much at attacking" is also very much missing the point. I know that there are counterexamples; part of the beauty of sports is that for every trend there is an exception. We are only talking about average numbers here. Moving on.
First, let’s take a look at the All-Time Premier League table (data from Wikipedia) and see what we can learn from that:
The obvious thing we see from the above is that Manchester United under Sir Alex Ferguson were really really good at football. They scored 210 more goals than anyone else in the history of the EPL, and have a very comfortable 240 point lead at the top of the table. Which only makes their troubles in the first season after Fergie retired that much more enjoyable and hilarious. But instead of just staring at an ugly table hoping to see patterns, let’s run a regression and see what that tells us (note: if you want to know more about my regression techniques, ask about it in the comments, although I didn’t do anything particularly fancy at all).
The regression analysis of the All-Time EPL Table tells us that for every additional game played, the average premier league team expects to get 0.86 more points, over the course of all of the premier league seasons. It also tells us that for every goal scored, they can expect about 0.82 more points on average, whereas they suffered an average loss of 0.44 points from conceding. Broadly speaking, this indicates that scoring has nearly twice as large of an effect on your expected points total as conceding does. Pushing that to its most extreme end, you get the argument that in the entire history of the EPL, attacking has been twice as important as defense, or at the very least that scoring a goal is twice as important as keeping one out. With this in mind, you’d have to say that all of the calls for Chelsea to splash the cash on a big-name striker this summer have more than a bit of wisdom behind them.
And that’s some good stuff, but I want to know what happens if we delve a little deeper (but not deep enough for any Balrogs please!). What about if we break just simple goals scored / allowed up into home and away games, and also look at teams on a season-by-season basis?
This data comes from statto.com, which was the first website that Google turned up when I was looking for historical EPL tables by home and away. Now, before running the regressions, let’s have a look at the top 20 teams, points-wise, in the history of the EPL:
The one thing that jumps out to me is that 15 the top 20 teams are from the last ten years. That indicates that to win the Premier League now is as hard as it’s ever been, and that it might take more points than it previously has to get there. However, seeing as this year has been as crazy and competitive of a season as I’ve ever watched, we may well see one of the lowest points totals for a champion as we ever have in the history of the EPL. For example, Chelsea, who are currently sitting at the top of the table (come on you Blues!), are only on pace for 83 points, which would only have been good enough for third place in 2008, 2009, or 2010, but this year it might (hopefully) be good enough to win the league. Parity!
Now let’s see what the regressions turn up:
The first that I notice is that the average number of points per team over time has gone down in the EPL, by about 0.28 points per season per team. This seems to go against what we noticed earlier, that the best teams have mostly come from the last decade, but can easily be explained by thinking a bit about how averages work. If the best teams are getting better but the average is going down, then that also means that the middle and lower teams are getting worse, and enough so that it outweighs the increase in points at the top. While we’re not as bad as Spain or Scotland in that regard, it does mean that the disparity between the top teams and the also-rans is getting bigger over time in the EPL.
The next thing to notice is that the numbers, while differing slightly from the previous regression, are not so far off enough as to be worrisome. We still see that scoring goals is more important for your expected points total than preventing them from being scored against you, although it’s closer to 1.5 times more important according to this data (versus the 2x more important that we saw from the previous data). Why that’s happening, I’m not exactly sure. My best guess is that the data differs in some minor way, but enough so that it’s creating this little oddity. Hypothetically, it should be the exact same sample and so we should be getting the exact same number, but either way, it’s not a huge difference and I’m not particularly worried. But now we get to the really juicy part: the home / away analysis.
We still see that the expected points per team goes down by 0.28 points per season. Mathematically, that probably mean we are seeing more draws on average than in the past, which is indicative of greater levels of parity. Interesting stuff, sure, but that’s now what we really care about. We want to know about goals! Well, first, the numbers – on average, we see that scoring a goal at home is worth +0.71 points, conceding a goal against at home is worth -0.61 points, scoring on the road is worth +0.83 points, and conceding on the road is worth -0.42 points. That means that not conceding at home is almost as important as scoring at home (so build yourself a fortress!), and one and a half times more important than conceding on the road, which is not even worth half a point lost. Further, scoring in away games is far and away the best thing a team can do for its expected point total. It is worth a whopping +0.83 points, which means that in the history of the EPL, every away goal was very nearly worth on average as much as playing an extra game (+0.86 points). That is shocking stuff, and shows that the bravest teams who go away from home and play to score are usually rewarded for their efforts.
Anyway, that concludes the analysis, and if you got this far, then you’re either really bored or really love stats. Either way, leave your thoughts / suggestions / interpretations in the comments. KTBFFH!