:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/24216905/153036690.0.jpg)
I think this week I'm going to start with the numbers and work backward. Tottenham Hotspur host Manchester United tomorrow in the early match, and I am projecting a win as the most likely outcome. It's not quite a 50/50 bet, but Spurs are favored by a good amount in my projections. This is, of course, not the outcome that the major English bookies are projecting. They have this match dead even, maybe leaning a little bit to United. I don't like diverging widely from the people who make their living laying odds on sporting events, but it's always going to happen occasionally. As I said, I wish this year's test case weren't Spurs.
Projections for Spurs-United
| TOT W | D | MUN W | |
|---|---|---|---|
| MCofA Projection | 48% | 28% | 24% |
| Bookie Projection | 35% | 29% | 36% |
| Top 4% | 53% | 41% | 32% |
Those are huge gaps. What's going on? Well, the basic thing that's going on is the same thing that's been going on for weeks. Spurs have radically weird, nearly unprecedentedly poor goal conversion numbers in attack, so we aren't scoring goals. These are Spurs and United's underlying stats in attack and defense.
For an explanation of the stats, see Shot Matrix I: Shot Location and Shot Matrix II / Shot Matrix III on pass type / shot type effects. The main things I'm going to focus on here are (1) shots from the "Danger Zone" close/central locations in the box and (2) distinguishing shots off crosses from shots not off crosses, as it is much harder to direct a cross goal-wards than it is to score off a normal pass. Spurs of course have taken many, many more shots from distance than United, but I want to focus just on the quality shots for comparison.
| Club | DZ Shots | DZ S Non-Cross | DZ Shots on Target | DZ SoT Non-Cross |
|---|---|---|---|---|
| TOT attack | 57 | 33 | 24 | 17 |
| MUN attack | 63 | 30 | 25 | 13 |
| TOT defense | 43 | 23 | 16 | 12 |
| MUN defense | 62 | 29 | 26 | 16 |
United of course have scored about double the goals of Spurs and about treble our goals not from penalties. But based on chances created, Spurs and United are about equal. (By Opta's "big chances" Spurs have a big lead, 23 to just 14 for United). And the defensive numbers show a significant advantage to Spurs as well. So since my numbers are based on chances created and shots on target instead of actual goals scored, you're going to get a really big divergence.
Over the four-season sample I have in my database, I find no meaningful persistence of G/SoT within or between seasons. As I've argued, this doesn't mean that there can be no meaningful persistence of G/SoT or that a poor G/SoT doesn't mean anything. It might! Football is incredibly complicated and our sample is never more than 38 matches in a season, which isn't that many.
The beauty of statistics as a field is that it absolutely refuses the concept of certainty. It is always possible that with a larger sample or a different means of measuring that an effect may be seen. But lack of certainty also doesn't mean a lack of knowledge. It's just that the knowledge always comes with error bars and caveats (like, really, pretty much all knowledge). I can say that based on the data available to me, I see no meaningful persistence of underperformance of expected goals. So the best conclusion I can draw is that Spurs' rate of goal scoring relative to chance creation should regress to the league mean. That doesn't mean it will happen—indeed, this was an argument I was making before the Newcastle match—but it's the best conclusion I can draw right now. I've tried a bunch of methods for identifying a persistent team skill in scoring shots on target, and nothing's come up useful yet. I will keep trying, especially if Spurs keep not scoring their chances, but all I can report is what I've found.
All of means, in short, that I'm still basically standing by my numbers. I'm not arguing that United can't win—stuff that has a 1-in-4 chance of occurring happens every day—but I think the chances of United winning are significantly less than are projected by the bookies. When you project using the underlying numbers, you get notably stronger numbers for Spurs, and somewhat weaker numbers for United than you would using points and goals. This means likewise that a Spurs win wouldn't prove anything either, but as I said above, it's getting frustrating. I'd really rather not have a heavy emotional attachment to the same club that's diverging in historical fashion from statistical norms. But here we are.
Come on you Spurs.
