I'm excited about my plus-minus statistic. For a game as dynamic as football, there is good reason to give all players on the pitch credit for the good and bad events that occur in the game. One winger not tracking back, one missed communication in midfield, one good outlet pass from the keeper to a fullback, one intelligent run by a forward who never receives the ball but opens the space for his winger to run into, all of these can be the determining event that leads to a goal, and no individual stat we have even begins to capture those effects. Plus-minus at least makes the attempt.
At the same time, I don't want to oversell the whole plus-minus thing. As I'll talk about in the full article, there are many confounding factors in trying to do plus-minus for football, and my solutions to these problems have been, let's say improvisational. But I do have some numbers that at least should be good to think with, as they say. That's my teaser for the week's articles.
Creating an adjusted plus-minus stat required, first, building a raw plus-minus stat. The adjusted plus-minus is arithmetically messy and even more of a mess methodologically. The raw plus-minus is at simple as it gets. How many goals did Tottenham Hotspur score with Mousa Dembele in the game? How many with him out of the game? How many shots on target from inside the box? How many shots on target from outside the box? For each player, I have a simple plus-minus for goals scored and expected goals scored.
As I'll explain later this week, I think the raw plus-minus is a not very useful for most players—that's why it needs to be adjusted. But in aggregating the raw plus-minus numbers, I found some interesting stuff to share on a Sunday.
First, the questions. Hugo Lloris won the starting goalkeeper job in late November from Brad Friedel, and he held in through the end of the season, missing only one match due to injury. By plus-minus, this was by far the most important change Tottenham made during the season. Spurs played 11 matches before the goalkeeper change, amounting to about 1047 minutes of football. There were 27 matches after the change, for 2575 minutes of football. These trivia questions concern those splits.
1a) What was Tottenham's goal difference in the first 11 matches of the season?
1b) What about our goal difference in the 27 that followed?
2a) Of the 1047 minutes played in the first 11 matches, how many featured William Gallas in Tottenham's defense?
2b) How many of those 1047 minutes featured Michael Dawson?
- As Brett Rainbow pointed out in a fan post about a month ago, Tottenham had a really terrible G/GA record with Scott Parker on the pitch. Spurs scored 21 goals in Parker's 1437 minutes played and conceded 19. (This is especially poor since Parker played every single minute of his EPL season with Hugo Lloris in goal.) However, the expected goals numbers are not that bad. Based on SoTiB and SoToB, I have Tottenham at 24 expected goals scored and 14 expected goals conceded with Parker on the pitch. Once adjusted for context, these numbers are actually a little worse than expected, but they're not horrible. It's an interesting question. Was Parker in part responsible for the poor conversion rates of Tottenham's attack and the excellent conversion rates of Tottenham's opponents in his time on the pitch? Or was it all randomness? I don't think that every single deviation of real goals from expected goals is "luck," but I don't currently see an obvious explanation for a gap of nine goals. What do you think?
- Surely you want the Dembro numbers. They're gorgeous. In the 1040 minutes that Sandro and Dembele played together on the pitch, Spurs scored 23 goals and allowed 8. The xG/xGA numbers line up too—21 expected goals, 8 expected goals conceded. The health of Sandro and Dembele is definitely one of the great what-ifs of the season, along with the delay in changing over the keeper's position from Friedel to Lloris.
- Are there any player combinations you want the plus-minus numbers for? Give me enough questions and I can do an all-request piece. I will run the numbers for all individual players in my plus-minus article next week. I'm particularly interested here if there are player combinations like Dembro, whether two-player pairs or back fours or wings or anything that you think will have notable plus-minus splits.