Premier League Power Rankings and Projections, Week 5: Whoa

"Whoa!" In your best Keanu. A friend of mine saw Keanu eating alone in a restaurant on the Upper West Side a couple years ago. He was just sitting there. Not reading, not conversing, just sitting. He would eat a bite, chew, reflect, then eat another bite. Also a couple years ago, I saw Keanu on the Lower East Side, walking alone, dressed in black and carrying a fancy camera. He would stride forward a bit, look around, then take an artsy photograph, often with  the camera held at an unusual angle. I like that in both of these situations, his inner dialogue seemed to be best expressed by, "Whoa." This has been the end of Keanu story corner.

The reason I feel like Keanu right now is best expressed in one number. 25%. That's the percentage of simulated seasons in which Tottenham Hotspur won the English Premier League. The bookies currently place the odds of a Spurs title between 12 and 15-to-1. To be clear, I'm not saying go out and lay money on this outcome. I like my statistics, I think my methodologies are reasonably good, but when I get way, way out of sync with the professional oddsmakers, I get nervous and I fear I'm doing something wrong.

So what is going on here? First, Spurs are effectively tied for the title right now. Obviously if the season ended today Arsenal would win on goal difference, but over a full season their one goal advantage is just rounding error. That means most of our competitors have a small hurdle to jump in order to catch us. But that should be baked into the odds already. What's new in my numbers?

Well, on to the second point, my numbers have seen the Premier League as a more open competition than the bookies and other experts since the beginning. I had both Tottenham and Liverpool with title chances of around 1-in-10 in preseason, when odds were running against them at significantly higher rates. This was mostly because, by the underlying numbers, Spurs and Liverpool were just as good as the top four clubs last year despite finishing a little bit back of the leaders. United and Chelsea both significantly outplayed their expected goals scored and conceded numbers, and so I projected them to fall back to the pack.

Third, Spurs have among the very best underlying numbers through the first five matches of the season. We have put 33 shots on target, 15 from inside the box and 18 from outside. We've won a pair of penalties and created 10 big chances as classified by Opta. This adds up to a little under nine expected goals scored on the season, though we've only actually converted five. Our defense has been the best in the league so far, allowing 14 shots on target, ten inside the box and four outside. We've conceded no penalties and only two big chances. That latter number is best in the Premier League. This adds up to just 3.5 expected goals allowed, again best in the league. This expected goal difference of a little better than +5 again leads the league.

So I don't know. I am definitely going to be running some numbers below which are significantly out of sync with how most people, even most Spurs fans, evaluate the likely outcomes of the season. I think I've been pretty clear that I think my numbers are interesting, and I think I've worked hard to make the methodology reasonably sound, but there is a whole lot about football that these numbers cannot possibly capture. Still, they should be good to think with. And I look over the last three paragraphs, and I'm convincing myself that hey, this club has been really good and they project to still be really good.

We will definitely learn more about them this weekend, when Chelsea travel to the Lane. I'll discuss that match and its implication later in the week.

Before I get to the projected table, I want to share the data I'm working with, or at least most of the key parts of the data I'm working these. This data comprises much of the reason for my rating of Tottenham's title chances. I've got expected goals scored and expected goals conceded for everyone in the Premier League, along with strength of schedule numbers, plus most of the key parts of the underlying data from which it's built. This is part of the data that feeds in to my projections. I'm slowly increasing the weight of the current season numbers compared to the preseason projections, at a rate of about 3% per week. This week, the projections are based off ~85% preseason numbers and 15% 2013-2014 statistics. Below in the table are those 2013-2014 stats:

SiBoT: shots in the box on target
SoBoT: shots outside the box on target
SoS: Strength of schedule, scaled to 100

• The main thing I'm not listing here that plays a part in my numbers is "big chances." I have to pay for a subscription to a selection of Opta's data in order to get access, so I don't feel like I should share the whole data table. The rest is based on freely available numbers, so I'm happy to share.
• As you can see, Arsenal have the best attacking numbers in the league, Manchester City second. Both have played some of the easiest overall schedules, so we should expect some regression, but I think it's reasonable right now to look at Arsenal and City as the two elite attacking clubs. Arsenal's had more problems at the back than City so far.
• My numbers continue to really like Swansea and Southampton, and Southampton nicely justified my continued optimism with a solid 1-0 defeat of Liverpool. Southampton's xG number these is a little lower than you might expect based on the shot-on-target data, this is because Opta has them producing very few big chances on the season, just three. So I have their shots on target rated as relatively less likely to be converted into goals than the average team's. Swansea's performance is all the more impressive because of the extremely difficult schedule they've faced so far (MUN, @TOT, @WBA, LIV, @CRY).
  

Power Rankings and Projections

Ok, that's rather enough of that. Here's the data I keep talking about, that table that projects Tottenham as one of the leading title contenders in the EPL. Have at it. (Do remember that because of rounding, not all the numbers necessarily add up quite right.)

• Expected Goals methodology
• Season Simulation methodology

• So, there it is. Spurs improved in the table (1) because they won an away game from which they were projected to take under two points, on average, (2) because United and Liverpool both lost, clearing space at the top of the table, and (3) because I am continuing to increase the weighting of the current-season numbers as the sample size grows, and Tottenham have the best underlying stats in the Premier League this season so far. With every week that goes by and Spurs don't have a stinker, their projections will improve. Or, they'll have that stinker and the projections will come back down to earth.
• The other big results were at the bottom and middle of the table. Hull City have been pretty impressive so far this season, compared to expectations. They have extremely blah attacking stats, but their defensive numbers would sit snugly among the various lower-mid-table clubs above the relagation scrap, and they've played a quite difficult schedule thus far (@CHE, NOR, @MCI, CAR, @NEW). If Hull keep performing this well, they'll start to push out of the relegation in the projections in the next couple weeks.
• By contrast, Sunderland took another bad loss. And it was a stomping. West Brom put four shots on target from inside the box and creating five big chances, while Sunderland didn't put a single shot on target over the 90 minutes. This performance pushed West Brom well aways back up the projected table, and stuck Sunderland solidly in the relegation places. They fired manager Paolo di Canio with good cause, but it will be interesting to see if anyone can turn the Black Cats around.
• I mentioned the big Southampton win above, but it was also a big Liverpool loss. The Merseysiders put only one shot inside the box on target while allowing five. This comes after last week's draw away to Swansea in which they were likewise outplayed but managed to scrape a point. Liverpool's hot start seems to be wearing off, though they will get Luis Suarez back next week. So another hot streak is reasonably likely to start soonish.
• I'm not yet confident saying that Crystal Palace are terrible, and Sunderland deserve a chance under their new manager. Fulham have some ridiculously awful defensive numbers, most impressively 26 shots in the box on target allowed. They will probably score some goals, but I don't see much of a path to defensive competence for Martin Jol's club, and that's bad. Terrible? I still want to wait and see.