Premier League Projections and Power Rankings, Week 7: Back to Normal(ish)

It's been a while, y'all. I had run the projections last week, but work intervened. And then i decided it would be more fun to work on the Shot Matrix database than to write up a set of projections anyway. So I have no particular shot matrix data to share this week, but I'm getting closer. I'm really rather excited about this, I think it's a degree of data granularity for football that hasn't been publicly available before. My hope is that by the New Year, I'll have information dating back several years for all the major European leagues as well as for international football. It's just a matter of drudgery and finding the time. But I figure if I flog it here publicly, that will shame me into completing my drudgery over time. So that's really the logic behind this paragraph.

I'm also stalling so I can avoid talking about the Spurs game. That was bad! Tottenham Hotspur had previously been leading the English Premier League with the lowest number of both shots on target allowed from inside the box and the lowest number of shots on target allowed from the most dangerous positions inside the box—from close and central positions. Before last week, Spurs had only allowed five shots on target in six games from these positions. Against West Ham United, they allowed four, nearly doubling the season total. The effect of this loss, which came in a home game from which we were expected to take on average 2.3 points or so, go beyond just the loss of expected points. Spurs defensive rating, which has been the driver of our strong projections, also took a medium-size dent.

And just to make things a little bit worse, my work with the shot matrix database identified a very minor coding error in my earlier data. This error didn't make much difference for 19 clubs, but it makes a small difference for Arsenal. A couple of shots allowed had been mis-coded as SiBoT when they weren't, giving Arsenal a slightly worse overall defensive rating than they should have had. So while Arsenal dropped points away to West Bromwich Albion, they also picked up maybe a half-point in expectation due to the fixing of the coding error. Great work, me.

Power Rankings and Projections

Do remember that because of rounding, not all the numbers necessarily add up quite right.

• Expected Goals methodology
• Season Simulation methodology

• Chelsea's been steadily moving up the charts, another good road win with five SiBoT / 2 SoBoT for and 3 SiBoT / 0 SoBoT against. Their defensive rating is now roughly tied with Spurs and Southampton for best in the EPL.
• Obviously it's fun to see Southampton doing so well, but I want to highlight Swansea instead for a minute. Swansea sit 15th in the table with just seven points, but their underlying stats are excellent. They have (1) played the toughest schedule so far of anyone in the league and (2) been relatively unfortunate / unclinical in their finishing. Based on the underlying numbers, Swansea should have a goal difference of roughly +3 in contrast to their real difference of -3. Last weekend, Swansea were somewhat unlucky to fall to Southampton despite doubling their opponents' numbers of both SiBoT and SoBoT (4 to 2 in both cases).
• Manchester City's victory over Everton was the big result of the weekend. City vaulted back to solidly top of the table and Everton's chances of a miracle run at top four took a solid hit. It's certainly not over for Everton, who I still think stand a reasonable chance of a brilliant Cinderella run. But given their relatively thin squad, you'd want them to build up a little bit of cushion now with most of their regulars healthy.
• I have worked up some visualizations of the data, just simple graphs showing the movement week to week in my projections of the races for the title, top four, and relegation. I'll have those ready for a post tomorrow.
• I chose "CRY" for my abbreviation for Crystal Palace before seeing that the official abbreviation appears to be "CRP." Either way, it seems insulting. I don't think they're very good, though.