Here is week 12:
Notes on the results:
- A lot of weird results in this week’s games. City thumped Villa on xG but couldn’t find the breakthrough and drew 0-0. Liverpool handily beat Palace on xG but lost 2-1. Chelsea beat Stoke on xG but lost. Arsenal beat Spurs on xG but was arguably lucky to draw after a late Kieran Gibbs equalizer rescued a result for them.
- This is the start of Leicester’s improved defensive record—only .2 xG conceded at home to a Watford team that loves to play on the counter.
- West Ham has had a bunch of games where they beat xG. This is one of them. (Don’t miss this Stats Bomb piece from James Yorke that briefly discusses West Ham’s finishing.)
- Of the teams in the top half, two of the most horrible to watch have been Manchester United and Southampton. This week’s results are consistent with that. These two teams produce results and can occasionally make highlights via an audacious bit of skill from one of their stars. But on the whole they aren’t known for being easy on the eyes.
How the model has done:
- Week 1: 5 correct results, 5 incorrect results—but one incorrect result was .8-.9 on ExpG and 0-1 in reality and another was .7-.8 on ExpG and 2-2 in reality. So not a bad week by any stretch if you basically grant those two as correct results because of how close the margins are.
- Week 2: 9 correct results, 1 incorrect result—and the one incorrect result .7-.6 on ExpG and 0-0 in reality. So that is basically 10 correct results, actually.
- Week 3: Only four correct results and six incorrect results, but there were a lot of really wacky draws this week.
- Week 4: Another week with only four correct results, but also another week with a lot of wacky results.
- Week 5: Eight correct results, with another “incorrect” result when the ExpG gap between teams is .1 in a game that ended in a draw. So basically nine correct results.
- Week 6: Four correct results and a handful of close results that are to be expected. The biggest oddball result is the City v West Ham result.
- Week 7: Six correct results, but nearly all of them from matches where one side clearly dominated the other.
- Week 8: Six correct results, but a couple close ones as well where ExpG is separated by .1 and the match ended in a draw.
- Week 9: Six correct results plus a narrow miss on Spurs-Liverpool that is close enough we may not even want to call it a miss.
- Week 10: Eight correct results.
- Week 11: Seven correct results, plus a draw in a game where the xG difference was .1.
- Week 12: Four correct results, plus a few close results on draws.
If you have questions about how this series works or about expected goals (ExpG) more generally, review our first post in this series.