I mentioned in my power rankings piece on Tuesday that Swansea City are one of my picks to improve over the next few weeks. They have 26 shots in the box on target this season and have allowed only 18, with also a 17-9 advantage in shots outside the box on target. Despite this, the Swans have conceded three more goals than they've scored on the season. This is the largest gap between actual goals and expected goals in the Premier League. I figured it would be fun to take a look at the whole EPL table in terms of expected goals and difference between expected and real goals. So here it is.
Expected Goals Table
Do remember that because of rounding, not all the numbers necessarily add up quite right.
- Expected Goals methodology
- "+/-" is how many goals difference plus or minus the club has outplayed or underplayed their expected goals.
- "SoS" is strength of schedule on a scale centered on 100, above 100 is harder and below 100 is easier by those percentages, so a 102 schedule is 2% harder than average.
|West Ham United||4.8||5.5||-0.7||7||5||+2||+2.7||97|
|West Bromwich Albion||5.9||7.5||-1.5||7||6||+1||+2.5||105|
So not only do the Swans have expected goals numbers right up there with the league's elite, they've put up these numbers while facing the second-toughest schedule in the league. Chelsea's numbers, while less impressive overall than Manchester City's, have come against significantly more difficult competition. They are quietly developing a title-contending resume. Fulham, by contrast, have played the easiest schedule of the bottom-dwellers. They don't look good.
Another interesting stat: Cardiff City lead the EPL in goals scored off corners with three, from only eight shots off corners. That is almost certainly an unsustainable rate of conversion from corners, and the gap between Cardiff's goals scored and expected goals scored can be mostly explained if you give them a more league average rate of corner conversion.
Finally, there you see Aston Villa in the middle of the table. They have roughly even expected goals scored and allowed numbers, but no one has faced better competition. Aston Villa have held their own or better against all of Arsenal, Chelsea and Liverpool this season. I am still not fully buying that they're this good, and my team projection numbers still include a significant preseason rating factor. So I'm projecting Tottenham as plurality favorites to win the game. But if Aston Villa are actually as good as they've played this season, which would make them a solidly above average EPL side, then this match is much closer to a 50-50.
Projections for Spurs-Villa
|Outcome||TOT W||D||AVL W|
The difference between me and the bookies is once again rather tiny. I'm projecting a slightly higher chance of a draw most likely because my numbers treat Spurs as an excellent defensive side and a merely pretty good attacking side. The fewer goals projected, the higher the chance of a draw. (As a note for bettors—specifically of course those living in places where such things are legal—that 3% difference is 100% useless to you. To create the bookie projections, I have to erase the "vig" or the automatic profit built in to all future bets of this sort, in order to get the percentages to add up to 100. That little gap between my projections and the bookies, even if you for some reason wanted to lay money on it, would be useless to you because actual odds, including the "vig" would give you no advantage at all.)
These are the sorts of games you need to win if you're going to make a run in the Premier League. It's a tough game, and at best we're 50/50 to drop points. Villa might be better than I have them projected, I doubt they're worse. Spurs need to be as good as they're projected in order to get the points, and I am moderately worried after the last couple matches that we might not be. A win would be nice, not only for the obvious reasons, but because I'd like to be more confident in the numbers than I am right now. Another bad loss and I'm going to feel like I should really go under the hood and see if I've been missing something in the data up to this point.