Premier League Power Rankings and Projections, Week 6: Whoa, Pt. 2

The projected table this week is certainly a thing. I can explain it, no doubt. Tottenham Hotspur have some of the best underlying numbers in the league, as I showed last week. The underlying numbers projected a draw in the Chelsea game as well. Both teams produced two big chances, Spurs had three shots inside the box on target, Chelsea had two, but Chelsea had one more shot outside the box on target. Both Spurs and Chelsea had just one shot on target from a central position inside the box. Last week's result looks like a fair one, by the numbers.

Let me talk about that last stat for just a minute. I've been building a new database recently, which I think I'm going to call the "Shot Matrix" database. It logs for every shot a location broken down into eight categories: 6-yard box central, 6-yard box wide, 18-yard box central, 18-yard box wide, 18-yard box tight angle, outside the box, outside the box tight angle, long range. I am hoping to have more time in the next month to break down my findings, but so far the biggest one is that not all SiBoT are created equal. The table below is just for 2013 stats in the Premier League, so the sample isn't yet large enough to be definitive. I am going to try to collect this data for previous seasons in order to get a sufficient sample, and then I plan to renoobulate all my power rankings numbers based on this data. But that's still a project in the pipeline, it's not complete. For now, here's just a table with data from this season demonstrating the quality of shots from different locations.

I have too few tight angle and long range shots logged for those percentages to even be publishable. The numbers the six-yard box are also very tentative, and everything here should be taken inside a wide confidence interval. I am struck by how, at least in this small sample, wide shots inside the box are not really much better than shots from outside the box, and may even be worse. I want to emphasize here that the sample is currently way too small, and there could be some pretty big flukes in these numbers. But I think they show something basically intuitive, that good shots inside the box come mostly from very close range or from central positions.

The Shot Matrix database will have a good bit more information, including type of shot (header / free kick / regular), how the shot was created (open play pass, open play cross, corner, set piece, through ball), and I'll be able to track some individual stats as well for the player who takes the shot and the player who makes the key pass. I'm hoping to build a better key pass stat that gives more credit for key passes leading to higher-percentage shooting opportunities. But those are all future projects.

Right now I've got my projected table. It's possible that in future weeks the projected table will change as I add in more information. My initial look at the numbers suggests that Tottenham will drop a little bit when this better data is figured in. We have a slightly below-average percentage of our SiBoT from central areas, though defensively we have allowed the fewest central SiBoTs in the league. My intuition is that the new stats will cost Spurs maybe a point in the projected table. That is work still very much to be done, and I probably won't be able to get to in until the international break.

So this is what I've been delaying a little bit. I've got Spurs projected at the top of the table. I should be clear, I really have a three-team tie at the top of the table, and you could easily include 4th place Chelsea in that tie as well. I can explain it. What happened was, in last week's projections, Spurs were tied with Manchester City and four points ahead of Chelsea, United and Arsenal. Spurs dropped about one expected point drawing with Chelsea while Manchester City dropped about two expected points losing away to Aston Villa. That puts Spurs at 73 points and City at 72. Arsenal picked up close to two points winning away to Swansea, and improved their underlying stats a bit with a big away win, but it wasn't quite enough to make up the full gap. Arsenal move up from 70 to 72. Chelsea's draw was roughly holding serve for them, they stay at 70 points. So you get what we've got below. Perhaps with the full Shot Matrix data this will change. But for now, your sort of projected EPL title winners, Tottenham Hotspur.

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

• As you can see, Spurs dropped a little bit in their projections, it's just that City dropped more, United dropped a lot more, Chelsea stayed steady, and Arsenal didn't quite pick up enough to make up the difference.
• So, Manchester United. They have played a difficult schedule overall, but the stats are just not good even accounting for schedule. They have conceded as many goals as they've scored, and it's not a fluke of conversion rates. United have put 15 shots on target from inside the box and have allowed 14 SiBoT. Of those, only 7 of United's shots on target from inside the box have come from central areas, while 11 of their SiBoT conceded have been of the most dangerous sort. This is a club in legitimate trouble right now. They are plain and simple not playing top four football.
• Another club with continuing doubtful results is Liverpool, despite a solid win over Sunderland this weekend. Liverpool created opportunities, but once again they allowed a weak opponent a bunch of legitimately good chances. Sunderland had two big chances and two SoT from central areas inside the box, along with a small handful of SoBoT. Liverpool's underlying stats this season suggest they are just too porous at the back to compete for the title, though they should be in the competition for top four. Of course, it's early and perhaps the new players on the back line just need time to jell. But for now, that's why Liverpool are sitting 5th in the projected table.
• Another big Southampton win. They're good. Don't be fooled by only five goals scored, three from open play. They're creating chances and stand fourth in the league with 15 shots on target from central areas in the box. The goals are coming for the Saints.
• Rogue Convict in the comments asked about the average points for the title winning team. I ran some sims and here are the numbers: The average points of the title winning side in my simulations is 81. So even though Tottenham are projected at the top of the table with an average of 73 points, that doesn't mean 74 points wins the title. In the 24% of simulations where Spurs won the title, they took a lot more than 73 points. More like 81. The same is true of the other clubs in the table.
• I've also got numbers for 4th place and 17th place. The average 4th place club had 70 points, and the average 17th place club, just escaping the drop, had 36.
• I think Crystal Palace are getting relegated.
• I think Martin Jol is getting fired.