We’ve finally finished gameweek 6 which means the stats are starting to take some shape. This post will look at some of the key team ratings so far this season. I’ll provide some brief explanations with each rating so you know what’s what, and (hopefully) in the next day or two I’ll have my player ratings finished also.
So let’s kick things off by looking at the league as a whole with Points Expectation. This looks at the league table and tells us whether teams are over/under performing based on a weighted equation of their shots wide, shots on target, and goal difference.
Fist point of note here of course is that Manchester City are outperforming their expectation by just over 2 points. Having won 6 of 6 it’s clear to see why. City have scored 18 goals in 6 games (3 per game!) and have only conceded 5 (0.833 per game). They have scored 78% of the goals in the games they have played and have scored 105% above the league average. It’s this huge goal imbalance that sees them above expectation, that being said they are well ahead of the pack already and their strong form shows no signs of stopping.
Moving to the far end of the scale, West Ham have had a howler of a start to the new campaign. There was much excitement around the Hammers powering on in their new stadium, but they just have not shown up so far. How a top-half team goes two goals up only to conceed four is just baffling. What’s even more surprising is that the PDO (more on that later) numbers show that they have fully deserved the results they’ve seen.
Stoke Stoke Stoke… There are no words. They are 0.048 points below expectation. That is all.
One of the notable over-achievers so far this season are relegation candidates Hull City. They have put in some admirable performances, but are 2.672 points ahead of where they should be and should back-track over the next few weeks.
Adjusted Massey Ratings
This rating is developed from Kenneth Massey’s ratings system. Kenneth Massey developed a system of rating teams’ past performances based on the strength of the opposition they have played (putting it basically). Anyone interested in the maths behind it should head over to the pena.lt/y/ blog. I’ve adjusted the method, and so far so good. I’ll be showing my method for formulating this in the near future.
These ratings range between +1 to 0 for attacking and 0 to -1 for defending. Any team above +1 when looking at attack ratings is playing some great football, any team below 0 has no serious attacking strength. With the defence ratings, any team above 0 can be considered a fortress at the back, and again, the closer to the lower boundary and beyond, the worse a team is at defending. The third column shows each team’s overall rating.
Here’s a general overview of each teams’ performance. Some things that stand out right away is the numbers way above 1 in attack and below -1 in defence. This is due to noise early in the season and tends to level out around the week 10 mark. The reason why I have included it here is to highlight how useful this measure is at charting runs of form. Palace have been improving week on week since the start of the season and have real confidence in their play. Everton on the other hand look to have reached a temporary peak, having lost out to Bournemouth in gameweek six, and have dropped back slightly in what was a dissapointing result for them after Lukaku bagged a hat-trick the week before.
Looking at Manchester City highlights the early noise in this measure. Over the four years I have been using this, the beginning of the season has alway had huge peaks and troughs for the first number of games before settling down. Odds derived from these ratings compare very well with bookmakers odds and are incorporated into my match prediction algorithm. Take a look at Manchester City’s overall performance from the start of 2012 until the end of last season:
Looking at the overall, attack, and defense ratings, its clear to see that at the beginning of each season, there are huge fluctuations in the ratings. This is simply due to lack of data for the new season. I’ll save further explanation of the ratings for another post.
Here’s a look at the attack and defense ratings together. Really helps to contextualise how the season is going so far:
Once again it’s clear to see Manchester City way out there, almost a full 1 point above the measure’s upper limit. Some other mentions here are the following best 5 teams in attack – Manchester United, Chelsea, Liverpool, Arsenal, and Everton. Also the next best 5 in defense – Manchester United, Chelsea, Everton, Tottenham, and West Bromwich Albion. Surprising to see Hull so prominent here, but as things even out they should move to the lower end of the pack.
The may not slip as far as Sunderland, who for yet another year have managed to cling on in the EPL. They currently have the worst attack and defense in the league but are close on the heels of Stoke, Bournemouth, and West Ham United.
TSR / SoTR
Total Shots Ratio is the bread and butter of the amateur stats analytics community. In general terms it is widely accepted as a usable metric with good predictive power, but it does have its drawbacks – one major one being that you have no context to the numbers you are seeing as it does not take into account the location the shot was taken from, or the circumstances in which the shot was taken. The formula for TSR is:
Shots taken/(Shots taken + Shots faced)
The predictive power of this metric has well been documented. A cursory search of google will give plenty of articles and information on studies other analysts have done in the area. Teams dominating proceedings are the big six, along with Everton, Southampton, and Crystal Palace. All of which have been playing nice football lately.
Here’s a look at a quick match prediction using TSR:
a look on oddsportal shows the odds at the time of writing are as follows:
Not too bad for a metric that does not require to much work to figure and maintain. I’ll get into detail about how I calculated the above in a future post.
Alright, so this post was quite long so I’ll finish up now with some extra stats for you.
Accuracy – Ability to hit the target as a %
FinishingI – Goals compared to total shots
FinishingII – Goals compared to shots on target
Save% – Percentage of shots the team’s keeper saved
SoTR – Shots on target ratio: SOTfor/(SOTfor+SOTagainst)
GoalR – Goal Ratio – same calculation as above
PDO – PDO is a measure of luck in very basic explanation. More on this in a future post.
If you made it through the whole thing, congrats. Any questions or thoughts feel free to comment below.