Once again it has been a few weeks since I last posted an ELO update. This is due to a new xGoals model I am working on for Bundesliga which I should have up and running in the next couple of weeks.
For this season I decided to experiment with ELO ratings. I started each team with a total of 1500 points to see how well they progress as the season unfolds.
ELO is a system that was originally created to rate chess players, but it has since been adapted to many other sports, including football. There are some great websites that track ELO ratings of teams from all over the world, one of the best I have come across is ClubElo. In a nutshell, ELO gives a rating to two teams/players and when they face one another the number of points gained and lost is equal so there is alway the same number of points in the pool.
Firstly, let’s take a look at the overall table
Chelsea continue their reign at the top of the ELO table with almost 73 points between them and Tottenham. Manchester United continue to climb the table, taking 3rd place. Liverpool find themselves again in the top 4 after an important victory against Tottenham. I tweeted a graphic last week showing the average rate of change by Everton and Liverpool this season. Last week Everton broke into the top 4, but after a joyless 0-0 at Middlesbrough, they have fallen to 7th.
At the other end of the table, Crystal Palace remain as unmovable as a passed out drunk Allardyce on a park bench. It is their second week at rock bottom but a grueling six weeks in the ELO relegation zone. Their highest record this season was 1523 points back in gameweek six. Since then they have slid 153 points.
The above chart shows Crystal Palace’s ELO performance so far this season. It highlights the difference in rating between Crystal Palace and their opponents. Palace have been rated lower than their opponents 15 times, and have only managed to pick up 10 points from a possible 45! More worryingly are the points they are dropping from games they managed to get results from in the first half of the season. They beat Middlesbrough, Stoke, and Sunderland on the bounce but lost out to Sunderland and Stoke in their last two games and face Middlesbrough on the 25th February. They are averaging just 0.7 points per game and have a very rough ride from GW 28, playing the Big 6 teams every second gameweek until season’s end. Palace have only come away from 12 home games with 7 points which is the worst in the league (0.57 per game). They have yet to host Tottenham, Arsenal, and Chelsea. Any hope for Palace will come in the interim of these games in the winnable(?) games like Boro(?), Watford(?), Burnley(?), and Hull (?) as their schedule yo-yos into the void.
Hull City, a team I had written-off a long time ago have pulled themselves out of the bottom spots after spending SIXTEEN WEEKS in the bottom three.
Another team that looks willing to duke it out is Swansea City, winning 4 games out of six since the start of 2017, with huge results against Liverpool and Southampton. After giving up on the American Experiment, they look like a team that knows how to actually play football again. How much of this is down to – or how much will he claim – Daniel Altman supposedly pulling the strings in the Analytics room?
Rolling Rates of Change
One thing I’ve started looking at lately are rolling rates of change. By looking at a five game rolling average we can look to when and how drastically teams improve and decline over the course of a season and minimise effects of bizarre results when we look at the season as a whole. I have organised these charts into cities/rivalries to make for a more interesting comparison between these teams.
Liverpool & Everton
City & United
Chelsea, Arsenal, & Spurs
Some points of note:
- As a team’s rating increases, the number of points it gains for a victory against inferior teams begins to decrease. This can be seen by looking at Chelsea’s rolling average.
- As a team’s rating increases, the number of points it loses for a defeat to inferior teams increases. This can be seen in Tottenham’s rolling average. After defeat to Liverpool, Spurs lost 26 points from their rating.
- The opposite for teams with a very low rating is true. For example, Crystal Palace would be rewarded very well for a victory against Chelsea, but would not lose so many points were they to lose to the Londoners.
Next time I will write about using these Seasonal ELO ratings for sports betting and discuss some of the results I have found in the Premier League and Bundesliga.
As usual, if you have any questions please feel free to ask below, or contact me on twitter. If you would like to see your team’s ELO charts send me a DM @petermckeever