Month: April 2014

Categories
Elo Ratings

Elo Match Rating System

Match Ratings

Currently I am producing match ratings for English (including Conference) and Scottish leagues.

The method I use is based mainly on articles found on clubelo.com. I say mainly as there are some differences. My ratings only take league games into account. FA cup, League cup, European matches and friendlies are not included.

I also apply a static home advantage of 90 points for English games and 50 points for Scottish games. Match ratings for English and Scottish league games start with the 2000/2001 season.

Team ratings

Elo Ratings

Calculating the Team’s Elo Rating

Match ratings are generated using Team Elo ratings therefore team Elo ratings where calculated a couple of seasons earlier than the match ratings starting with the 1998/1999 season. Each team in the four English and Four Scottish divisions are given 1,500 Elo points for the start of the 1998/1999 season.

As the fixtures progress the changes in the team rating begins to reflect the strength of each team relative to the other teams in the division.

Elo Team Ratings

English

Scottish

 

Ratings for the Conference league started with the 2005/2006 season.

Relegation and Promotion

At the end of each season an adjustment is made to deal with teams that are relegated and promoted. In order to keep the number of points in each division constant I have opted to simply reallocate the total points of the division’s relegated teams and assign them between the newly promoted teams and vice a versa.

This may not be the best approach particularly when a division has only one promotion spot like the Scottish Premier. At some point I’ll look closer at how relegated and promoted teams perform in their following season.

Rating Factors

A team’s new Elo rating is recalculated after each league game with the new rating depending on a number of factors. These include:

  • The team’s old rating
  • The Elo points difference between the home and away team
  • The result of the match
  • The expected result of the match
  • The goal difference of the match
  • An adjustment for home advantage

A team’s rating may increase or decrease or occasionally stay the same.  Rating points are exchanged between the two sides. For example, if a home team’s rating is increased by 7 points then the away team’s rating is decreased by 7 points.

A match rating for the next game is created by simply taking the home team’s Elo rating and subtracting the away teams’s Elo rating. A high positive number indicates that the home team is the stronger team and a high negative number indicates that the away team is stronger.

My next task is to understand Bivariate Poisson Distribution to calculate probabilities and odds for match results and correct scores.

Categories
Football Rating Systems

Rating Systems for Football Matches

As mentioned in my first post I imported English and Scottish league results from Football-data and ran some basic rating systems against the data. You can view match ratings for past results and fixtures as well as recent form for the teams prior to each game.

The rating methods I’ve used are based on articles found online and are not meant as any kind of sophisticated method for picking winners or making a profit in the long term. I ran these on UK football results from 2000-2001 up to the current season just as an interest and they are only a starting point for any analysis. The screen shot belows shows how you can select and navigate through past results back to season 2000-2001.

Example Table Showing Match Ratings

football-ratings5

Basic Football Rating Systems

I’ve looked at three match rating methods including an Elo rating system, goals superiority and a basic Win = 3, Draw = 1, Loss = 0 points method. However, no reference is made to true odds, value, goal expectancy or Poisson distribution etc. These are related topics with plenty of in-depth discussions online that I still hope to read and fully understand.

Win/Draw/Loss Match Rating System

The basic rating takes the last 6 league results for each team (home or away) and totals the number of points gained: Win = 3 points, Draw = 1 point, Loss = 0 points. The match rating is the home team’s points minus the away team’s points .

Goal Superiority Rating System

The next rating method is based on the goal superiority rating system mentioned in Football-Data’s FREE PDF guide to Rating Systems for Match Prediction. This one also looks at the last 6 league games. In this method a team’s ratings is calculated as their total goals scored minus their total goals conceded.

The match rating is then the Home team’s rating minus the away team’s rating. Football data’s article goes on to describe how to calculate a home win – draw – away win probability distribution and true odds for the three outcomes.

Elo Rating System

The Elo rating method is a bit more involved and you can read more background information here: Elo Rating System. I’ll go into a little more detail on how I use the Elo method in my next post but at the moment I feel it is a bit more accurate than the other two methods for predicting match outcomes.

The example results table below highlights last weekend’s Elo match ratings where the top 11 games all went to form as home wins. However, most weeks this is unlikely to happen as some games won’t go to form.

Highlight Key

= Home = Away = Draw
Elo Match ratings for games Fri 4th Apr 2014 to Mon 7th Apr 2014
Date Div Home Score Away Rating
05/04/14 E0 Man City 4 : 1 Southampton 279
05/04/14 E0 Chelsea 3 : 0 Stoke 252
05/04/14 EC Luton 1 : 0 Aldershot 235
05/04/14 SC3 Peterhead 4 : 0 Montrose 220
07/04/14 E0 Tottenham 5 : 1 Sunderland 219
05/04/14 EC Gateshead 5 : 0 Tamworth 210
05/04/14 E2 Brentford 3 : 1 Notts County 190
04/04/14 E1 Leicester 2 : 1 Sheffield Weds 186
05/04/14 EC Chester 3 : 2 Hyde United 184
05/04/14 EC Cambridge 3 : 1 Southport 166
05/04/14 E1 Wigan 1 : 0 Leeds 159
05/04/14 E1 Nott'm Forest 1 : 2 Millwall 158
05/04/14 E2 Wolves 2 : 0 Peterboro 155
05/04/14 SC1 Falkirk 5 : 0 Cowdenbeath 143
05/04/14 E3 Chesterfield 1 : 1 Newport County 136
Full list for: Matches between: Fri 4th Apr 2014 and Mon 7th Apr 2014

Categories
Football Rating Systems

Getting Started on a Football Rating Strategy

So I finally get round to starting this blog. Why am I doing this? Well there is the fun element of having a bet on football (and a picking winner) but also I’m revisiting my interest in analysing football data. This is something I enjoy and started to do in  the mid 90s using soccer data-sets and before that it was helping my dad choose his v-plan 81 selections for the pools.

However, I am not a statistician or betting expert but a software developer that hopes to learn more about betting on football in general, staking plans, value and finding an edge.

What is Since a Win?

The domain is a reference to the football sequence tables and recent form runs that appeared in newspapers prior to upcoming football fixtures. These tables would list football teams in descending order of how long since a particular result i.e. since a win, since a draw etc.

At the moment I am importing English and Scottish league results from Football Data for analysis and as a starter have created some simple rating methods based on articles found online. Football-data.co.uk has been around a while and is an excellent resource for footy data if your into analysing and creating your own football betting systems. Originally I used to pay a tenner to access the data files but they have been free to download since 2007.

Most of the time I shall be coding or analysing rather than writing but when I find something of interest I’ll share it.