Introduction

World Club Football Elo

An Elo rating is a method originally used for games like Chess and Go to rank players.  My understanding of the system is that there are two main features;  there are a fixed number of points within a system and the number of points transferred between players depends on the likelihood or  probability of a win for either player.

So if a weaker player wins against a stronger player, the weaker player would take more points from the stronger player.  If the stronger player won, less points would be taken.  Given that point are taken, the overall points within the game should remain constant.

There have been attempts to apply Elo to football, most successfully within the national game.  Outside of the major football organisations like FIFA, Elo ranking has respect.

Application to Club football seems to be limited, the best attempt I can find is the www.clubelo.com site.  this only deals with European football and even then only goes back to 1950s with the start of European football.  Nevertheless, it is a very good site, with some interesting data and some brilliant arguments for method.

My aim is to go right back to early football and try to include results from all over the world. Ambitious eh ;)

Method

I've applied the Elo rating pretty much as described in:

http://en.wikipedia.org/wiki/World_Football_Elo_Ratings
http://www.eloratings.net/system.html
http://clubelo.com/Articles/RatingSystem.html

What I have done differently is the starting point, home advantage and weighting.

Starting Point
For the Elo systems above, a new team will be given a starting rating of some arbitrary value.    The international teams seem to have a starting value that ranges from 1000-ish to about 1800.  ClubElo has to content with teams being relegated and then returning to competition.  I aim to overcome this by tracking all divisions (see my paragraph on 'official competitions')

Therefore I have set all new teams a start of 0.000.  This means some teams will go negative.  I expect a poor run of results to result in a negative rating.  I also expect most teams who fold to have a negative value, therefore the average points of active teams should steadily increase.  We will see over time if this is a good decision.

Weighting
EloRatings uses a system of weightings depending on the competition.  For the FIFA rankings there is also a weighting system.
It seems that ClubElo however has done analysis that shows a weighting value of 20 gives the best prediction for the next match.  Although I agree that different competitions or even games (cup finals etc) should have a different weighting, I will use the value of 20 for the time being.  When I have more data, this may be something to look at.

Home advantage
EloRatings simply applies a factor of 100 for home advantage.  However ClubElo has analysis that home advantage can change over time.

For my analysis, I have taken the last 10 home games to apply a factor of between 50 and 100.  So if a team is on a good run of form at home, and has won all 10, they would have a factor of 100.  A new team or poorer team would have a factor of 50.  A win will contribute 5 to the factor while a draw will contribute 2.5.  Again, we will see how this is working out over time.

For the moment, I'm using an Excel sheet set up to do all the calculations and produce tables and graphs.  I can use Excel and am pretty nifty with formulae.  I do accept that as this becomes bigger I will need a database.  I'll cross that when it comes to it.

Official Competitions

My aim of this is to have as many football matches contribute to a clubs Elo rating.  The first recognised competition using current rules is the English FA Cup in 1971.  Marlow and Maidenhead played in this and according to Wikipedia, have only missed 1 competition each since.  Therefore they should have a lot of games to create a rating.  How far do I go though?  There were many friendlies at the time (there still are really) but a lot of results, particularly from that era disputed, unknown etc, so which games should count towards a teams Elo?  With there being fewer games, international ratings use friendlies but gives them a lower weighting.

My view is to take this as it comes but with the following basic rules of what is an 'official competition':

• Each country has a major cup that most teams from that country can enter
• The top, national division should count as official
• Any leagues below that have direct promotion/relegation are included

That is the plan anyway, the first stage is to get all results from 1871 to 1888 in.  This will comprise of English FA Cup, Scottish Cup, Welsh Cup & Irish Cup.  I'll do a blog when those results are in, reviewing the findings, and the methods up to that point.