Thursday, 15 July 2010

Weighting For Elo

Anonymous from London has a thought:

When trumpeting ELO and the fact most of these statistical models use ELO, have a think:

If ELO is so widely used then the market-makers (vague term I know, not time to be more specific) will of course be aware of it.

If ELO is good then the market-makers will use it.

If that is the case then you have no edge from using ELO - you need a different approach.

Or, you need to use an ELO rating system which is superior to those used by others.

Typed this in a bit of a rush, may not quite make sense so feel free to pull apart.
Agreed, it would indeed be a big shock if market-makers were unaware of Elo ratings, and it is true that if all Elo rating systems were the same, then it would not be possible to gain an edge using them, but the key is that not all Elo ratings systems are the same.

The basic Elo algorithm is fairly simple. The complexity comes in to play when you start adding variables, for example how much weighting to give a certain competition, how much allowance do you award for home advantage, what differences do you include to account for the different ‘personalities’ of leagues, how do you reward a one goal win versus a five goal win, do you accept the result alone or do you factor in other statistics such as shots on goal or corners, is a 2-0 win with the second goal in stoppage time the same as a 2-0 victory sealed 20 minutes earlier etc.?

Everyone has different ideas on these. The data for football these days is available to all of us and while I do not know if the bookmakers use Elo, another method, or a combination of methods (most likely) to price up matches, given that differences exist between the prices offered by individual bookmakers and on the exchanges, it is fair to assume that similar, but nevertheless different, pricing models are used.

And where differences exist, opportunities exist.

All rating models are likely to have their strengths and weaknesses. For example, it was apparent last season that, for whatever reason, the selection of draws in my Elo based system worked very well in the English Premier League, not so well in other leagues.

So it’s worth remembering that a big advantage punters have versus bookmakers is that the former always have the choice of whether to bet or not. We have the luxury of being patient and using a sub-set of the predictions if and when we determine that we have an edge.

It would be highly unlikely that an Elo based system would consistently be more accurate across all leagues than those used by bookmakers, but the wealth of data available and modern day computing power means that at least the playing field is relatively level, and that a good rating system, selectively applied, has the potential to be profitable.

Finally, a small point perhaps, but it is Elo, not ELO. Elo is not an initialism. Elo ratings are named after Arpad Elo, a chess master who come up with the basic idea for the United States Chess Federation to improve upon the Harkness rating system they formerly employed.


Anonymous said...

You may be interested in a book by Arpad Elo "The ratings of chess players". (1978).

Anonymous said...

None of the top professionals use ELO ratings. You are heading down the wrong path especially at the start of a new season.

Laurie Wakeham said...

Can you tell me how I can add shots:shots on target:goals to my Elo rating model? I stumped. :(
I'd be most grateful

Laurie Wakeham said...

Can you tell me how i can add shots:shots on target:goals to my Elo model?
I'd be most grateful.