Sunday 30 August 2009

Nerdy Anorak Elo Stuff


The football Elo ratings are starting to settle down. I have three numbers that I am using. The Elo rating, the current form rating, and the power rating.

The Elo ratings are adjusted after each match, and in many ways are like the 200-day moving average used in stocks and options trading. On the plus side is that they give a good idea of the base strength of the team, but this is offset by the slowness with which they react to recent form, which is why I started using a couple more parameters.

The second is to account for a team's current form (over the last six games) while the third, a power rating, gives an overall reflection of each team's performance over the past six games, and places less emphasis on goals and more on domination. Because goals are such a relative rarity in football, a team can lose 0-1 but dominate the game, and this power rating is to help smooth out such anomalies. A good example is yesterday’s Manchester United v Arsenal game where there was little between the two teams, and a subjective decision to award a penalty arguably turned the game.

I have been converting these numbers into probabilities and comparing them with the odds available on the exchanges.

It’s hard to draw any conclusions from the results to date, but as we are at a mini-break, I’ll try.

Laying the draw would have been the most lucrative strategy to date. (Hardly surprising given the remarkable lack of draws so far – 3/36).

Here is how the other numbers stack up. (Results are to a level stake of 1 point and exclude commission. Prices used were the best available on the exchanges as close to kick-off as was practical).

Using the base Elo ranking, backing the top rated team would result in a profit of +7.58 (+21.06%).

Backing the team with best Form: +20.22 (+56.17%)

Backing the top rated Power team: +3.91 (+10.86%)

Although the form rating reaches back to last season, and thus is of limited value until we have a couple more games in the bag, this is the number that shows most promise. However, the power rating is only just getting started so the jury is out on this.

Anorak Section

As for the season so far, my ratings now have Bolton(19th in table), Portsmouth(20) and Hull(14) as the bottom three teams (in bottom up order). I’m not sure why no one seems to be talking about relegation for Bolton, but they are clearly not very good. At the top are Manchester United(3), Chelsea(1) and Liverpool(7).

In the Championship, the top three are Cardiff(2), Preston(6) and Sheffield United(7), with the relegation spots filled by Plymouth(21), Scunthorpe(20) and Crystal Palace(19).

League One: Leeds(2), MK Dons(3) and Charlton(1). Relegation – Wycombe(20), Stockport(14), Yeovil(12), Hartlepool(21).

League Two: Rotherham(2), Bournemouth(3), Shrewsbury(11), Dagenham(1). Relegation – Grimsby(22), Hereford(20), Macclesfield(19), Notts County(5).

Conference National: Oxford(1), Stevenage(11), Luton(3). Relegation – Forest Green(23), Ebbsfleet(22), Gateshead(21), Chester(24).

Conference North: Droylesden(1), Southport(3), Alfreton(9)

Conference South: Hampton & Richmond(11), Eastleigh(15), Newport County(4) / Woking(3).

Scottish Prem: Rangers(1), Celtic(2).

One: Queen of the South(2), Dundee(4).

Two: Stirling Albion(2), Alloa(1).

Three: Livingston(7), Forfar(1).

1 comment:

Anonymous said...

I'm sorry these football ratings systems don't work. One-size-fits-all constants for short and long term form/ home team advantage don't work. Rating have no consideration of team line-ups or team tactics , they might pick up the odd rick in League 2, but the clued up football punter will always trump them and pick up any value. Do yourself a favour Cassini and spend your time on something more fruitful. If you do make it pay I'll give you the money myself...