Sunday, 8 February 2009

February Football

February football is off to a profitable start, with my ratings producing six winners from seven games in the Premier League yesterday. The only loser was Chelsea who were expected to beat Hull, but at 1.31, I didn’t back them anyway.

Having given my Elo based ratings a few months to settle down, they are still a work in progress, but I am starting to use them with a little more confidence. Early results seem to suggest that the value is when my ratings predict a lay of the home team. Aston Villa, Newcastle United and Fulham all obliged.

For the record, Arsenal and Manchester United are not surprisingly rated to follow suit this afternoon.

I am still researching the best way to use the ratings. For example, I picked the Wigan - Fulham game as a draw based simply on the fact that the two teams were rated so closely (1168 and 1161 respectively), and the draw 'seemed' to be a value bet. I'm looking for something a little more scientific.

Anorak Section

As for the ratings, Manchester United (2456) are far and away the top rated team in the Premier League, ahead of Liverpool (1966), Aston Villa (1788), Arsenal (1721) and Chelsea (1581).

The other top ranked teams in the leagues I track are currently Cardiff City, Leicester City, Darlington, Burton Albion, and AFCs Telford and Wimbledon. In Scotland I have Rangers, St. Johnstone, Raith Rovers and Cowdenbeath.

From table-toppers to table-proppers I have West Bromwich Albion, Charlton Athletic, Cheltenham Town, Chester City (not a good time to be a team starting with ‘Ch’ apparently), Lewes, Solihull Moors, Fisher Athletic, Inverness Caledonian Thistle, Queen of the South, Stranraer and Elgin City.

From now to the end of the season will give me four months to see whether the effort is worthwhile. If it is, then I may start rating some of the overseas leagues as well.

It may be more profitable to use these ratings for promotion / relegation markets – it’s something I’ll be tracking.


Anonymous said...

link exchange?

let me know

Cassini said...

Link added...

Anonymous said...


Keep it up with the ELO stats mate although I must say, the points differential between your top teams are lot wider then mine! Hopefully (for both of us!) you're simply using a bigger scale then I do.

Like I said before, once teams have had 30 games rated, you'll start seeing a lot more accuracy within them.

The trick though is to carry your final ratings this season into the next one when it starts. This sounds odd to most as teams have transfers etc and can change. This is true but the beauty part is that your "starting figures" for next season are more accurate then simply starting all over again and youll find that only about 10 or 12 games are needed before they reflect reality a lot more.

Dont forget, every time a game is rated, your entire set of ratings will get slightly more accurate then before.


Cassini said...

Hi JPG - I started my Prem teams off on an average of 1200 points each, more for the top teams based on last season's finishing positions. I was planning on rolling the numbers forward to next season, and I have high hopes for them! My big problemo is how to convert the values to odds, especially with the stupid draw. NBA and NFL are so much easier!

Anonymous said...

Yeah, you're right, the draw element is tricky. However, what you can do quite easily is convert your ratings into a Draw No Bet price. Once you have that, its easier (although not necessarily as scientific) to convert it all into a 1x2 format.

Regarding yuor ratings, Im still a little surprised at the disparity in them (scale is probably the factor.) To use your current ratings of Man U (2456) and Liverpool (1966), if say Liverpool won 1-0 away at Man Utd in their next match, what would both teams ratings be? My calculations show that they would be Man U (2408) and Liverpool (2014).

Is this the case?


Cassini said...

If Liverpool won at Manchester United, then United's rating would go down 7% to 2371, and Liverpool's would increase to 2144. My formula is for the home team to 'risk' 7% of their rating, the away team 5%, and in the even of a draw they share the spoils. It was something I first read about in a book called "The Punter's Revenge" many years ago, and those %ages were the ones that the authors claimed to work best. Time will tell.

Anonymous said...

Cassini, that makes a lot more sense! You actually compute your ratings in a different way to myself (or it appears to) but I can tell almost instantly that your weightings for home and away etc roughly tie in with the way I do it.

Youll notice that my ratings moved a lot less then yours given the example of a Liverpool win yet the gaps in your ratings are lot more. In short, your using a different scale to my own (plus a slightly different way of calculating) which is fine - was just curious to see if either of us had got the wrong end of the stick!


Cassini said...

I'm sure my method could be improved. For example, the ratings would be the same of Liverpool won 1-0 or 8-0. I did toy with adjusting slightly based on margin of victory up to 3 goals, but decided it wasn't worth the effort, and a team losing 1-0 that conceded a late second goal while pushing for an equaliser would be unfairly punished. I also feel that the losing team should perhaps not lose the same flat percentage regardless of who they lost to. Like I've said, it's a work in progress.