I recently mentioned that I was developing an Elo based football ratings system. I'm rating all the English teams down to the Conference North and South and all the Scottish League teams.
JPG commented that I should be "aware that the accuracy only comes into place after around 30 matches (for each team) have been rated". While I was aware that the ratings would take a while to settle down, I honestly hadn't considered that it would be February before I could start making my fortune. I have kids! I need money for Xmas! Thanks a lot JPG!
Seriously, the comment was appreciated, and as with any system, I shall be 'paper trading' this one first, or perhaps placing small 'live' bets as I did last Saturday, and it will be interesting to see how much the accuracy improves from now until the 30 games.
I would like to know where this value of 30 games comes from? I was thinking of trying it with American Football, but waiting for two years before it settles down is not appealing.
Anyway, with the lack of games tomorrow that I will be betting on, I thought I'd look into one game that Talkbet has mentioned on his blog - http://talkbet.typepad.com/talkbet/2008/10/october-10th-update-i.html - where he says to back Salisbury City at home to Kettering Town at 2.88+.
The game caught my eye, not only because Kettering have been my non-League team for many years going back to the 80s, but because Salisbury have been in something of a slump over the last five games (recent form is taken into consideration as well as the overall rating).
I ran the numbers and have Salisbury at 3.25 and Kettering at 2.49 and at the prices currently available on the exchanges, the value here in my opinion is with a lay of Salisbury. But only a small lay.
Come on the Poppies!
Update: A pleasing come from behind win for the Poppies, who stay second in the Conference, but more importantly remain atop my own personal rankings for their division as do Chelsea, Reading, Leicester City, Rotherham, Tamworth, Team Bath, Celtic, Dunfermline, Brechin City and Stenhousemuir.
About Me
- Cassini
- I have had a life-long interest in sports and after studying Pure Mathematics with Statistics at secondary school, have been fascinated by odds and probability. The first system I came up with was a simple one - back the favourite and double up after a loss until a winner. Simple enough in theory, and I told my Dad about it. Not being a betting man himself, he ran it by some of his colleagues, and came home to tell me that it wouldn’t work because a long losing run would mean that the bank would be empty. Then there was always the possibility that the winner would be returned at odds-on, meaning that the total returns would not match the outlay. Not what a ten year old wants to hear! Only slightly daunted, I then went on a search for the Holy Grail, the secret to riches that I knew was out there somewhere. Finally in 2004 I stumbled across an article about Betting Exchanges and four years later I was able to make a steady profit.
Nice call! Any thought on the Eastbourne game tomorrow ?
ReplyDeleteCassini, not all is lost regarding ELO! I made a follow up comment about it under the "Sunday Value" thread before I read this one. Im not sure of the mathematical reasons but I can confirm from experience that the 30 match figure is a noteable marker. Remember, each game rated is another step towards accuracy, whether its the 1st game or the 1000th, just that each game makes less of a shift towards accuracy then the previous one. By the way, Id be very interested to hear how youve applied ELO to US Football as Ive been trying without success so far on this. (I dont mean the betting on the ratings, but the actual compiling of the ratings.) Cheers - JPG
ReplyDeleteJPG - For American Football I rate the teams so that the difference between then (with an adjustment for home advantage) represents the expected points difference. I compare that with the handicap and attempt to find value. E.g. today: Saints (25) v Raiders (16) - I expect the Saints to win by 9 so Saints -7.5 at 2.16 is to my mind great value.
ReplyDelete