Create Own Ratings and Kelly Staking

[SniffTheGlove]

I am looking at creating ratings for all the teams in my database. I have looked around this board and got quite a few ideas, so starting off simple I am using the details from Jenspm's Thread http://www.punterslounge.com/forum/f21/how-beat-bookies-value-betting-guide-inc-spreadsheet-102626/ to create the ratings then using Kelly Criterion for the staking. I have created my own initial test spreadsheet to make sure all the maths and equations work before I code it all into my software. Could people look at my spreadsheet and tell me if it's good, stable or bad with bad maths and also ways to improve the model. You only need to change the data in the yellow cells, all other cells have formula's based on the yellow cell data. The data is taken from a match played on 20th of Jan 2010 (Arsenal - Bolton), the match ended up 4-1 to Arsenal. The odds used are from Bet365 taken from my footbal-data data. The ratings are based per said Jenspm "How to beat the bookies – a value betting guide" thread I just want to know it's all correct before making a fool out of myself. Thanks

[Shinigami]

Re: Create Own Ratings and Kelly Staking Personally, i do not think it's wise to use the last 'X" number of games as calculation of value. i don't find it a reliable form of measure. Recent form and winning rates can change dramatically in a short time, but usually ability to score change slowly. Reversion to the mean is almost a law of nature with respect to team performance. More often than not, teams whose recent performance has been poor tend to perk up and improve. Whereas, a team who has been on a hot streak in the last "x" number of games may just falter. Besides, when you're only drawing recent sample from the last "X" number of games, it's a form of "recency bias" - putting emphasis in recent wins/losses.