Sample size and age of data

[Paparazzo]

Dear all, Just an open question for an odds' statistical based model: How many matches would you consider to have good conclusions? To me even 50 matches would be enough, ideally more than 100. Due to increasing payouts during the years, how far would you go in order to catch data? In my opinion any data prior 2007 would be misleading... I mean that a 2,00 given in 2000 might be a 2,10 today... Thanks for your replies!

[Valiant Thor]

Re: Sample size and age of data Betting tax was abolished for the punter in 2001 so the prices actually contracted as books were responsible for the tax and therefore made their overound bigger to cover the cost. So any odds based stat model should be focused on post 2001 data to avoid statistical 'noise' from earlier odds. As for what size of sample data should be used ,you could use a simple benchmark 'Margin of Error' formula for your data. An acceptable size is around 95% (+/- 0.025%) MoE (or confidence interval) should give you a ballpark figure. Sample Size = (1/.025)^2 =40^2 =1600 samples So if you had a sample size of say 2500 you would have a MoR of around... MoR=1/?2500 = 1/50 = .02% =+/-.02 = .04% MoR Or better still if your data is in excel and it contains varying odds use the Confidence formula in there which when you +/- the figure from the mean of your sample data will give you the range to 95% confidence that your results will fall in between that range (that is if you set the alpha to .05, of course). Then you could do a monte carlo simulation using your strike rate 'v' estimated range over 1000's of random simulated bets to check profitability over a greater timescale. hope that is of some help :ok