Quote:
Originally Posted by dude45
I figured as much but how many samples of that exact situation would I need ? The main reason I learned preflop charts and study pokersnowie and now zenith is because someone else has put on the leg work some might call that lazy or cheating but I could play for the rest of my life and not get a large enough sample of say 3betting AQ In the BB versus a CO 2 bet and a button call
Someone with better stats knowledge than me might point out where I'm wrong or have a better method, but I would get an estimate of the standard deviation out of pokertracker and calculate the size of the sample required to reach a desired confidence interval.
Since your sample is small and we only want an estimate you could group similar hands to get the SD stat for that group when you open. If it's 500BB/100, for example (no idea if this is close to what it would actually be), and you observe a win-rate of 10BB/100 with the hand that you want to know about, the 1 standard deviation confidence interval takes 250,000 hands to get to +/- 10BB/100.
1*500/sqrt(n) = +/-10
n = 250,000
If the you are looking at something like a suited connector UTG that's 4 combos out of 1326 and 1 position out of six, so your total sample would need to be 250,000*6*1326/4 = 497,250,000 hands to be about 85% confident that your open is profitable if you observe a win-rate of .1BB on average when you open that hand.
You could grind 24 tables full time every day for your entire life and still not confidently know whether 98s is a profitable open from the LJ. And over the course of that amount of time the games will likely be totally different and the sample you modeled after won't really be relevant anymore. There are probably methods to analyze this spot better, like looking at similar hands and extrapolating, but you need a lot of hands.
There might be a better way, but the point is that a hand on the cusp you will just never know, and a hand not on the cusp is clear enough that you don't need statistics to know it's a profitable open. I would just use charts as a baseline and adjust based on game conditions rather than try to use database analysis on specific combinations of hands and spots. Database analysis needs to be more general.