What's a decent enough sample size for SNGs/STTs, 500? 1000?
Thanks
Full Version: Question About Sample Size
Depends really. I think I read somewhere you're mostly playing turbos? Maybe I'm making that up... anyway, if you are, a few thousand games probably gives a good idea of where you're at. If you're playing non-turbos where the variance is lower, a thousand games is probably enough. To get a really accurate picture probably more liked double those figures though. I've played about 800 turbos in the last few weeks and I'm still at the point where a few hours playing can change my roi by 5%.
EDIT: This was a 30 buyin downswing in 60 games, which is slightly... abnormal (disgusting), but you get my point.
In short: lots
EDIT: This was a 30 buyin downswing in 60 games, which is slightly... abnormal (disgusting), but you get my point.
In short: lots
Also, it depends how accurate you want to be. Calculating your confidence intervals for winrate/ROI is a good way of showing your ROI within certain percentages of accuracy.
I've not done that for ages and my math skills are rusty at best, so maybe a mathsy guy can come help with that.
I've not done that for ages and my math skills are rusty at best, so maybe a mathsy guy can come help with that.
QUOTE (cheesies @ Wednesday, March 26th, 2008, 7:48 PM) 
Also, it depends how accurate you want to be. Calculating your confidence intervals for winrate/ROI is a good way of showing your ROI within certain percentages of accuracy.
I've not done that for ages and my math skills are rusty at best, so maybe a mathsy guy can come help with that.
I've not done that for ages and my math skills are rusty at best, so maybe a mathsy guy can come help with that.
*cough*Simo*cough
Imo, your sample size shouldn't be that big of an issue because any large enough sample size is going to take a long time and your play at the begining of the sample is going to differ from your play at the end and in the middle so you're still not getting a true picture imo.
I don't think you should worry about it, we all know variance is a big factor, but determining just how big is pretty impossible (i may be wrong on this), and your past results shouldn't effect you, you should still allways try and learn/improve whatever your results as long as you play poker. I used to look at my ROI% a lot, but i think it's a bad habbit and something you shouldn't focus too much on.
QUOTE (Sheiky @ Thursday, March 27th, 2008, 7:06 AM)
*cough*Simo*cough
Imo, your sample size shouldn't be that big of an issue because any large enough sample size is going to take a long time and your play at the begining of the sample is going to differ from your play at the end and in the middle so you're still not getting a true picture imo.
I don't think you should worry about it, we all know variance is a big factor, but determining just how big is pretty impossible (i may be wrong on this), and your past results shouldn't effect you, you should still allways try and learn/improve whatever your results as long as you play poker. I used to look at my ROI% a lot, but i think it's a bad habbit and something you shouldn't focus too much on.
Imo, your sample size shouldn't be that big of an issue because any large enough sample size is going to take a long time and your play at the begining of the sample is going to differ from your play at the end and in the middle so you're still not getting a true picture imo.
I don't think you should worry about it, we all know variance is a big factor, but determining just how big is pretty impossible (i may be wrong on this), and your past results shouldn't effect you, you should still allways try and learn/improve whatever your results as long as you play poker. I used to look at my ROI% a lot, but i think it's a bad habbit and something you shouldn't focus too much on.
If youre going to try and make a living on STTs then you need this information to properly decide between buy in levels and bankroll size. It is a random sample so fluctuations over time shouldnt be that great except for improvement, which will tend to understate your "true" ROI and SD, which isnt a bad thing.
I dont think it needs that many tourneys though. ROI is equivalent to finish position as long as the stakes are constant, so youre only looking at the distrbution of 4 results...1,2,3 and >= 4. With the varianceof the average finish position being reduced by the square root of the number of trials I would think 400-600 should give a pretty accurate sample mean and variance.
QUOTE (copernicus @ Thursday, March 27th, 2008, 4:34 PM)
If youre going to try and make a living on STTs then you need this information to properly decide between buy in levels and bankroll size. It is a random sample so fluctuations over time shouldnt be that great except for improvement, which will tend to understate your "true" ROI and SD, which isnt a bad thing.
I dont think it needs that many tourneys though. ROI is equivalent to finish position as long as the stakes are constant, so youre only looking at the distrbution of 4 results...1,2,3 and >= 4. With the varianceof the average finish position being reduced by the square root of the number of trials I would think 400-600 should give a pretty accurate sample mean and variance.
I dont think it needs that many tourneys though. ROI is equivalent to finish position as long as the stakes are constant, so youre only looking at the distrbution of 4 results...1,2,3 and >= 4. With the varianceof the average finish position being reduced by the square root of the number of trials I would think 400-600 should give a pretty accurate sample mean and variance.
I just read this in the mathematics of poker so i vaguely get what you're talking about
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.