# How to calculate my sample size to establish my real-WR?

I've been tracking for a period my hour rate and I want to know how much data I've to collect in order to establish my winnings per hour. I've been searching online and when it comes to getting a handle on real-WR potential 50k hands (1600 hours) is a starting point, 100k (3200 hours) is pretty good, and 500k+ (16200 hours+) is solid. But I'm not sure about this and I would like to calculate it myself. Any help?

• rule of thumb is the more hands you have played the more accurate your real-WR calculation will be. You can calculate it with any number of hands but smaller sample sizes will be less accurate. The examples you gave in the question seem good to me, i am not sure if there is a calculation that will tell you how many hands is enough. Commented Jan 25, 2020 at 19:49
• Normally there should be some way to calculate the sample size, some that should be representative enough to get my real-WR but it might require advance statics knowledge that I certainly don't have yet. Thanks anyway! Commented Jan 27, 2020 at 13:38
• this may be a better question for mathematics or statistics stack exchange, I would be interested to hear what they have to say. Commented Jan 27, 2020 at 20:03
• I agree. I'll post it on my mathematics account to see what they answer! Commented Jan 29, 2020 at 11:27

The problem I see here is that if you are starting to become a professional player, your skills should be improving, thus I think you should track your performance based on shorter period of time, say per month or per three month. You can then test if the increase of your hourly rate in each period is statistically significant (assuming the increase is a constant).

• Not bad idea. I think that there I'll be able to conclude if there's a periodical improvement in terms of hourly rate but this measurement won't be conclusive enough to give me information about my real-WR, consequently won't tell if I'm a profitable poker player or I'm not. Commented Jan 29, 2020 at 14:47
• If your focus is to have a conclusion that you are a profitable player, then the problem is "if your average WR is significantly greater than 0". Then you should do a one-tailed t-test assuming your WR is normally distributed. For this, to get a significant result (at 5% or even 1%), I don't think you need so many samples. Assuming one hour is a sample, then something like a hundred should have enough statistical power to detect the difference.
– Z Gu
Commented Jan 29, 2020 at 17:28
• To do statistical inference, you have to first have a hypothesis, say "my WR is 50 euro/hour", then you can do test and draw conclusion just as I described. Simply averaging your data gives you no conclusion, only an estimate, regardless of your sample size
– Z Gu
Commented Jan 29, 2020 at 17:33
• lol this is why real professional poker players almost NEVER track their results. A pro who "keeps track of his hourly" as if it means something, is a guy who will be on the rail in 14 months. Not only should your own personal game change, but the competition ALWAYS changes. Stats are useless except to determine if you're a winning player or not. It's like trying to figure out what your win rate is at being elected Governor or how likely you are to get married. Your poker records, all of them, represent ONE DATA POINT TOGETHER. Commented Jan 30, 2020 at 22:49