how many hours is a good sample size in live poker? - Gambling and Probability

Two Plus Two Forums Other Topics Probability

how many hours is a good sample size in live poker?

Post Reply Subscribe

...

07-18-2018 , 12:15 PM

josofo

grinder

Join Date: Oct 2009 Posts: 653

how many hours is a good sample size to determine your win rate?

Last edited by josofo; 07-18-2018 at 12:21 PM.

Quote

07-18-2018 , 01:19 PM

pucmo

adept

Join Date: Mar 2016 Posts: 1,187

One hour, or are you blind?

Quote

07-18-2018 , 03:53 PM

josofo

grinder

Join Date: Oct 2009 Posts: 653

Quote:

Originally Posted by pucmo

One hour, or are you blind?

but variance

Quote

07-18-2018 , 04:52 PM

browni3141

Pooh-Bah

Join Date: Aug 2015 Posts: 5,678

It is not possible to accurately determine your win-rate in live poker using only statistical methods. By the time you've played enough hours to narrow the confidence interval on your win-rate the game conditions will have changed.

Quote

07-20-2018 , 02:30 PM

lawdude

Carpal \'Tunnel

Join Date: May 2009 Posts: 10,305

Quote:

Originally Posted by browni3141

This is formally correct.

However, it is nonetheless true that you can at least get an idea as to whether you are a winning or losing or close to break-even player, with some confidence, if you play enough hours.

Online, 35,000 hands was a good guideline, which translates to about 1000 hours of live limit play and 1400 hours of live no limit play.

Basically if you play full time for 3/4 of a year in a particular cash game, you should have some idea by the end.

Quote

07-20-2018 , 10:22 PM

browni3141

Pooh-Bah

Join Date: Aug 2015 Posts: 5,678

Quote:

Originally Posted by lawdude

Right, but I would add that the sample size required to determine whether a player is a winner or loser is much smaller live, as the edge is much bigger. It's definitely achievable to do this before game conditions significantly change. However, the sample size required to know your win-rate, let's say accurately enough to judge whom between several players is best, is much larger and not achievable before game conditions change significantly.

Quote

07-29-2018 , 02:19 AM

leavesofliberty

Carpal \'Tunnel

Join Date: Jul 2010 Posts: 6,296

Agree with comments about sample size. It should be understood that you should be learning as you play so even over a year of solid play conceivably you've improved your game. Your opponents are also conceivably learning too. You can get a picture of what's happening, but you still couldn't really draw a conclusion. If the data says you lost for the first nine months, and won the last three months, it doesn't necessarily mean you're a losing player and may indicate that you're learning how to play.

Quote

07-30-2018 , 10:15 AM

festivegorgon

newbie

Join Date: Oct 2017 Posts: 25

it really depends on signal to noise.. you might not realize how extreme this effect is..

example:

assume std deviation is 200BB/hr

2.5BB/hr win rate.. you'd want 67,000 hours to detect this

25BB/hr win rate.. you'd want 670 hours to detect this

Quote

Post Reply Subscribe

...