How many hours does it take to establish a true win rate? B&M only.

[A_C_Slater]

You take amount of money won dividied by the number of hours you played to get your win rate, right?


How many hours would it require of B&M play before you could claim a "true" winrate.

Assume that you are staying at one limit and not moving up or down. This is for limit hold em.


Also, the poker room has those automatic shufflemasters.

[Djogolf]

My "guess"...

150,000 hands...

30 hands per hour...(This number is more dependant on level than shufflemaster, IMO. The higher the level, the more hands you will see per hour, because there are less players seeing the flop, and more experience w/ regards to action, and the "oh, is it my turn?" sort of stuff)

30 hands per hour = 5,000 total hours...
40 hours per week = 125 weeks...
Works 50 weeks per year...
so...2.5 years playing 40/week, 50 weeks/year

--Golf

[A_C_Slater]

Thank you.


But why do you pick 150,000 hands?

What is so special about that number?

[Djogolf]

[ QUOTE ]
why do you pick 150,000 hands?


[/ QUOTE ]

No specific reason. Mainly I tried to pick a middling sort of number that will give you a fairly good idea.

My thinking was that 100,000 may not be enough, and 200,000 should be plenty, so I split the difference. There was no "mathematical" reasoning for it.

Hopefully others that have more experience w/ SD, variance, probabability, ect., ect., can provide you with a more concrete response, with some specific SD/Var. numbers to back it up.

--Golf

[jtr]

If you want someone to answer this question, you need to put a number on what you mean by a "true win rate". The truth about your win rate is, well, elusive. But if you play enough hands we can give you an estimate -- you tell me what level of confidence you're happy with, and what margin of error.

So, an answerable version of the question runs "How many hours of B&M do I have to play before I there is only a 1% probability that my real win rate is any further than 0.1 BB/hr from the estimate?"

If you like that version of the question, then the answer (assuming a conservative standard deviation of 12 BB/100) is 9,585,200 hands (about 159 years of full-time B&M play). So you can see there's a problem getting to the long run.

If you give up some precision, and are happy just to know your win rate within plus or minus 1BB, but you insist on the same 99% level of confidence that your true win rate is within that range of values, then you can get there in only 95,852 hands (100 times faster, but still a year and a half of full-time play).

[dogmeat]

I had some spikes playing B&M, but after about 500 hours my overall numbers varied only about 10% from month to month.

Dogmeat /images/graemlins/spade.gif

[Rudbaeck]

One amazing number of hours. Let's say you win 1BB/hr and you have a standard deviation of 12BB/hr after 2,000 hours. (This is one year of 40 hours a week, 50 hours a year at the tables!)

Now we are 95% certain you are winning between 0.5BB/hr and 1.5BB/hr.

You can calculate the confidence interval yourself. First you calculate the standard error by this forumla: SE=SD/sqrt(population size)

In this case we get SE=12/sqrt(2000)=12/44.72=0.268

We can be 95% certain that our true win rate is within 1.96*SE of our measured win rate. 1.96*0.268=0.525, so lets just round that to 0.5