Good or Lucky? PT stat question

[MAxx]

How many hands do you need to evaluate your play with confidence at a particular low limit holdem level. For example for the BB/100 stat. Would you say that you can draw solid conclusions from this stat after 500, 5000, 25000, or how many hands? After how many hands would you be confident to draw conclusions about your rate because of your play, and not due to the cards running over you in the shortterm?

[astroglide]

100k. after 300k i think you can be sure.

[EWillers]

I know the math guys can answer this much more precisely, but. . .

100K minimum I think is quite a bit too high. It depends, of course, on how accurate you want your rate to be. But in the time it takes to accumulate 100k hands (unless you're playing 2 or 3k a day) I imagine that the "goodness" of the games will have varied enough over the time the hands are played to outweigh any more precision in the winrate created by playing 100k hands vs. say, 20k hands.

I think that it's a scale. The smaller the # of hands, the more off your winrate may be. As the # of hands goes up, the more accurate the winrate tends to become. Beware of comparing hands from 6 months ago to hands played now tho. 1) You may have gotten better (or worse) and 2) (more likely) the games are noticably better or worse now than they were 6 months ago.

Bottom line, in my opinion, 10k or so will give you a pretty good indication and 20-30k will give you solid enough of a win rate to feel comfortable about (assuming the level of play remains relatively constant--big assumptioin).

[astroglide]

10k is not REMOTELY enough. i have played hundreds of thousands of hands and have had many breakeven or losing stretches over periods that long.

[MMMMMM]

I pretty much agree with 100K hands for a dependable reading, just based on B&M. That is about 2 years full time play at B&M and that is about the time it should take for a really good reading IMO. Playing B&M, even very good players sometimes do run significantly good or bad for 6 months or even for a year. Of course more than 100K hands would be better still.

I don't have nearly the number of hands logged in PT that you do, Astro, but it is nice to know that your online stats roughly match my feel for B&M fluctuations.

[balkii]

Hi maxx, if you're trying to decide who to listen, go with the guy with 4K+ posts, not the one with ~70.

Astro's right. Here a way of thinking about how long the long run is, as far as win rates go. After 110K+ hands, I am a net loser with 33. 33 is not a difficult hand to play. You flop a set, and either win a whole bunch or lose a whole bunch. So far its the latter.

Since your overall winrate is made up of the combined winrates of your hands, that cant be accurate until your hand EVs are accurate. I am sure that my 33 EV is inaccurate, thus my overall winrate is inaccurate.

[Thythe]

I agree with astroglide. 100K is the bare minimum to start drawing convincing conclusions.

[BruceZ]

[ QUOTE ]
How many hands do you need to evaluate your play with confidence at a particular low limit holdem level. For example for the BB/100 stat. Would you say that you can draw solid conclusions from this stat after 500, 5000, 25000, or how many hands? After how many hands would you be confident to draw conclusions about your rate because of your play, and not due to the cards running over you in the shortterm?

[/ QUOTE ]

See my post in this thread:

Standard error and confidence intervals (http://archiveserver.twoplustwo.com/showflat.php?Cat=&Board=inet&Number=464228)

[o0mr_bill0o]

is there a standard deviation stat in pokertracker? that makes it oh so easy....

[Yeknom58]

I think your conclusion is correct but your example seems poor.

33 is probably only slightly profitable in the long run and only if you're calling in the right spots/games and getting paid when you flop that big hand. So if you're losing with 33 it could easily be your play not just the numbers.

[steveyz]

Yes, there is. Go to the "session notes" tab, and click on "More Detail..." in the upper right hand corner.

[parachute]

[ QUOTE ]
Hi maxx, if you're trying to decide who to listen, go with the guy with 4K+ posts, not the one with ~70.

[/ QUOTE ]

How many posts do you need to evaluate a 2+2 poster with confidence?

[Louie Landale]

I'm no stat guy. But I'd guess that if you are winning near your fair-share of hands then your BB/100 will be pretty close. With 10 people at the table your "fair share" would be winning 10 of 100 hands. But if you are playing tight you should be winning closer to 7 or 8 (you fold more eventual winners than the loose players, no duh). So if you are winning lets say 7.5 hands every 100 then your other stats are probably pretty close to accurate.

- Louie

[dogmeat]

I have to agree that 10K is not nearly enough. I have seen a lot of fluctuation in 5k and 10k samples. However, in my case, once I reached about 30K in pt I have seen only small changes in my overall hourly win. I have been playing mostly the same limit for almost 50K now, and my hourly rate has only varied by about a $1 up or down. Even with a cold-streak of about 10 days, once the next week went by my numbers were the same as before the streak.

I guess that is a long winded way of saying that for a player that is not necessarily improving, 30K hands should be a very good indication.

Dogmeat /images/graemlins/spade.gif

[dogmeat]

See "Some Simulations" thread for some more numbers. I stand corrected to a minimum of 250,000 hands.

Dogmeat /images/graemlins/spade.gif

[droidboy]

[ QUOTE ]
After how many hands would you be confident to draw conclusions about your rate because of your play, and not due to the cards running over you in the shortterm?

[/ QUOTE ]

It depends on what you are doing with the hands. If you are just looking at the pure results, you'll need on the order of 100,000 hands. If you were doing curve fitting, you could probably get an estimate which was about as good with only 10,000-20,000 hands.

- Andrew

www.pokerstove.com (http://www.pokerstove.com)

[Wahoo91]

How many posts do you need to evaluate a 2+2 poster with confidence?

LOL /images/graemlins/grin.gif Can someone run a model?