variance

[Sciolist]

[ QUOTE ]
My game selection is top of the line, and I never have a problem with variance too much, and my winrates always been good. Sample size? Probably, I'm not clocking 500k a year, but still a substantial amount. Who knows.

[/ QUOTE ]
I've wanted something like this for MSNL (or any cash game to be honest) for a while now. My data doesn't go back very far - I have like 200k hands from Party and 300k hands from stars (from 2003 and earlier). I COULD request a gazillion hands from PS from the best players, but that'd be pretty unethical/likely to get me fired.

Anyone out there with a couple hundred million hand databases who could run it through Poker Grapher and let us know how many/how long breakeven stretches they have, same with winning/losing runs?

Anyone who uses PokerStars, wins, and has 1m+ hands who would give me permission to run my own analysis on their hands? I'd promise not to use them for any other purposes.

[Goldstone]

[ QUOTE ]
[ QUOTE ]
My game selection is top of the line, and I never have a problem with variance too much, and my winrates always been good. Sample size? Probably, I'm not clocking 500k a year, but still a substantial amount. Who knows.

[/ QUOTE ]
I've wanted something like this for MSNL (or any cash game to be honest) for a while now. My data doesn't go back very far - I have like 200k hands from Party and 300k hands from stars (from 2003 and earlier). I COULD request a gazillion hands from PS from the best players, but that'd be pretty unethical/likely to get me fired.

Anyone out there with a couple hundred million hand databases who could run it through Poker Grapher and let us know how many/how long breakeven stretches they have, same with winning/losing runs?

Anyone who uses PokerStars, wins, and has 1m+ hands who would give me permission to run my own analysis on their hands? I'd promise not to use them for any other purposes.

[/ QUOTE ]

You don't need actual data to work out what kind of runs (good and bad) are possible & how likely they are. You can just run a simulation in Excel using the macro function.

[Sciolist]

[ QUOTE ]
You don't need actual data to work out what kind of runs (good and bad) are possible & how likely they are. You can just run a simulation in Excel using the macro function.

[/ QUOTE ]
Really? How? What kind of suppositions do you have to make?

[Goldstone]

[ QUOTE ]
[ QUOTE ]
You don't need actual data to work out what kind of runs (good and bad) are possible & how likely they are. You can just run a simulation in Excel using the macro function.

[/ QUOTE ]
Really? How? What kind of suppositions do you have to make?

[/ QUOTE ]

You need to know your Standard Deviation per 100 and then you can assume various win-rates within a likely range. Then you use the macro function to run a series of sets of hands (i.e. 10000 x 500k hands / a years worth or whatever).

Excel then runs 10000 simulations of a years worth of play and you can see how likely it is given varying win-rates that you run bad / break-even and for how long. It's quite useful if you play for a living and have a reasonable nut to hit every month.

[Sciolist]

Do you happen to have a copy of yours? :]

What kind of sample size do you need to get a reliable standard deviation?

[Goldstone]

[ QUOTE ]
Do you happen to have a copy of yours? :]

What kind of sample size do you need to get a reliable standard deviation?

[/ QUOTE ]

Not sure on that, but if you have a few hundred k hands I think your SD is probably pretty accurate. When I find my copy I'll PM you.

[aejones]

Everyone thinks about stuff like this. I think that I geniunely play better than some people whose WR is twice as good as mine.

It always makes me laugh when people post a hand that says something about how confused they are in this situation... then I read the hand, and it's VERY basic... and the results are like "I didn't know what to do- so I pushed- and I sucked out."

I swear I feel like every "hard lesson" I learned in poker has cost me money.

Also note, how important is tilt in all this? Sick important, if I could get back all the tilt where I've spewed money bluffing or pushed on bad draws, I'd probably run at like 10 ptbb/100.

Thinking about this makes people (including me) sick. All you can do is play the best poker you possibly can and pray for the best.

[johnnycakes]

here's a quote of a post from GuyOnTilt. I can't find the link, but found this text in another thread. The subject is LIMIT hold'em, but it still applies.

[ QUOTE ]

[ QUOTE ]

[ QUOTE ]

GoT did some calculations and concluded that it's possible for a "true" 1.8/100 winner to win like 3+bb/100 or .5ish bb/100. Combine this with the fact that players are bound to change their play from the beginning of a meaningful stretch of hand to the end, and obsessing over your bb/100 rate is little more than an act in futility.
-James



[/ QUOTE ]

Come on. This way oversimplifies things. Just because it's possible for a 1.8 player to run at 3.0 or 0.5 doesn't mean it's likely. As the number of hands increases the level of confidence in the BB/100 number undoubtedly increases, but that does NOT mean that the number is meaningless after 25k or 50k or 100k.

In other words, to paraphrase Peter_Rus' idea, is it possible that someone running at 2BB/100 after 50,000 hands is really a losing player? Yes, possible. Is it likely? No. Stated another way, just because something isn't "statistically significant" doesn't mean it's meaningless.



[/ QUOTE ]

Eh, I'm getting questions now about this so I figured I'd specify. I ran 100 samples of 100k hands each for a 1.80 wr, 16.90 sd player (or 100 different players with the exact same true winrate and sd under the circumstances). Note this test would assume winrate is constant per 100 hands, i.e. no changing game conditions, no tilting, etc. Out of the 100, the highest wr was 3.47 and the lowest was 0.42, with the total wr over the 10 million hands being 1.95, meaning the player(s) was/were running hot for these 10M hands, and not just by a little, yet still one of these samples was as low as 0.42 bb/100.

On the subject of sample size, obviously 100 trials is far too few to come to any reliable conclusions. But these results made me think of variance and the long run in LHE quite differently. If two people playing the same game were to present to me their last 100k hands and one was earning 0.5 bb/100 and the other was earning 3.5 bb/100, who would I think was the better player? Obv, the 3.5 guy. But how much doubt would there be in my mind as to whether he was better or not? Apparently there should be room for some. Winrates just do not converge NECESSARILY until millions and millions of hands. For some they will, sure. Some of us will run close to our true earn for our lifetimes and will rarely or never venture to the upper amplitude of our SD. Others will run hotter than our true earn lifetime; some a little and some A LOT. Same goes with running cold. Some of us will find the very outer bounds of what our SD is capable of, and others won't even get close.

So what determines who among us will get rich and who stays stagnant or drops down? Better players will have a better chance at success of course, and success on a greater scale. But even a WCP could very conceivably end up having to drop down to lower limits while a mediocre player may rise to the big games, maybe never even realizing they're as good as they truly are. It might not even be a stretch to say this HAS happened.

So poker skills being equal, what determines who becomes and millionaire and who keeps playing 15/30? I don't know. QM? Sure. Maybe God DOES play dice with poker, I don't know. What I do know is that this (along with continuing to learn and appreciate Zen philosophies) has helped me come to realize that results, even on an extremely broad or lengthy scale, should be meaningless to me. And I don't mean meaningless in the sense of how I view the game now. I mean in the sense of how I feel I should STRIVE to view the game. We as a group have trained ourselves to not care about 200 bet swings, about 20k hand down periods. None of that comes naturally of course, but as we learned more and more about the nature of LHE we came to accept those things as just part of the package and we learned to deal with it. In the same way, I'm attempting to continually make myself immune to results, period. Not just short-term, but long-term as well. I want to approach this game theoretically and conceptually, without the hint of any wins or losses clouding my judgement. Ridding my conscious from any and all results, period; that is the goal. I'm not there yet by a long shot, but given what I think I know about this game and the philosophy and approach I feel is best for me, my goal is to be constantly progressing toward that state.

GoT



[/ QUOTE ]

[/ QUOTE ]

[Sciolist]

[ QUOTE ]
Also note, how important is tilt in all this? Sick important, if I could get back all the tilt where I've spewed money bluffing or pushed on bad draws, I'd probably run at like 10 ptbb/100.

[/ QUOTE ]
I agree that this is huge. When you see people saying "oh, I run at 10ptbb/100", they are just discounting the big down swings they've had. They either put it down to tilt that they won't ever repeat, or confuse downswings with tilt and vice versa. I really don't think that 10pt/bb is possible in MSNL for many people at all, if anyone.

[Rotterdaum]

Can someone please do that excel simulation for

300k hands, 3ptbb/100, 50ptbb/100 SD
300k hands, 5ptbb/100, 60ptbb/100 SD
300k hands, 8ptbb/100, 50ptbb/100 SD

and the like ...
and show the results?