Ok, really, how long until the long term
July 9th, 2014 at 11:08:03 PM
permalink
I'm defining "the long term" here as the length of time before the central limit theorem kicks in and your possible results over your amount of play start to look normal (for some reasonable definition of "looks normal")
The measure "n_0" comes up a lot in blackjack circles, which is useful for a low variance game like blackjack, but for a game like video poker, after n_0 games, your results won't necessarily look normal, and therefore standard deviation is not a useful measure of anything (in particular, it won't be true that you have a 2/3 chance of being within 1SD of the mean, so the useful property of n_0 that you have a 2/3 chance of being ahead (or, behind if the mean is negative) is not true).
I feel that this has something to do with the skewness of the probability distribution of a single game (rather than the variance) but I'm not sure that skewness is enough info.
So, say I define a probability distribution to "look normal" if:
1. It is fairly symmetrical, that is, the difference between the median and mean is less than some previously-chosen epsilon, and
2. Its derivative is 0 at only one point outside of the "long tail" -- ie, in the middle (100% - epsilon) of results for some previously-chosen epsilon (so it has only one local maximum; it does not have multiple peaks and valleys, except for the long tails which will always be wavy)
So, how many hands of something like video poker until my distribution of possible results starts to look like this? Is this just a function of skewness, or something more?
The measure "n_0" comes up a lot in blackjack circles, which is useful for a low variance game like blackjack, but for a game like video poker, after n_0 games, your results won't necessarily look normal, and therefore standard deviation is not a useful measure of anything (in particular, it won't be true that you have a 2/3 chance of being within 1SD of the mean, so the useful property of n_0 that you have a 2/3 chance of being ahead (or, behind if the mean is negative) is not true).
I feel that this has something to do with the skewness of the probability distribution of a single game (rather than the variance) but I'm not sure that skewness is enough info.
So, say I define a probability distribution to "look normal" if:
1. It is fairly symmetrical, that is, the difference between the median and mean is less than some previously-chosen epsilon, and
2. Its derivative is 0 at only one point outside of the "long tail" -- ie, in the middle (100% - epsilon) of results for some previously-chosen epsilon (so it has only one local maximum; it does not have multiple peaks and valleys, except for the long tails which will always be wavy)
So, how many hands of something like video poker until my distribution of possible results starts to look like this? Is this just a function of skewness, or something more?
July 10th, 2014 at 2:34:09 AM
permalink
Quote: AxiomOfChoice
The measure "n_0" comes up a lot in blackjack circles, which is useful for a low variance game like blackjack,
I asked a blackjack dealer for an n-0 low variance game but the Floor told me to get my tail out the door and take my epsilon with me. Was this normal behavior?
July 10th, 2014 at 1:14:02 PM
permalink
what looks normal and how many
all nice
From Jufo81
from my data (maybe to follow)
JOB 9/6 (1 unit = 5 coin)
1k hands
10k hands
100k hands
looks tilted
I think most that play video poker do not really care about all this stuff,
the return of the game over a few thousand hands played is either there or it is not
for those that do, yea!
most vp players also think the games are truly rigged if it is not something close to expectation
I hear this all the time playing video poker
public perception is very strong
What is the coefficient of skewness for video poker?
https://wizardofodds.com/ask-the-wizard/video-poker/jacks-or-better/
as the number of trials increase the skew starts moving towards 0
it could take some time to get there within +/-
why the attempt again?
Sally
all nice
From Jufo81
from my data (maybe to follow)
JOB 9/6 (1 unit = 5 coin)
1k hands
10k hands
100k hands
looks tilted
I think most that play video poker do not really care about all this stuff,
the return of the game over a few thousand hands played is either there or it is not
for those that do, yea!
most vp players also think the games are truly rigged if it is not something close to expectation
I hear this all the time playing video poker
public perception is very strong
What is the coefficient of skewness for video poker?
https://wizardofodds.com/ask-the-wizard/video-poker/jacks-or-better/
as the number of trials increase the skew starts moving towards 0
it could take some time to get there within +/-
why the attempt again?
Sally
I Heart Vi Hart
July 10th, 2014 at 1:27:30 PM
permalink
100k is starting to look kind of normal-ish but there is still some significant skew there, as you point out.
How many hands before the median result is, say, within 5% of the mean (ie, 0.95μ < median < 1.05μ).
More importantly (because I don't actually care about JoB) is there a way to calculate (or even estimate) this without simulations, from the probability distribution of 1 hand?
How many hands before the median result is, say, within 5% of the mean (ie, 0.95μ < median < 1.05μ).
More importantly (because I don't actually care about JoB) is there a way to calculate (or even estimate) this without simulations, from the probability distribution of 1 hand?
