|
beauwillie
|
03-24-2010, 01:42 AM
Post subject: Probability Analysis
|
#1 (permalink)
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
I wonder if you could help me with have a probability question I have. There is an on-line poker site I play that seems to deal me bottom end hands in the blind positions at a frequency far greater than probability analysis would suggest. These become almost automatic check or fold hands and takes the blinds, at least for me, out of the game at frequencies significantly greater than normal. If you can, please check my analysis and tell me if I am looking at it correctly.
First, let me define bottom end hands as the ten pocket card combinations ranked 160 through 169 using a typical hand ranking scheme, say the Skalansky rankings for example. Each of these two-card combinations (unsuited of course) can occur 12 ways. Thus the probability of being dealt one of these hands is 0.0095 or 1 chance in 110.5 opportunities. The probability of being dealt any one of the ten bottom end hands is then just ten times this value or 0.095 or 1 chance in 11.05 opportunities.
Assume a ring game with nine players. The chance of being dealt one of these bottom end hands in a blind position would then be two-ninths of the 10 bottom end hand probabilities or 0.021 or 1 chance in 50 opportunities. In practice, this means it shouldn’t happen, on average, more than once every 25 hands. Of course, you need a statistically significant sample to evaluate before you can begin to reach any tentative conclusions, 2500 minimum and preferably more like 10,000.
Is this analysis correct or am I looking at it incorrectly?
|
|
|
Play for FREE and practice your game at...
Join the FTR Poker Forum to disable these banners and start posting!
|
|
RockyMoose
|
|
Straight
Join Date: Jul 2008
Posts: 196
|
|
You're looking at it incorrectly. It doesn't matter if you are in the blinds or not; your chances are the same no matter your table position. If the odds are 1:11 of getting a crappy hand, as you've defined it, that holds true on every hand, every position.
When you have a pocket pair and see a flop, do your odds of flopping a set change based on your table position?
(It hurts to fold the binds since you put money in the pot, and that may be the reason you notice. Compare this situation to when you are UTG, and it doesn't hurt at all to click the fold button.)
|
|
|
|
beauwillie
|
|
03-24-2010, 06:05 PM
Post subject: Thanks
|
#3 (permalink)
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
Thanks
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
Turns out your answer may not be correct. Re-think it. The net probability is the combined probability of two independent events: (1) bottom pocket cards AND (2) in blind position.Net probability is product of two independent probabilities.
|
|
|
|
ChezJ
|
|
Full House
Join Date: Oct 2004
Location: Washington, D.C.
Posts: 1,456
|
|
what the hell does this have to do with live poker?
|
|
|
|
eugmac
|
|
Full House
Join Date: Jun 2006
Posts: 739
|
|
if this is another post implying that online poker is rigged, for what it's worth, i feel much more vulnerable to collusion and false dealing in live poker much more than online. one time in a casino i used to visit, they used to sometimes run self-dealt games when there weren't enough professional dealers to go around. they caught somebody trying (apparently badly) to deal from the bottom of the deck. i hear they beat the crap out of him, à la Rounders. shame i wasn't there that night...
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
I'm not implying anything, I am just making an observation. I haven't yet reached any conclusion re the implications if I am correct. So far I am saying "so what."
Like you, I have encountered a situation in a B&M casino that I was not comfortable with. The chemistry between the dealer (female) and a player (male) seemed a little too heated for me, especially when he won several big hands playing rags that most would have folded on the deal. No proof, just itchy feeling. All you can do is cash out and walk away.
|
|
|
|
ChezJ
|
|
Full House
Join Date: Oct 2004
Location: Washington, D.C.
Posts: 1,456
|
|
please go away.
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
I don't know much about you except you're a democrat
|
|
|
|
oskar
|
|
4-of-a-Kind
Join Date: Apr 2008
Location: in ur accounts... confiscating ur funz
Posts: 2,452
|
|
This is actually somewhat interesting as a maff problem. You can expected to get more top ranking hands in the blinds against a btn steal. DUCY?
spoiler: it won't proof that online poker is rigged. Since the sites make money with rake, the idea that they would deal people hands that they are less likely to play is among the most retarded I've heard so far.
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
Please note that I haven't said on line poker is rigged. I have only reported what I believe to be an accurate observation. I have then said, if it is true, so what? I don't know what to make of it.
Your 1st sentence is unintelligible and is among the most retarded I've received so far.
|
|
|
|
oskar
|
|
4-of-a-Kind
Join Date: Apr 2008
Location: in ur accounts... confiscating ur funz
Posts: 2,452
|
|
if the question is just how many times you're supposed to get x range in what position, then the answer is: it's supposed to be completely random. There are 1326 combinations of hands in nlhe, if it's a full ring game then you'll only see the blinds 2/9 times and I'd want to have a sample where I see every hand 100 times, so around 600k hands should suffice (easily obtainable from any regular). And then you look for any abnormality that is outside the margin of error, and then you better get your life back asap.
The problem with your method is that you're not taking into account that there are more combinations of unpaired hands than pairs.
Where I was getting at in my last post is that the hand you have influences the action or vice versa. If you get Ax in the BB, then now there are just 12 combinations of Ax hands, 3 instead of 6 combinations of AA that can get dealt - you know, stuff that people open with. That's why your range of hands in the BB is actually different whether it has been folded to you or if there has been a raise.
It's subtle and not very useful, but more fun than ZOMG I FINK MAH NUMBER GENARATER HATEZ ME!!!!  (( - threads. We already have enough of those.
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
You clearly do not know how the numbers work. Do some homework, then come back.
|
|
|
|
oskar
|
|
4-of-a-Kind
Join Date: Apr 2008
Location: in ur accounts... confiscating ur funz
Posts: 2,452
|
|
If you do know "how the numbers work", why do you even bother starting a thread? Just do it, and post the results.
|
|
|
|
JR9477
|
|
Flush
Join Date: Aug 2008
Location: Iowa
Posts: 426
|
|
I'd stay away from Heads-up type games if I were you.
|
(Josh)
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
TY. Very good advice. What I think you're saying is that in Heads Up, the numbers are almost meaningless. So if you let them dominate your decisions you are in for a heap of hurt. But they are very important in tourneys and ring games.
|
|
|
|
eugmac
|
|
Full House
Join Date: Jun 2006
Posts: 739
|
|
beauwillie, you just don't get it. the original post is flawed, the premise is absurd. this thread is finished.
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
It would be more helpful if you could provide support for your unfounded assertions. But like so many self-styled experts, you think you know a lot. Unfortunately, most of what you know is wrong.
|
|
|
|
eugmac
|
|
Full House
Join Date: Jun 2006
Posts: 739
|
|
you were the one coming in here with an "unconventional" theory, and we don't really understand where you are going with this. so please either clarify your position or please stop berating us for not understanding how numbers work. state your hypothesis, show YOUR math homework and the data that you are analyzing, and maybe we will be interested in hearing about your findings! thus far we've seen no data analysis from you of any sort, so what do you expect us to do for you?
besides, this thread does not belong here, as stated above. this forum is for live poker.
|
|
|
|
eugmac
|
|
Full House
Join Date: Jun 2006
Posts: 739
|
|
sorry to double post, but here is the math that you're looking for:
72o has a certain probability of being dealt to you at any given time. This probability never differs according to which position in which you are seated. You have the same odds of being dealt this hand on the button as you are in the small blind.
using standard deviation and sample size you can analyze your data and determine whether the number of times you are dealt 72o in the blinds are within reasonable bounds. i'm no probability buff, so i can't tell you much more than that. you need to have a decent sample size to be able to determine anything for sure, so let's use oskar's suggestion of 600,000 hands.
so you can do the above homework and then get back to us?
|
|
|
|
beauwillie
|
|
High Card
Join Date: Mar 2010
Posts: 10
|
|
OK, I assume you would like to understand the implications of my posts. First, let me re-iterate that I'm not implying that the games are deliberately rigged at the poker site observed. I am just reporting my observations.
Now, as to the numbers. As you point out, 72o has a certain probability of being dealt to you at any given time. To be precise, there are 12 ways to make 72o in 1326 possible pocket card combinations. So there is exactly 1 chance in 110.5 that you will be dealt 72o in any given hand. As you point out, with a data base of 600k hands, on average you should be dealt 72o 5,430 times with a confidence factor of greater than 99.9% (or margin of error of well less than 1%).
Without getting wrapped around the axle with a whole lot of analytical gibberish, here is the point. In a perfect system, with a large database of hands, say the 600k for example, the frequency distribution of actual hands dealt should agree with mathematical predictions with a margin of error of well less than 1% (all this assumes a normal distribution). If, after having collected a statistically significant sample of data, the frequency distribution of actual hands dealt, when looking at all hands dealt, begins to show significant departures from predictions, however randomly they may be distributed, then it suggests, perhaps, a problem with the shuffling RNG. This doesn’t mean someone is trying to deliberately game the system. It just means that maybe there is a problem with the RNG.
So what? Well, be aware that if the shuffling RNG has a problem, there are methodologies that would enable sophisticated Nerds to crack the RNG in ways that would enable them to not only see every card in every hand face-up but to also predict what cards are coming. If such a thing can be done, with the way technology is going, it probably is or will be done. Its sorta like identity theft. Its not something to worry a whole lot about but just stay alert.
|
|
|
|
only_bridge
|
|
3-of-a-Kind
Join Date: Dec 2008
Location: Linköping, Sweden
Posts: 100
|
|
Ok, lets explain this in a simple way.
First you set up a hypothesis, H0.
For instance that you get dealt 72o more often than should.
Then you assume the opposite, and this will be the one hypothesis, H1.
In this example, that you get dealt 72o just as often, or more seldom, than you should.
Then you collect the data.
Note that you cant go back and look at your old hands, as this would be data snooping, and would drasticly influence the result.
Then you decide what confidence level you want to test your Hypothesis at.
Note that you have to do this prior to your calculations. If you do it afterwards then that will have the same effect as data snooping.
Lets say 99.9%, like you mentioned before (ie 0.1% margin of error).
Then you make a simple T-test, or some other common statistical test, of your hypothesis.
Now if you have decided to have the confidence level 99.9%, and you fail to falsify your H1 hypothesis at that level, but you almost manage to falsify your H1 hypothesis at this level. (Lets say that you would have managed to falsify H1 at the 99.5% level.)
Then you get an inconclusive result, and can therefor not veryfy, nor falsify your H1 hypothesis.
Lets say that you manage to prove that your H1 hypothesis wrong. Then you have managed to prove this with a certain probability.
Now another thing that you have to take in to conclution.
Lets say that you have, not only one, but a number of different things you want to test.
Then that will also influence your level of probability.
|
|
|
|
only_bridge
|
|
3-of-a-Kind
Join Date: Dec 2008
Location: Linköping, Sweden
Posts: 100
|
|
Oh, and given your sample size is big enough, you can always assume appoximate normal distriution.
|
|
|
|
only_bridge
|
|
3-of-a-Kind
Join Date: Dec 2008
Location: Linköping, Sweden
Posts: 100
|
|
Quote:
Originally Posted by beauwillie
Turns out your answer may not be correct. Re-think it. The net probability is the combined probability of two independent events: (1) bottom pocket cards AND (2) in blind position.Net probability is product of two independent probabilities.
|
This is correct yes, but you have to plan your test carefully, and first testing that you get bad cards in late position, or on the blinds, and then test if you get bad cards no matter what position, on the same sample, would be very dependent indeed.
|
|
|
|
RockyMoose
|
|
Straight
Join Date: Jul 2008
Posts: 196
|
|
OP is a troll.
|
|
|
|
ChezJ
|
|
Full House
Join Date: Oct 2004
Location: Washington, D.C.
Posts: 1,456
|
|
Quote:
Originally Posted by chezj
what the hell does this have to do with live poker?
|
lock this fucking thread already willya mr mod???
|
|
|
|
JR9477
|
|
Flush
Join Date: Aug 2008
Location: Iowa
Posts: 426
|
|
Why'd you have to bump it?
|
(Josh)
|
English
Deutsch
Español
Français
Italiano
Nederlands
Pусский
Svenska
06-02-2012, 08:10 PM Merge Network 6.0 Looks to Retain Current Player Base
Linear Mode
