Two Plus Two Poker Forums FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, Bodog
 Register FAQ Search Today's Posts Mark Forums Read Video Directory TwoPlusTwo.com

 Notices

 Probability Discussions of probability theory

06-19-2012, 10:06 PM   #91
journeyman

Join Date: Nov 2009
Posts: 290
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Sherman This is where you went wrong. Anything that changes the way cards are dealt to something that is non-random can be detected.
Your gonna have to explain that a little closer.

What if the rig was made so that the outcome would be what we see in this analysis?

06-19-2012, 11:31 PM   #92
Carpal \'Tunnel

Join Date: Jun 2005
Location: Psychology Department
Posts: 7,430
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Easyonkemp Your gonna have to explain that a little closer. What if the rig was made so that the outcome would be what we see in this analysis?
You are going to have to explain that a little better. I don't understand what you mean by what if the rig was made so that the outcome would be what we see in this analysis?

Certainly you can set up a rig that is undetectable by a particular analyses. My thermometer is quite terrible at telling me what time it is. But any rig can be detected by using the right analysis, just as the proper tool for telling time can tell me what time it is.

If you define the rig, I (or someone else) can define a method for detecting it (so long as the rig is some deviation from a random process). Many people theorize that the rig could be "so small" that it cannot be detected. But that theory fails because with a large enough sample size (say millions of poker hands) even small rigs will be detected. Others theorize that rigs are "self-correcting" so that a rig will apply one way in certain situations, but another way in other situations. However, such "self-correcting" theories fail to recognize that in doing so they have actually invoked a non-random process twice(!) both of which can be detected.

Now, I will admit that it is possible for something to be rigged in such a way that no one has thought of and therefore no one has ever tested for. But every testable theory for how "Stars rigs the game" that I am aware of has been soundly dis-proven by appropriate data analysis or by pointing out how such rigs are logically impossible.

So it is possible that poker websites are rigged in some way that no one is aware of. But in my opinion it is extremely unlikely because the data that have been actually analyzed by people who know what they are doing indicates the games are products of random processes as one would expect.

06-20-2012, 05:27 PM   #93
journeyman

Join Date: Nov 2009
Posts: 290
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Sherman You are going to have to explain that a little better. I don't understand what you mean by what if the rig was made so that the outcome would be what we see in this analysis?
This analysis showed that the hands that were tested the underdog got favored at those 60k hands or w/e that were tested. What if that was just because of rigging? Like what wykh said. Enhancing the degree of how much the worse hand won.

Quote:
 Originally Posted by Sherman If you define the rig, I (or someone else) can define a method for detecting it (so long as the rig is some deviation from a random process). Many people theorize that the rig could be "so small" that it cannot be detected. But that theory fails because with a large enough sample size (say millions of poker hands) even small rigs will be detected.
I dont understand why you would always detect a small rig in millions of hands?

Quote:
 Originally Posted by Sherman Others theorize that rigs are "self-correcting" so that a rig will apply one way in certain situations, but another way in other situations. However, such "self-correcting" theories fail to recognize that in doing so they have actually invoked a non-random process twice(!) both of which can be detected.
By self-correcting you mean that if you would want to give one player an aces vs kings setup, that will be corrected later by the same setup the other way around? Or similiar scenarios. Yes, then you have two processes that are not random, but they are working towards the goal of making it look random. So how would you detect that then?

What about a rig that keeps track of the players and give better cards and more luck in some situations (situations based on playing cyclus, history, profits etc.), but will also correct what cards it has given before later. So the HH at any point dont have too many aces or has hit the flop too much, too good EV etc.

 06-20-2012, 05:33 PM #94 Carpal \'Tunnel     Join Date: Jun 2005 Location: Psychology Department Posts: 7,430 Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B Any changing of the cards dealt in a non-random fashion can be detected if the sample size is large enough. Period. It is a fundamental principle of statistics. If there is a true effect (i.e. not a null effect) and the sample size is large enough, the effect will be detected beyond reasonable doubt. In the example you are referring to, there was an effect detected, but it was not so large an effect that we would be confident it occurred because of a non-random process. That is, an effect size (i.e. deviation from what we would expect if nothing was wrong) that is as small as the one you are referring to has a reasonable probability of occurring even if things were dealt fairly. I just don't know if I can do it any better than that without taking you all the way through a basic statistics course.
06-20-2012, 09:49 PM   #96
journeyman

Join Date: Nov 2009
Posts: 290
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Sherman Any changing of the cards dealt in a non-random fashion can be detected if the sample size is large enough. Period. It is a fundamental principle of statistics. If there is a true effect (i.e. not a null effect) and the sample size is large enough, the effect will be detected beyond reasonable doubt.
I still dont understand why this is?

Quote:
 Originally Posted by Sherman In the example you are referring to, there was an effect detected, but it was not so large an effect that we would be confident it occurred because of a non-random process.
Well, exactly? Rigging it that way to that degree would go undetectable.

Quote:
 Originally Posted by heehaww I wanna point out what should be obvious. If your rig theory started as a result of your gut feeling ("Omg the odds of this are just too low"), then your gut feeling was that there was a large rig, not a microscopic one. When you got sucked out on a few times in a row, you didn't think, "OMG the percentages must be off by at least 0.0000004%", you thought they were off by much more than that. If there is a rig, and it's so small that powerful statistical analysis can't find it, then your intuition didn't detect it either. It amounts to a few extra cents rake, so it's not like you would notice the effect.
I dont think you should look at where this is comming from and why, but rather what it is.

I also think that you can get blinded a little by only looking at the statistics of the hands and by that define the whole story of what is going on. Because its quite complex, and there are only so much you can actually get good statistics of anyway. Like: Ok, I got a trillion hands here. And I have gotten all the starting hands correct amount of times, I have seen various flops correct amount of times, won all-ins, lost, everything looks fine etc. But it doesnt tell me when I lost or won, how much, and to who. The statistics isolates the players. And at the end of the day it is what players wins whose monies that matter.

What I would like to see (not expecting) is stats of player types. Are the profitting players running like the not-profitable EV wise? Do new players get more lucky on a site at first? Do moving up in stakes often makes the players run bad? I am having trouble neglecting these kinds of things to happen by analysis that are only made on pokerhands.

06-20-2012, 10:38 PM   #97
Carpal \'Tunnel

Join Date: Jul 2008
Posts: 7,390
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Easyonkemp I still dont understand why this is?
Because any change you make means the cards are no longer randomly dealt.
Quote:
 Well, exactly? Rigging it that way to that degree would go undetectable.
Except that the rigging itself would be easily detectable based on card distributions.
Quote:
 I also think that you can get blinded a little by only looking at the statistics of the hands and by that define the whole story of what is going on. Because its quite complex, and there are only so much you can actually get good statistics of anyway. Like: Ok, I got a trillion hands here. And I have gotten all the starting hands correct amount of times, I have seen various flops correct amount of times, won all-ins, lost, everything looks fine etc. But it doesnt tell me when I lost or won, how much, and to who. The statistics isolates the players. And at the end of the day it is what players wins whose monies that matter.
You can't change those things without changing the cards. Once the cards are changed, it's game over for the "undetectable" rig.

 06-20-2012, 11:17 PM #98 Carpal \'Tunnel     Join Date: Jun 2005 Location: Psychology Department Posts: 7,430 Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B The analysis you refer to has been done (at least it was attempted). I forget which site or person did it, but I distinctly recall seeing an analysis where someone tried to identify players as "winners" and "losers" and then looked to see how they ran in terms of expectation. In any case...let me try to give a basic overview of how these analyses work. First define a null hypothesis. In cases such as these, the null hypothesis is that no rigging occurs (i.e. that the deck is being dealt randomly). Then define an alternative hypothesis...in this case that the deck is not being dealt randomly. To get more specific, let's take a simple example. Say you believe AA is being dealt too often. The null hypothesis is that AA is being dealt the correct amount of times. The alternative is that it is being dealt too often. Now collect the data. Lets say you collect 10,000 hands and record the number of times AA occurs. We would expect to see AA approximately 45 times in 10,000 hands. Let's say that in your data you have seen AA 52 times. 52 is more than 45, but how confident are you that the deck is not being dealt random? One way to address this problem would be to think about how many times in 10,000 hands you would actually observe 52 (or more) AA hands? We can use probability and the normal distribution to answer this question. One method that seems appropriate for this data is a one sample t-test. Without going into the details, the formula for a one sample t-test is as follows: T = (X - M) / SE where X is the mean of the observed data, M is the expected result under the null hypothesis, and SE is the standard error of the observed data. In this case, mean of our observed data is .0052 and the expected mean is approximately .0045, for a difference of .0007. The standard error of the observed data is given by the standard deviation of the observed data divided by the square root of the sample size (10,000). The standard deviation is a measure of how spread out (or variable) the observed data is and it is given by: SD = sum[(x - X)^2] / (N - 1) where x is a single observation (either 0 or 1...either we didn't get AA or we did), X is again the mean of our observed data, and N is the sample size. Using this formula we get a standard deviation of approximately .06693. Dividing the SD by the square root of our sample size (10,000) we get a standard error of approximately .000719. Now we divide our numerator (.0007) for the t-test formula by our standard error to get a t-value of approximately 1.0. I used rounding here, but without rounding a more precise answer is T = 0.9386. Now, we can use the t-distribution to determine the probability of observing such a t-value if the null were actually true. Or in other words, the probability of observing our 52 AA if the deck were dealing randomly. You can find this answer yourself by entering the following in an empty cell in Excel: =TDIST(0.9386,9999,1) The resulting answer is approximately .174. That means that if we deal out 10,000 hands fairly from a random deck we would expect to see 52 or more AAs a little over 17% of the time. In my view, that is hardly a rare event. That tells us that we cannot be too confident that the deck is somehow rigged to deal AA to us too frequently. Does that mean the deck is not rigged? Well, not yet. It is possible that the RNG is dealing AA to us at exactly a 52 per 10,000 hand rate. Unfortunately our sample size is not big enough to confirm that we have a true difference. However, if we repeated the study again but this time collected 100,000 hands and the effect was still the same size (i.e. we get AA 520 times out of 100,000), we would have much more convincing evidence that something is amiss, because this time the probability would be approximately .00015. The basic equation though is this: Statistical Test = Sample Size X Size of Effect A larger statistical test (such as the T-value) indicates that the probability of observing such an effect in a truly random data set is less likely. This equation is instructive because we see that smaller effects require a larger sample size to detect. However...effects that are truly null (i.e. effects that don't exists) will never be detected no matter how large the sample size is. So it is possible that poker websites rig decks in ways that the effects are so small we would need trillions of hands to detect them, but that is very unlikely in my opinion as such a small effect would have almost no impact on their bottom line. That is, the smaller they make the rigging effect, the less benefit they get out of rigging. So it doesn't really make any sense. This is consistent with the other poster's comment that most people who claim rigging are claiming a big (and obvious) effect exists. And if the effect is big, it should not take a large sample to detect those effects (i.e. a million hands is easily plenty).
 06-21-2012, 11:47 AM #99 Pooh-Bah   Join Date: Dec 2006 Posts: 5,133 Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B I estimate with 100% confidence that this explanation will not change his beliefs, but it was rather interesting to read. Another factor which needs to be stressed is that there are thousands of qualified eyes always watching the data, and that is what helps instill confidence in the marketplace. Whenever valid, quantifiable concerns arise we see threads which micro-analyze the data to get to the truth of the matter, and we have found rooms that have tried to hide (like Pitbull) basically cease to exist as a result. I realize it is part of some human's nature to be paranoid about basically everything, and who knows, maybe some education will make some of them feel better in some ways, but others reading this thread can have confidence that plenty of people exist that can do the needed research that they cannot.
06-21-2012, 01:51 PM   #100
journeyman

Join Date: Nov 2009
Posts: 290
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Sherman So it is possible that poker websites rig decks in ways that the effects are so small we would need trillions of hands to detect them, but that is very unlikely in my opinion as such a small effect would have almost no impact on their bottom line. That is, the smaller they make the rigging effect, the less benefit they get out of rigging. So it doesn't really make any sense. This is consistent with the other poster's comment that most people who claim rigging are claiming a big (and obvious) effect exists. And if the effect is big, it should not take a large sample to detect those effects (i.e. a million hands is easily plenty).
I agree that giving out aces too much is not a good idea. I dont see the rigging beine done in that way. Giving out too much or too little of anything. Either you will just get caught by HHs test, or the effect will be so small that its not really worth it. There are however, times when aces will be worth more. Like when other players have strong hands or will be hitting flops. Basically, you can have a HH that is "correct", but is worth more than another HH that is the same, because you are getting strong hands while others have quite strong hands too. So you will gain more money of your strong hands, but you dont get dealt them more often and dont hit the board more often. What do you think the main "obstacle" of rigging it in such a setup-way is for example?

I would also like to get your views on what I wrote earlier about the rigging being done on a player-basis and is at the same time working towards looking random in the card distribution.

And, was wykh not right when suggesting that a rig that enhanced the underdogs chanses to the degree that were done in this analysis would go undetected? (the reason I actually bumped this thread). Or are there other kinds of testing that would catch such a rig, since this analysis didnt?

06-21-2012, 02:27 PM   #101
Carpal \'Tunnel

Join Date: Jul 2008
Posts: 7,390
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Easyonkemp Basically, you can have a HH that is "correct", but is worth more than another HH that is the same, because you are getting strong hands while others have quite strong hands too.

06-21-2012, 02:29 PM   #102
Pooh-Bah

Join Date: Dec 2006
Posts: 5,133
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Easyonkemp I agree that giving out aces too much is not a good idea. I dont see the rigging beine done in that way. Giving out too much or too little of anything. Either you will just get caught by HHs test, or the effect will be so small that its not really worth it. There are however, times when aces will be worth more. Like when other players have strong hands or will be hitting flops. Basically, you can have a HH that is "correct", but is worth more than another HH that is the same, because you are getting strong hands while others have quite strong hands too. So you will gain more money of your strong hands, but you dont get dealt them more often and dont hit the board more often. What do you think the main "obstacle" of rigging it in such a setup-way is for example?
You need to anticipate how players will behave in essentially an infinite number of combinations in advance. Unless "mind control" is part of your theory, this would be an impossible "rig" to pull off, and you would always run the risk of someone behind the rig revealing what they know (remember, hundreds of poker rooms have come and gone, so plenty of opportunity for that has passed).

Don't worry - I realize this and everything Sherman and everyone else has said will do nothing to change your beliefs, but others reading this may appreciate how impossible your theory is in practice. I know you will not agree and that is fine, I am not trying to change your mind on anything.

Quote:
 Originally Posted by Easyonkemp I would also like to get your views on what I wrote earlier about the rigging being done on a player-basis and is at the same time working towards looking random in the card distribution.
Impossible and impractical with millions of players playing in infinite combinations of player mixes (often changing dozens of times per table each hour). Don't worry, I realize this will not change your beliefs, and I am not trying to change your beliefs.

Quote:
 Originally Posted by Easyonkemp And, was wykh not right when suggesting that a rig that enhanced the underdogs chanses to the degree that were done in this analysis would go undetected? (the reason I actually bumped this thread). Or are there other kinds of testing that would catch such a rig, since this analysis didnt?
He was not right. He said he made a mistake in his calculations, and explained that (as he did for all of his various rig theories). You were told this in the rigged thread with appropriate quotes from wykh but I appreciate minor details like that do not change beliefs such as yours.

All the best.

06-21-2012, 02:45 PM   #103
Carpal \'Tunnel

Join Date: Jun 2005
Location: Psychology Department
Posts: 7,430
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Easyonkemp I agree that giving out aces too much is not a good idea. I dont see the rigging beine done in that way. Giving out too much or too little of anything. Either you will just get caught by HHs test, or the effect will be so small that its not really worth it. There are however, times when aces will be worth more. Like when other players have strong hands or will be hitting flops. Basically, you can have a HH that is "correct", but is worth more than another HH that is the same, because you are getting strong hands while others have quite strong hands too. So you will gain more money of your strong hands, but you dont get dealt them more often and dont hit the board more often. What do you think the main "obstacle" of rigging it in such a setup-way is for example?
The largest obstacle in the way of that sort of rig is that it is ridiculously complicated. Can you imagine writing the software code that sets this rig up? "Ok, we want to set it up so that the hands are equally balanced, but in certain situations (let's say if people have big stacks of chips in front of them) we will deal more big hands against each other." That might sound simple, but writing a piece of software that can do all that will not be easy...at least no where near as easy as writing a piece of software that simply deals the game fairly.

But that isn't really your question. You really want to know if this sort of rig would be detectable. The answer is yes. And analysis have been done looking at these sorts of things. We can do analysis to look at what hands others tend to get when I have a certain hand. We can look at what hands tend to go to showdown against each other. It is trickier to analyze for sure (because it is a simple fact that two big hands are likely to go to showdown whereas as a big hand vs. a weak hand is not nearly as likely).

Quote:
 Originally Posted by Easyonkemp I would also like to get your views on what I wrote earlier about the rigging being done on a player-basis and is at the same time working towards looking random in the card distribution.
I already gave my views on that in the very first sentence of my last post. Player basis rigging theories have been examined and seem to have failed.

Quote:
 Originally Posted by Easyonkemp And, was wykh not right when suggesting that a rig that enhanced the underdogs chanses to the degree that were done in this analysis would go undetected? (the reason I actually bumped this thread). Or are there other kinds of testing that would catch such a rig, since this analysis didnt?
No. Wykh is wrong in that post. The pokersite cannot set up a small effect that is "1.5 SDs off the mark" forever and go undetectable. The standard error of the sample mean decreases with the square root of the sample size (SE = SD / sqrt(N)). If the effect remains the same and sample size increases, the likelihood of detecting the effect increases as well.

Or in other words, if at the population level there is an effect (i.e. the computer software is rigged) a large enough sample will detect such an effect with great confidence. This takes us back to the basic equation I wrote before which shows us that a small effect can be detected if the sample size is large enough.

06-21-2012, 08:41 PM   #104
journeyman

Join Date: Nov 2009
Posts: 290
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Sherman I already gave my views on that in the very first sentence of my last post. Player basis rigging theories have been examined and seem to have failed.
Yea I saw that, but you didnt say anything about your thought on the process and why such rigs would get detected. Would not this usually be data that you would have to get from voulenteering players?

Anyone else know anything about these analysis?

06-21-2012, 09:45 PM   #105
Carpal \'Tunnel

Join Date: Jun 2005
Location: Psychology Department
Posts: 7,430
Re: FLOP ALL-INS analysed for bias. 6 million+ hands Stars, Party, Ongame, Merge, Entraction, B

Quote:
 Originally Posted by Easyonkemp Yea I saw that, but you didnt say anything about your thought on the process and why such rigs would get detected. Would not this usually be data that you would have to get from voulenteering players? Anyone else know anything about these analysis?
I think it may have been an analysis done by a poker hand database website like Poker Table Ratings. Because they have nearly entire samples of hands for lots of players, they could easily identify people who had been losers and people who had been winners. So you wouldn't need volunteers.

 Thread Tools Display Modes Linear Mode

 Posting Rules You may not post new threads You may not post replies You may not post attachments You may not edit your posts BB code is On Smilies are On [IMG] code is On HTML code is OffTrackbacks are Off Pingbacks are Off Refbacks are Off Forum Rules

All times are GMT -4. The time now is 05:27 AM.

 Contact Us - Two Plus Two Publishing LLC - Privacy Statement - Top