The great "Poker is rigged" debate - Collected threads edition
If you asked me if the air I breathed contained argon I'd probably decline to answer that, too.
Keeping with the dangeraw theme, maybe you should do something like this...
Suppose you are a better player than I am by some order of magnitude. Let's forget flush and straight draws and concentrate just on the hot and cold equity of our respective hands.
You start with a 74/26% equity to mine. After the flop, you're 88/12%. Now this is just a what if, but...
What if the site took a portion of your hand's equity and distributed it to mine? So if your hand is statistically supposed to win 88% of the time and it only wins 71% against me, would you ever be able to detect this disparity even over a significant sample size? Now let's complicate it by one other factor:
Suppose this disparity in true equity grew with every bet the pot size increases. So in small pots, your 88%, wins very close to 88% (maybe only slightly less). Yet, in large pots when multiple bets go in pre-flop, on the flop, and on the turn, you go from 88% to 64%. Of course, the difference goes to me.
Again, I'm not making accusations. I am simply curious if there is any way that this could ever be detected by anyone? I'm not a math geek, so if I'm misunderstanding hot and cold equity, please correct me. But hopefully, you get the gist of what I'm asking.
While I'm not saying sites do this, it's not like I'd be completely shocked if I learned that they did. Better hands and draws would still win money from worse hands and draws. Better players would still beat poorer players. It would just take more time is all. In the process, worse players "hang in there" longer and more rake is collected. There are many who say that a poker site makes far too much money to risk ever doing this. This is why I'm asking if such a practice could ever be detected and if so, how?
Personally, I don't think this is the case and that it's anemic win rates in the face of rake that gives the illusion of being rigged. To anyone who's struggling on Stars and fancies themselves a decent player, just look at what you paid in rake! You're beating the players, just not the rake. And for someone like myself, I'm barely beating it even though I think I've got awesome skilllz
fr3sh stuff itt
Small samples can (and often times do) skew results.
You start with a 74/26% equity to mine. After the flop, you're 88/12%. Now this is just a what if, but...
What if the site took a portion of your hand's equity and distributed it to mine? So if your hand is statistically supposed to win 88% of the time and it only wins 71% against me, would you ever be able to detect this disparity even over a significant sample size? Now let's complicate it by one other factor:
Suppose this disparity in true equity grew with every bet the pot size increases. So in small pots, your 88%, wins very close to 88% (maybe only slightly less). Yet, in large pots when multiple bets go in pre-flop, on the flop, and on the turn, you go from 88% to 64%. Of course, the difference goes to me.
Again, I'm not making accusations. I am simply curious if there is any way that this could ever be detected by anyone?
What if the site took a portion of your hand's equity and distributed it to mine? So if your hand is statistically supposed to win 88% of the time and it only wins 71% against me, would you ever be able to detect this disparity even over a significant sample size? Now let's complicate it by one other factor:
Suppose this disparity in true equity grew with every bet the pot size increases. So in small pots, your 88%, wins very close to 88% (maybe only slightly less). Yet, in large pots when multiple bets go in pre-flop, on the flop, and on the turn, you go from 88% to 64%. Of course, the difference goes to me.
Again, I'm not making accusations. I am simply curious if there is any way that this could ever be detected by anyone?
Your second factor does nothing but make the overall effect smaller when all hands are summarized.
Further, if you are the favored weak player class, you get results above expectation, and all your opponents get results below expectation. And this is not a 1-1 relationship, it's 1-to-many. So favoring you means skewing (screwing) a hundred or a thousand others. How do you balance this out to not be detectable?
MTTs on Jokerstars today. Played 14 cashed in 2.
Lets have a look at the funny hands that KO d me today...
How can this bs site not be rigged. I would make money every day in low level MTTs if I didnt get KO d by this constant bs.
hey shills if you wana give % for the hands do it when the money goes in.
Lets have a look at the funny hands that KO d me today...
How can this bs site not be rigged. I would make money every day in low level MTTs if I didnt get KO d by this constant bs.
hey shills if you wana give % for the hands do it when the money goes in.
Nobody wants to indulge your rigtard fantasies by doing equity calculations on all the hands you have cherry picked (made up? no hand histories as usual) to "prove" your point. Why are you rigtards incapable of using pokertracker etc to do some analysis of your own rather than constantly moaning? Are you all too stupid to use the many tools available to you or are you just worried they would contradict your delusions?
For somebody to claim something is legit then put it on the market is one thing, but to prove something is legit then put it on the market is another thing. Like that stock brocker that was ripping everybody off, showing people that this person made this much money and that person made that much money. So it's real! Just ask XXXX.
If I'm incorrect, meh. I'm sure the post is here somewhere, but why should I bother proving what I've said is true? Prove it isn't, shills!
I'm not talking about any one player getting special or worse treatment at the expense of hundreds of thousands of other players. I'm simply saying that if WE have access to software that is capable of determining inferior play from opponents (such as PT3 and HM), then you can bet a poker site would have access to much more advanced software to track ALL its players. Now...
It's not about screwing one person or favoring another. It's about the subtle pushing of money back and forth between players. About having the worst players lose at a slightly slower rate. The advantage to a poker site is obvious. Again, the end result is that winning players still win and losing players still lose. When you look at your stats, all seems fine. AK beats A7, AA wins approximately what it should and so on for every situation. And if you have enough hands against an inferior playing opponent, you'll find that you are indeed beating him and that he is indeed losing. The only thing is that the whole whole process just takes a little longer. This means more players at the tables for longer periods of time.
Maybe I have an over-active imagination, but it's not hard for me to conceive of a software like PT3 coupled with a more sophisticated software that calculates and manipulates hand edges based on player stats. So if I'm a clear losing player at 52/12 and you're a clear winner at 31/26, why couldn't they just tweak the equity percentage just a little? Steal a little of ace/king's 88% equity (when in the hands of a solid player), and give 10% of it to A7 (in the hands of an inferior player) on that A63 flop? (just pulling these figures off the top of my head. Don't hold me to them).
I want to make clear that I'm not suggesting this is what they're doing. I just want to know if they DID do it, would it be detectable? My gut feeling is, no. Mainly because we as players have no way of knowing what opponents fold when they miss.
Basically, what I'm doing is turning your explanation to another poster around. You can calculate how far off the mean your flush draws are hitting or missing, but have no way of knowing this same information on how your opponents run against you, since you cannot know what they had when they fold before the river.
Like I said, I'm not a math or software person. But if I can envision such a software that a). Track an opponents skill level (ala PT3), then track the equities of each player's hand (ala Pstove), and then manipulate the card generator to steal a portion of the strongest player's equity and deliver it to weakest player's equity, then I'm sure it could be done. And if I understand you correctly, unlike collusion, this could NEVER be proven. That's all I want to know. I appreciate the time you've taken to respond. Thanks.
It's not about screwing one person or favoring another. It's about the subtle pushing of money back and forth between players. About having the worst players lose at a slightly slower rate. The advantage to a poker site is obvious. Again, the end result is that winning players still win and losing players still lose. When you look at your stats, all seems fine. AK beats A7, AA wins approximately what it should and so on for every situation. And if you have enough hands against an inferior playing opponent, you'll find that you are indeed beating him and that he is indeed losing. The only thing is that the whole whole process just takes a little longer. This means more players at the tables for longer periods of time.
yeah, this would be hard to detect, but it's pretty far-fetched. the software necessary to track and rate all their players across the site and then change outcomes during hands account for the often unpredictable choices made by players seems like more trouble than it's worth. I'd just go the old super-user route, myself.
You gut feeling is wrong. The re-distribution of equity would show up.
I'm not talking about any one player getting special or worse treatment at the expense of hundreds of thousands of other players. I'm simply saying that if WE have access to software that is capable of determining inferior play from opponents (such as PT3 and HM), then you can bet a poker site would have access to much more advanced software to track ALL its players. Now...
Maybe I have an over-active imagination, but it's not hard for me to conceive of a software like PT3 coupled with a more sophisticated software that calculates and manipulates hand edges based on player stats. So if I'm a clear losing player at 52/12 and you're a clear winner at 31/26, why couldn't they just tweak the equity percentage just a little? Steal a little of ace/king's 88% equity (when in the hands of a solid player), and give 10% of it to A7 (in the hands of an inferior player) on that A63 flop? (just pulling these figures off the top of my head. Don't hold me to them).
I want to make clear that I'm not suggesting this is what they're doing. I just want to know if they DID do it, would it be detectable? My gut feeling is, no. Mainly because we as players have no way of knowing what opponents fold when they miss.
Like I said, I'm not a math or software person. But if I can envision such a software that a). Track an opponents skill level (ala PT3), then track the equities of each player's hand (ala Pstove), and then manipulate the card generator to steal a portion of the strongest player's equity and deliver it to weakest player's equity, then I'm sure it could be done. And if I understand you correctly, unlike collusion, this could NEVER be proven. That's all I want to know. I appreciate the time you've taken to respond. Thanks.
I want to make clear that I'm not suggesting this is what they're doing. I just want to know if they DID do it, would it be detectable? My gut feeling is, no. Mainly because we as players have no way of knowing what opponents fold when they miss.
Like I said, I'm not a math or software person. But if I can envision such a software that a). Track an opponents skill level (ala PT3), then track the equities of each player's hand (ala Pstove), and then manipulate the card generator to steal a portion of the strongest player's equity and deliver it to weakest player's equity, then I'm sure it could be done. And if I understand you correctly, unlike collusion, this could NEVER be proven. That's all I want to know. I appreciate the time you've taken to respond. Thanks.
Where? How?
In your scenario, equity is redistributed from the favored hands to the underdogs. Analyzing a large enough sample of hands would reveal that the hand equity in the poker client is not equal to the actual equity of the hand.
Thus, if this is happening, someone would have noticed it when the perform a detailed statistical analysis, as described Spadebidder.
Thus, if this is happening, someone would have noticed it when the perform a detailed statistical analysis, as described Spadebidder.
not on an individuals hands. By the way since most people here make assumptions about other people without reading or understanding what they are saying I am not saying sites rig. They are innocent until proven guilty(statistically wise or someother wild case like misappropriating funds/deposits)
Really though it wouldnt come up if you skew it in a way that makes it statistically unmeaningful. You could distribute the winning %'s by minor or even bigger amounts but until it becomes statistically significant under a given alpha level and large sample size youd have to attribute it to running bad really. Basically you could skew it not to meet expectation but in the range of expectation. You wouldnt be able to tell if a site did that on purpose or if it was bad luck really. And if you increase the power of the test you increase the chance of making a type 1 erroe which statisticians fear more than a type 2. No one wants to claim their is treasure there when it isnt. I mean how would you?
for ex say you tracked all in hands where you an avg of 72% equity. You tested 19 trials and 11 were a success. The PMF of x is 7%. The expected value is 13.78 the variance is 3.788 and the sd is 1.95.
If you were to convert this to a p value and stated your hypothesis test was all in 1 pair hands with 72% equity are not normally distributed and meant to be below expectaton for my account(ha) and your HO was that all in 1 pair hands on site x are normally distributed for my account and is meant to meet expectation(one tailed prediction).
Converting this to a p value you geet -1.41 as a z score which is about 7.93%. If you used a alpha level of 5% this test isnt statistically meaningful. But is "almost" borderline. At -1.65 it becomes statistically meaningful. If the ratio remained you would never really be able to say it was. I mean you could raise the power of the test and use a alpha level of 10% but you increase the chance of type 1 error like i said before. If you always use a 5% significance level it would not be meaningful but it could be skewed(just havent had enough evidence so you technically cant say yeah it is). It could be skewed to run bad but not be anything to detect that raises an eyebrow.
So really how would you know? Sites also deal billions of hands and statistically speaking their will be people at times who do get bad luck. So its weird. IMO you really couldnt detect it unless it was something crazy but why would a site ever skew it to where 2 eyebrows get raised?
And whose really going to check it? I am sure the serious players do while others who go off intuition dont and make themselves look bad.
To frog you should pick up a stats book if you havent. Its a really great topic and is useful in a lot of areas in life. Its great sometimes whe you can find things out for yourself and not have to ask people because you have the skills. So many ppl lie in stats it is rediculous. First time i really started to learn stats the person teaching the audience did a test on someone who claimed being hothoused increased IQ. They tested it and he goes how do we decide what area to reject this? he went basically WHEREVER I WANT! It was kind of a joke but he really wasnt joking. He said we use 5% and explained why but still told us its a gamble.
Also if you were to test a site and make a two tailed prediction and say they rigged it in ways not to meet expectation(wther above or below) you are making a two tailed predicition in which case your significance level becomes 2.5% on each tail totalling 5%. If you used 5% on each tail ur using an aplha level of 10%.
If one tailed prediction the 5% alpha level will due
Basically on the two tail prediction you are hedging your bets.
Really though it wouldnt come up if you skew it in a way that makes it statistically unmeaningful. You could distribute the winning %'s by minor or even bigger amounts but until it becomes statistically significant under a given alpha level and large sample size youd have to attribute it to running bad really. Basically you could skew it not to meet expectation but in the range of expectation. You wouldnt be able to tell if a site did that on purpose or if it was bad luck really. And if you increase the power of the test you increase the chance of making a type 1 erroe which statisticians fear more than a type 2. No one wants to claim their is treasure there when it isnt. I mean how would you?
for ex say you tracked all in hands where you an avg of 72% equity. You tested 19 trials and 11 were a success. The PMF of x is 7%. The expected value is 13.78 the variance is 3.788 and the sd is 1.95.
If you were to convert this to a p value and stated your hypothesis test was all in 1 pair hands with 72% equity are not normally distributed and meant to be below expectaton for my account(ha) and your HO was that all in 1 pair hands on site x are normally distributed for my account and is meant to meet expectation(one tailed prediction).
Converting this to a p value you geet -1.41 as a z score which is about 7.93%. If you used a alpha level of 5% this test isnt statistically meaningful. But is "almost" borderline. At -1.65 it becomes statistically meaningful. If the ratio remained you would never really be able to say it was. I mean you could raise the power of the test and use a alpha level of 10% but you increase the chance of type 1 error like i said before. If you always use a 5% significance level it would not be meaningful but it could be skewed(just havent had enough evidence so you technically cant say yeah it is). It could be skewed to run bad but not be anything to detect that raises an eyebrow.
So really how would you know? Sites also deal billions of hands and statistically speaking their will be people at times who do get bad luck. So its weird. IMO you really couldnt detect it unless it was something crazy but why would a site ever skew it to where 2 eyebrows get raised?
And whose really going to check it? I am sure the serious players do while others who go off intuition dont and make themselves look bad.
To frog you should pick up a stats book if you havent. Its a really great topic and is useful in a lot of areas in life. Its great sometimes whe you can find things out for yourself and not have to ask people because you have the skills. So many ppl lie in stats it is rediculous. First time i really started to learn stats the person teaching the audience did a test on someone who claimed being hothoused increased IQ. They tested it and he goes how do we decide what area to reject this? he went basically WHEREVER I WANT! It was kind of a joke but he really wasnt joking. He said we use 5% and explained why but still told us its a gamble.
Also if you were to test a site and make a two tailed prediction and say they rigged it in ways not to meet expectation(wther above or below) you are making a two tailed predicition in which case your significance level becomes 2.5% on each tail totalling 5%. If you used 5% on each tail ur using an aplha level of 10%.
If one tailed prediction the 5% alpha level will due
Basically on the two tail prediction you are hedging your bets.
Sounds like you're running above expectation.
Increasing your sample size does wonders in decreasing the statistical error.
Really though it wouldnt come up if you skew it in a way that makes it statistically unmeaningful. You could distribute the winning %'s by minor or even bigger amounts but until it becomes statistically significant under a given alpha level and large sample size youd have to attribute it to running bad really. Basically you could skew it not to meet expectation but in the range of expectation. You wouldnt be able to tell if a site did that on purpose or if it was bad luck really.
To your point and partially in answer to PeaceFr0g also, the above is true of your own history, but it is not true of a large pooled history. The larger the pooled sample, the closer the mean must come to expectation, and the closer the full distribution must come to the expected distribution curve, and the more pronounced any shifting between groups of players would be. At large enough sample sizes, the actual sample distribution curve should be virtually indistinguishable from the full population (calculated) distribution curve. The sample size needed for this can be calculated. And if the distribution doesn't match, something is fishy.
1)one person could filter their hands. So you would never really know if all the data was really there. You only have your data and what you can go by. Someone could bring you their data but filter it in a way that pursues their self interest or honestly be making a mistake
2)samples among players would be relatively small too how many hands a site deals
3)if a large group of people did pull their samples together why would they tell if they saw something significant? its useful information. You can now use the data to get an edge. Most people would have too pay for it.
4)people could pull samples together and find nothing statistically significant
5)if something was found people could go well site XYZ deals X amount of hands so something of that nature is bound to happen to a player y% of the time during this period.
theres so many possibilities who knows? there is a suitable trial number for every margin of error but how many players actually reach that number? A bigger sample is always better and will be more reflective no doubt. But smaller samples are only what some ppl have. And they will keep updating their info as they get more and more data. For ex if someone played 100 k hands and had 100 all ins 1 pair hands and found something statistically significant by using a 5% MOE after roughly 400 hands theyd have a pretty accurate sample.
1/.05^2
but if someone wanted a 10% MOE theyd only need 100 hands to test.
Obviously the lower the moe level the more trials needed but the more accurate the data. But after a 100 hands and you noticed something fishy why would you want to go on if using a MOE of 10%? whats your motivation to use 5% if at 10% you increased the power of the rest and will show somethign thats there more than when its not? Obviously you run the risk of a type 1 error( i know .8 is the convential power level)but you increased the power of the test. What i am getting at here is alpha levels sometimes are set on the amount of risk the tester is willing to take. One person might find something statistically significant, while another might not. It comes down to power of the test but if sites were to skew it on a players personal game(I AM NOT SAYING THEY ARE ITS AN EXAMPLE) they could do it where player A will be below expectation in this area but another player wont. If the two players got their data out it most likely wouldnt show anything suspicious. Basically I am saying individual encounters could be skewed different for each player. It be super hard to detect anything suspicious if so.
Player a says my all in hands with aces arent meeting expectation at a pretty significant level over a nice sample size. Player B goes mine are running about at expectation. This could keep going for each players events. The other person could go well your just running bad. I am doing just fine over her ein that area. And vice versa. If both players noticed something they could pull in 2 more people who have a significant sample size and they could be running at acceptable levels of expectation. You can test all players or get ALL the sites hands so..
BTW I AM NOT SAYING ITS RIGGED. DO I THINK IT IS? IDK. I STILL PLAY EVERYDAY AND REVIEW MY DATA MONTHLY. I am neutral ont he subject until my data tells me something else. But i still think about the possibilities.Right now im willing to take that risk until data shows me(if ever) somethings weird. Until then, I will be clicking my mouse and hitting the books.
everysingle time i get 2 pair on that site someone 4 sure gets trips or a higher 2 pair and if this happened to pros all the time there would be no such thing as pros. Everytime I get pocket kings or qq and it does not go all in preflop its almost an automatic that someone catches a set
everysingle time i get 2 pair on that site someone 4 sure gets trips or a higher 2 pair and if this happened to pros all the time there would be no such thing as pros. Everytime I get pocket kings or qq and it does not go all in preflop its almost an automatic that someone catches a set
Thanks onthelow for such a well thought out post. You're right that I should study statistics/probabilities for myself. It would be great not to have to rely on other people. As is, I wouldn't have the faintest idea of even where to start. But just one more question for you...
When you say:
If done extremely in an extremely subtle and gradual fashion , how could we ever make the distinction between running bad, from whatever our long term win rate is? In other words, would you REALLY know if your win rate was chopped by 5,10,20,30% over 1.5 million hands? In order for you to know this, you'd have to be able to calculate millions and millions of examples of EV in various situation based on opponent tendencies, etc. I doubt anyone is that familiar with what their true win rate should be. It's much more likely that they'll eventually come to accept whatever win rate they're left with as their true earn from a game. I myself, wouldn't know if my win rate should be 2.7bb/100 when it's really .8bb/100. I think I know how to play poker, but just don't have the mathematical expertise to tell if I'm running bad or if my results are exactly where they should be after 1million hands.
This is basically why many people claim it's rigged. They feel their edge is or should be, higher than it is. They claim it is statistically unlikely to run bad so often for so long. Admittedly, 99% of these people don't understand variation among sample sizes.
I myself have been guilty of this. I could SWEAR that if you put these same opponents in front of a live dealer at a live table, I'd beat them for much more than I do online. Likewise, I've felt that if you put a bunch of 1st graders in front of a computer and had them randomly hit fold, call, and raise buttons, they'd kick my ass online! But I'm the first to admit that I don't truly understand statistics and I could easily be suffering from delusions and illusions. I'm also not a world class player who knows the intricate complexities that need to be understood in order understand exactly what EV certain lines have with a given hand/flop texture, etc. But I'm pretty sure this is why some people are convinced it's rigged.
Btw- I should mention that I play LHE where edges are lower. From what I understand, a good NL player can make 3-5 PTBB's/100. I doubt even world class limit players beat their games for much more than 1.5bb/100. Many are closer to 1bb/100. I haven't read this whole thread, but I'd imagine that it's the limit players who complain the most about rigging. I could be wrong.
When you say:
You wouldnt be able to tell if a site did that on purpose or if it was bad luck really.
This is basically why many people claim it's rigged. They feel their edge is or should be, higher than it is. They claim it is statistically unlikely to run bad so often for so long. Admittedly, 99% of these people don't understand variation among sample sizes.
I myself have been guilty of this. I could SWEAR that if you put these same opponents in front of a live dealer at a live table, I'd beat them for much more than I do online. Likewise, I've felt that if you put a bunch of 1st graders in front of a computer and had them randomly hit fold, call, and raise buttons, they'd kick my ass online! But I'm the first to admit that I don't truly understand statistics and I could easily be suffering from delusions and illusions. I'm also not a world class player who knows the intricate complexities that need to be understood in order understand exactly what EV certain lines have with a given hand/flop texture, etc. But I'm pretty sure this is why some people are convinced it's rigged.
Btw- I should mention that I play LHE where edges are lower. From what I understand, a good NL player can make 3-5 PTBB's/100. I doubt even world class limit players beat their games for much more than 1.5bb/100. Many are closer to 1bb/100. I haven't read this whole thread, but I'd imagine that it's the limit players who complain the most about rigging. I could be wrong.
1)one person could filter their hands. So you would never really know if all the data was really there. You only have your data and what you can go by. Someone could bring you their data but filter it in a way that pursues their self interest or honestly be making a mistake
2)samples among players would be relatively small too how many hands a site deals
3)if a large group of people did pull their samples together why would they tell if they saw something significant? its useful information. You can now use the data to get an edge. Most people would have too pay for it.
4)people could pull samples together and find nothing statistically significant
5)if something was found people could go well site XYZ deals X amount of hands so something of that nature is bound to happen to a player y% of the time during this period.
2)samples among players would be relatively small too how many hands a site deals
3)if a large group of people did pull their samples together why would they tell if they saw something significant? its useful information. You can now use the data to get an edge. Most people would have too pay for it.
4)people could pull samples together and find nothing statistically significant
5)if something was found people could go well site XYZ deals X amount of hands so something of that nature is bound to happen to a player y% of the time during this period.
There was also the recent Cigital study where Stars provided over 100 million hands for a study, including all hole cards. They weren't looking for bias in the deal, but the database exists.
Again, single player histories usually can't tell you much about the randomness of the deal unless they are very large. But these massive databases can.
At large enough sample sizes, the actual sample distribution curve should be virtually indistinguishable from the full population (calculated) distribution curve. The sample size needed for this can be calculated. And if the distribution doesn't match, something is fishy.
Take Ac,Kh V. 7d,6d on a flop of Qd,8d,2s. According to stove, he should win this hand a little over 43.5% of the time. Well, let's say a poker site takes 2% of your equity (leaving you with 54.3%), and gives it to your inferior playing opponent (who now goes from 43.6% to 45.6%)?
No matter how large your sample size, how could you EVER detect this? Remember, you have no way of knowing what an opponent held all those times he folded the river (did he miss his flush draw? Gutshot? Pair draw?). If I'm wrong about this, please correct me.
I would think being robbed of a measly 2% equity in this particular spot would be absolutely undetectable no matter how large your sample size. In fact, if they were to scale this equity slippage according to pot size (with the smallest pots being zero), I don't see how it would be detectable even they took up to 20% in really big pots, since they come up even less frequently.
All I'm saying is that I see no way for this to be detected without knowing your opponents hole cards in all hands. And there's only one entity that has that kind of info... The poker site
I'll leave you alone for now. I know this is nothing more than pie in the sky, and as a rule, I HATE conspiracy theories myself. It's just something I was thinking about. Thanks again for your help.
Feedback is used for internal purposes. LEARN MORE