The equity bias would favor the regs with < 200k hands, or whatever your criteria for "large enough sample size" would be.

And some of these regs would not even look like regs because we managed to observe only 100 hands before they moved up to nl400 due to running good / at neural EV.

I think that if you compare group A and B within a certain limit and "number of hands" is at least one criteria for defining these groups, EV results will always be slightly skewed due to the effect of "hot runners" moving up.

I'm not sure how significant this effect would be, but if someone would do a statistical EV analysis of "how do regs run" it should at least taken into consideration.

You'd also have regs that run bad and move down.

I've found this thread by accident and i am pretty amazed.

On June 2, 2007 I've posted on internet(even 2+2) such a method for checking the fairness of the RNG and everybody make me crazy.
Even I've written the program that reads the PT2 database to make all the math stuff and graph but give up the research since making poker just a hobby.I will quote myself because maybe someone will get one or two interesting ideas.

"
Some people say that by rigging the odds in big pots online poker sites are trying to curb the flow of money from bad players to good players in order to boost their profits.

Is online poker rigged or not?

Hard to say because we need aggregated hand histories to do serious research.
But I can do some math against specific situations.

Let's say I have hands H1,H2,....Hn where you are against one opponent and one of you two (or both) moves ALL IN.
Let's name your equities E1(H1),E2(H2).....when an ALL IN situation occurs.

Let's define a function that returns the size of the pot when you WIN and 0 when you loose.
R(i)={ 0, if you loose
POTSIZE(i)/2, if we have a split
POTSIZE(i), if you are the winner
}


R(1)+R(2)+R(3)..................
and
E1(H1)*POTSIZE(1)+E2(H2)*POTSIZE(2)+.............. .

should have very close values!!

Now let's define the problem and use statistics to solve it:

N is the number of the ALL INs games logged in a database.
In each of these games there was an equity W(i) I have in pot P(i)

P(I) is the potsize when one of the two players moves ALL IN
0<=W(i)<=1 is the equity I have from the P(i) pot

So in each game I can win R(i) dollars:
R(i) = {
P(i) , I win
P(i)/2 , pot split
0 , I lost
}

Theoretically after all these N games I should have win something like E = sum after i ( W(i)*P(i) ) with i=1,N

In reality I have won S dollars.

How likely is to have such a big abs(E-S) value??

The math part was solved with help from JakeD here
It's about 'standard deviation' and 'condidence interval's

http://www.mathhelpforum.com/math-he...viation-4.html

I'm sure this has been discussed already,


But why can't we only analyze data from HU sngs/cash games when all in preflop. Wouldnt this eliminate hand selection issues? e.g. AA vs 6s7s/JsTs aipf is 2x.xx% underdog hu- even if villain has skewed handselection pf.

Or why can't we just fill up a 6max 5nl table (with 4 ppl sitting out) and shove everyhand preflop?

Or best yet... we could fill up a 6max/FR 5nl table with only 2+2ers and to just play poker. Everyone ships their HHs and we can analyze different scenarios more deeply.

I am sure that there are some mid/highstakes guys who are willing to donate/ship some money to explore/fund such experiments where absolutely no data is missing- especially say if you're a 3/6reg who is suspicious. At 5nl we wouldn't need a ton of money to attempt this. Even if it is a small sample size, the data is still perfect and can always be used in future.

I also would be interested in seeing EV graphs of pots of say, >170bbs, and just not allin-luck for all pots.

What interested me most about this thread was the bunching microbob effect. And the regs with negative luck graphs seem to have the most legimate claim of "riggedness". I hope this discussion does not die.

This is the first I've heard/seen the luck graph. The few graphs that were displayed were quite linear, leading me to believe that the number of hands required to see "results" is not too big anyways. Except that guy that had to leave Russia had 200k, played weak-tight, had positive "luck", but did say he didn't get it all in very often.

I would like to congratulate micorbob for what I believe is the solution. I think that "luck graph" is linear with slope dependent on playing style.

Is there not someone with many hands on both pokerstars and fulltilt to run a test?

Would you guys channel your energy into finding Usama or something?

There's a lot of trash in this thread cluttering up a legitimate gripe and discussion. If you don't know what you're talking about, please stop. I couldn't get past the first 100 posts because there was so much garbage. MicroBob, your essays are counterproductive.

Here is the question: Does PokerStars make grinders lose more often than they should when they are all-in?

Possible explanations:
A mistake in the calculations.
Sample size.
Maybe something about people being more likely to play hands with A's in them and thus skew the EV calculation?

How to prove or disprove this:
Someone needs to collect a LARGE HH DB and analyze it. THEN COMPARE IT TO A SIMILAR DATABASE ON ANOTHER POKER SITE E.G. FTP

Please only post if you have something meaningful to contribute.

not only counterproductive, but also torturous. i stopped reading bob's posts a month ago, and i feel liberated. (100% serious)

there's absolutely a threshold between useful and pure clutter. imo, he's way over the threshold and is a good example of why posting limits (quantity) should be considered here.

i guess that was not meaningful, but it had to be said.

I, on the other hand, have posts of utterly high contents & qualities.

I also had to suffer through the first 100 posts. I just skipped all of Microbob's since someone would quote his relevent points (I do thank you though MBob for all the work and analysis).

Anyways, I'm a SNG player and have had torturous "luck" results as well. I'm sure I'm in the top 10 volume players on Stars (SNE), so hopefully my results and database are large enough. Here's a post on my results and analysis:

http://forumserver.twoplustwo.com/36...report-316556/

I honestly don't know what the reason for this is. Like I posted, there are 4 possible reasons I can think of:

1) Mine, Juk's, and Phil's programs are wrong (unlikely)
2) Someone has hacked Stars and is fixing the cards/games (unlikely)
3) Stars is deliberately screwing the sharks and redistributing to the fish (unlikely)
4) I'm on the worst 75,000 SNG stretch of all time (unlikely)

I'm going to run my luck calculation software on my SNG data set to see how the top 20 regs faired to see what their luck was.

Keep up the great discussion!
-BK

Some of you guys are so invested to prove something is rigged that you just don't get it.

I have found all MB posts very serious and they offer a pretty good explanation about what could may happen.

It's not that your programs are "wrong"

What it could be wrong it's the way you interpret such results.

Sorry guys, but a lot of you don't have a minimum clue of how to apply the scientific method. If you want to prove a theory and let out a possible explanation because just don't fit with what you want to prove, your research is worthless. Same if you know there exist variables who may affect your research and you just dismissed them because your "intuition" tells you so, your research is equally worthless.

Disclaimer: (I just read the first 100 posts)

I guarantee if we look at the data closely we will find that high cards are favored on flops - and overall hand values are higher than the standard %'s we all know and love.

This does not mean the cards are "rigged", it just means that you are more susceptible to variance, and you should not go AI with low pairs. A cashout curse (at least mine anyway) could be occuring because we are on a downswing following an upswing.

I've analyzed some of my HH's and I think I may be right. Don't have enough data though.

But one thing I noticed even when I first got to PS was the high number of Ace high flops. I was never one of those who accepted the fact that the cards would play like a real deck.

aren't you that guy that made all those ******ed posts in stt about how it was always +ev to complete in the sb with atc because you can just cr the flop and its actually like being in position, and the guy that said you had something like a 30% roi in sngs because you had played like a couple hundred and had a big score in a 180 or something? i'm pretty sure that was you. not too surprising to find you here.

Actually I think you may the one lacking the scientific method in your thought process. The hypothesis here has been that regulars at sites, Stars in particular, tend to run below expectation. Microbob's posts have assumed the hypothesis and tried to counter it. This is bad science since it still hasn't been entirely proven that regulars run poorly at Stars in the first place.

And jumping somewhat ahead of myself, there's also the issue that his explanation is effectively irrelevant since data can be compared in contrast to other sites - where any 'natural' irregularities would be compensated for naturally. So far we've failed to do comparative analysis in a systematic manner. All we have is anecdotal data that seems to suggest most of those regulars who have complained about running significantly below EV, have come from Stars. But at the same time it is not unlikely that more players plan on Stars than other sites - thus compensating for the disparity [which itself has not even been definitively established].

So rambling aside - the point is that we need to start doing comparative analysis of data in a systematic manner. If the average regular is running below expectation at Stars, to a relevant degree, but the average regular at Full Tilt/Cake or whatever other site we use as a control sample, is not running below expectation - then we have some very conclusive data where we'd need to start looking for potential confounding variables that would be explicitly relegated to Stars. In this case, Microbob's theory would not possibly be a reason - since any sort of bunching would be game and site independent meaning the comparative analysis would naturally account for it.

I've mentioned this multiple times before, but I'm not sure some people are understanding the concept of comparative analysis with a control sample - since bunching keeps getting brought up over and over again.

Totally agree with you, and to be honest, I'm surprised it hasn't been done

Haha. You didn't understand. Sng's allow lead outs from the SB because it's +EV to have low stacks to your left and big ones to your right. Leading from the SB tends to flow chips to the right, as you pass a bet through the BB.
The less chips the person to your left has, the better off you will be as the blind pressure increases. I was just trying to explain ICM calculating the size of the stacks but not calculating how those stacks were positioned. You guys countered with "a penny saved is a penny earned" without ever learning how to use your "blocks" of chips.

Intersting how you interpreted that as ALWAYS complete. I probably said something like: you should always complete at 9:1 odds from SB because you can check any flop for temporary position or lead draws to redirect chips.

Only Gigabet understands.

My memory of that is I pissed off all the losers in that forum by telling a poster how to take advantage of all the crappy players doing SNG's.

So funny cause I was winning SNG's so often I got bored of them. Haha.

Can you elaborate?

I'm looking for datamines for the month of August in 6-max NL 100 from Full Tilt for comparative analysis with data from Stars. Similar data for the same time period/stakes at other sites would also be very useful.

They all do that now. The next card dealt is not known to the site software until it is ready to be dealt (all betting is completed), at which time the RNG creates a card to be dealt from the remaining deck (or a random deck with X remaining cards) and sent to the users software. This prevents any possibility of cheating by knowing the card in advance, even if you are a site insider.

This also means that when you see the flop after you folded and you say, "Wow I would have flopped a straight", you are wrong, it would have been something different. The RNG doesn't even kick in until the round is done and it is time to deal a card.

This is incorrect. As described on their site: http://www.pokerstars.com/poker/room/features/security/

There is an abstraction of a shuffled deck from which cards are drawn from - rather than dynamically instantiated cards.

I guess I should not have generalized, but FullTilt among others does do what I described.

Bumping this before bed. I'm already developing the software for comparative analysis. All I need now is more data from different sources.

OK it only took me a little more than 2 hours to get through this thread. I have some questions if anyone would like to answer.

Aren't poker software RNG's referred to as PRNG's (P = pseudo) in the cryptology industry because they use a seed that is based on a mathematical formula rather than a truly random number?

My understanding was that it starts with a seed that is randomized through "thermal noise". I'm not sure of the exact definition but I thought I read somewhere it was randomized by mouse movements of the users.

Kind of like a game of telephone, where a person starts with a word (say it's 8 letters long) and whispers it to one person, who whispers it to another, and so on. When it gets to the end, the word is completely different.

But the initial word is not exactly random, since the subset of 8 letter sequences that are actually words is <<<< than the full set of 8 letter sequences. You can predict the original word by working backward and asking each of the people in the chain what their word is, but in a computing environment, there is no processor fast enough to work it's way back to the original seed in real time.

Is this not how an RNG works? If so, can the cards be unpredictable, not perfectly random, but still not be "rigged".

I ask because if we do come up with a scientific method of testing the shuffles, and we do find a bias for some particular kind of deal, that does not mean it's rigged.

So even if we prove that it's not random, does that matter? Should we even expect it to be random?

Joely, I hear often about how the "seed" is not truely random and I just don't get it - surely it is a bit like shuffling a proper pack of cards but someone telling me it can't be a random shuffle because I started my shuffle with the ace of spades at top.

Not exactly. I'm pretty sure it doesn't work this way but the easiest way that I can think of to shuffle mathematically is to have one number for each of the ~1.7*10^80 different combinations that a 52-card deck could make. A mathematical formula could never even approach that. In fact, I suspect that only a tiny, tiny fraction of the possible combinations are used. This will make certain cards more or less likely to appear in certain spots.

This may be the source of what is percieved on the user end as "rigging". Let's just say that a 5 is more likely to occur on the flop than every other card.

A fish comes into the game. Knowing nothing about the importance of position, he limps EP with A5.

A skilled player raise on the BTN with AK.

The flop is A57, and the skilled player goes broke to the fish. The skilled player may think he is a 75% fav pre-flop, when he is actually more like 68%.

I am not saying that it is not rigged. I wouldn't put it past anyone in this environment. But even if we establish some sort of bias or favoritism for bad hands, it does not prove that someone is behind it doing it intentionally.