a note to you conspiracy theorists (or good news for stars)

[C M Burns]

No doubt many of you have at least considered after a couple bad beats, maybe the system is biased somehow, favoring all those fish. Well I decided finally to do a little "reseach" on this subject. What I did was keep track of a number of pre flop head up allins to see if they check out.

These hands are all at Stars over the last moth or so. Mostly sng (turbo since i was impatient) but a few from multi's. To qualify two, and only two people had to be in an allin situation pre flop, then i recorded the hands and who won, only about 1/3 are mine, i computed the expected win rate and compared it to observed (tie hands were excluded)

175 hands in total. Expected win probability for favorite was .66, and the favorite won .67 or 67% of the time. You do not need to be a statistician to see that this is right in line with what is expected.

And 175 hands may seem small but it is actually a reasonable sample to make some conclusions. I won't go into detail but this is very much like telling whether a coin is biased, how many flips do you need? Of course it depends how big a bias but 100 flips would be a reasonable amount to detect a .1 bias (e.g. .6 head) There is more variablility in the hands of course but, we are talking hundereds of hands not thousands to make a reasonable conclusion.

However, what about my hands. 61 hands were mine, of those I was favored 69% of the time. Ok I guess since most of these were high blind small stack situations. My expected win was .55, and observed was .46, a considerable diference, not rediculous but one would expect this only about 10% of the time (run of these 61 hands). I guess I should be happy i was slightly profitable over than run. Of course then if you take my hands out the favorite wins much more than expected (.75 to .66).

So the bottom line is that Stars tourney allins seem to be fair for all of you. Whether they are out to get me is still to be decided.

[TylerD]

Good news and good work Monty. Still stings when your opponent spikes his set on the river though /images/graemlins/mad.gif

[JWise]

your post proves nothing and actually leans toward stars having a bias shuffle.

[gabyyyyy]

[ QUOTE ]

your post proves nothing and actually leans toward stars having a bias shuffle.

[/ QUOTE ]

His post really does prove nothing. 175 hands? You also only analyzed when 2-3 players were involved.

[MicroBob]

i agree that you need more than 175 hands to begin to prove the validity of the stars all-in.

certainly there are several out there who's expectation isn't even close to the actual result after only 175 situations.


please note - most people know that i am not generally with the 'deal is probably rigged' crowd.


your experiment is a start....but that is all it is.

i also agree that your own personal results tend to validate the 'conspiracy' gang's suspicions more than it refutes it....but with that few hands, it REALLY doesn't do either.

[daryn]

</font><blockquote><font class="small">In risposta di:</font><hr />
your post proves nothing and actually leans toward stars having a bias shuffle.

[/ QUOTE ]


you realize you just contradicted yourself in the same sentence right?

i agree his post proves nothing, that's about it. i also believe the site is legit, not that it matters.

[JWise]

advice for next time if you want to prove the deal is not bias towards fish youre going to have to identify the fish. I know you wont believe me when I say this but fish sometimes go in as the favorite too. Your conclusion for next time should be:
Fish won ?% when they shouldve won ?%

[JWise]

[ QUOTE ]
[ QUOTE ]
your post proves nothing and actually leans toward stars having a bias shuffle.

[/ QUOTE ]


you realize you just contradicted yourself in the same sentence right?

i agree his post proves nothing, that's about it. i also believe the site is legit, not that it matters.

[/ QUOTE ]

when i say proves nothing I mean proves nothing

[MicroBob]

i agree with JWise.

i think it can LEAN one way without really fully PROVING anything.

although 'fully proving' may be redundant...as i'm not sure it's possible to 'partly prove' something. you either proved it or you didn't, right?? thats a different topic though.

[lorinda]

How does stars know who the fish are?

Lori

[TheNutz]

Lorinda, they have their own special pattern mapping software that tracks all deposits, cashouts, average pot won vs average pot lost, average tournament finish, and the speed of losing the entire deposit they made on payday!

Duhhh ! /images/graemlins/grin.gif

[MicroBob]

"Your conclusion for next time should be:
Fish won ?% when they shouldve won ?% "

on this particular issue i agree with lorinda.

this idea doesnt make a whole lot of sense to me??

weren't we trying to prove that the 'favored' hand supposedly doesn't hold-up as frequently as it is supposed to?
i thought those who suspected a rigged situation at stars believed this to be true because the 'better' hand going in wasn't holding up enough.




i think any efforts to try to 'prove' the legitimacy of the deal are really a moot effort because there are just too many viewpoints among those who think the deal is not properly random.

to wit:
some think it is rigged to keep the fish around longer....

others think it is rigged to keep the fish from playing too much....

while others think it is rigged simply to generate more rake at a quicker rate without regard to whether fish OR winning players stay there longer or not.


additionally - some players think the tourney all-in's specifically are messed up while others think the tourney hands are fine but it's the ring-game hands that should come under scrutiny.

[Pokeraddict]

Stars must be rigged if I can win there /images/graemlins/grin.gif

[C M Burns]

What it more or less proves is that at least allin pre flop, things are as they should be. As for the amount of hands, if there was say a .05 bias in the shuffle/system, where the worst hand won 5% more often than it should. after 175 hands you would get results that "significantly" (in a statitical sense) differ from true chance about 97% of the time. If you wanted to get this up to 99% you could get another 100 hands or so, but this seems close enough for me.

And not that i ever thought it was biased, i just though it was nice to see that things work out how they should.

So the point is this shows that it is very very unlikley (about 95% sure), that stars shuffle shows any sort of bias at least after the flop. I guess you could come up with some other way it could be biased but now the theories must be more complex and involve pattern maps.

And way to disregard something without any explanation for the oposite. And if posts had to prove something to be worth posting here I think there would be alot of empty space.

Edit: and just after i write this i bust out of a stars tourney w/ AQ against 39, i take it all back it is clearly rigged

[C M Burns]

The thing is with all of these theories there must be an overall deviation from chance and anything that shows some consistency with chance to some degree disagrees with the theory. But anyhow proof is always in the eye of the beholder, and some just will not believe and the more people who have these theories the better, since they often will be the ones thinking they are due so they will chance their draw with 1-1 odds and other such plays. hmm, maybe i will post the oposite on RGP.

[MicroBob]

"anything that shows some consistency with chance to some degree disagrees with the theory."

i agree....but some of those don't believe in the validity of the online game will likely disagree.


i admit to not knowing diddly-squat regarding math and/or probabilities....and you seem to have done your homework and believe your claim....but i simply fail to see how 175 hands can even be close to a large enough sample size to prove anything (much less with 97% certainty as you state).


i would think that in a truly random environment it would still be extremely easy for the end results to deviate by 10 or more.

i know you have more 60-40 or even 80-20 situations on the all-ins....but i'm just thinking about flipping a coin....
after 175 trials i would not expect the break-down to be terribly close to 50-50...and if it deviated by 10 or 15 i wouldn't think that would be too far out of the ordinary.

again, I'm no Daryn or Lorinda (or BruceZ or Homer) in matters such as these so it's VERY possible my assumptions are way off or even completely irrelevant.

[Gonzoman]

[ QUOTE ]
i agree with JWise.

i think it can LEAN one way without really fully PROVING anything.

although 'fully proving' may be redundant...as i'm not sure it's possible to 'partly prove' something. you either proved it or you didn't, right?? thats a different topic though.

[/ QUOTE ]

Just to clear some stuff up mathematically, it is impossible to 'prove' anything using statistics. One can only show that the results achieved are highly improbable assuming the null hypothesis is true. What the orignal poster has done here is shown that assuming the deal is fair, the results are well within the range expected.

If someone wants to show that the deal is not fixed towards fish, one would have to make the null hypothesis 'the deal is fixed to allow the fish to win an extra x% of the time'. Then run an experiment with enough data to determine that the null hypothesis is highly unlikely.

[JWise]

[ QUOTE ]
"Your conclusion for next time should be:
Fish won ?% when they shouldve won ?% "

on this particular issue i agree with lorinda.

this idea doesnt make a whole lot of sense to me??

weren't we trying to prove that the 'favored' hand supposedly doesn't hold-up as frequently as it is supposed to?

[/ QUOTE ]

[ QUOTE ]
No doubt many of you have at least considered after a couple bad beats, maybe the system is biased somehow, favoring all those fish. Well I decided finally to do a little "reseach" on this subject. What I did was keep track of a number of pre flop head up allins to see if they check out.

[/ QUOTE ]

It was suppose to be about the deal being bias towards the fish.

[Fred Duke]

I love your experiment. I see two issues that need further elaboration.

You state " Expected win probability for favorite was .66...". This .66 is the key to the whole thing and yet you make no mention of how it is derived. Did you do a Monte Carlo simulation for each of your 175 match-ups? That would clearly be the most rigorous methodology (and time-consuming). Anything less involves a measure of hocus-pocus that queers the conclusion to one degree or another. Please tell us what you did to get this .66 figure.

The second potential problem comes from sampling bias, using only some hands in a session, ignoring some that 'look odd'. It can be intentional or accidental and only the experimenter is able to judge the honesty of the sampling methodology.

As a student of statistics and experimental science I can state for those that doubt it, 175 hands here is a significant sample size upon which one can reasonably draw preliminairy conclusions, all other issues being resolved.

[Fred Duke]

One sane voice in a mob of torch-carring Phillistines.
Thank you.

[C M Burns]

ah good questions. To get this average probability, I computed the probability of a hand winning using twodimes.com, I believe their method is some sort of smulation, e.g. if it was AsQs vs. JcTc prob win = .X, then I averaged these to get the .66, maybe not the ideal way but it seemed reasonable. And you are right there could be some sort of sampling bias, I did try and not choose hands after the fact (e.g bad beats), but before the deal was complete. Of course all I show is that things turn out as they should, and how most would expect them too, so there is no reason to suspect a bias, even though you know how things should be it is just reasuring to play with some numbers and see they work out.

Now as for the person asking how 175 could be enough of a sample. Intuitvley it may not seem like enough, since 175 hands is nothing for gaging your results. Now if I were doing something like seeing if I got AA or KK the right amount of times or how often my set was drawn out on, 175, would be a ridiculously small and meaningless sample. But what I am doing is actually quite different. It would be more like having 175 times you had a set and seeing how often you are outdrawn. And even there how the hand was played would have much to do with the outcome. What I have is 175 hands where the expected outcome is totally dependent on chance (or the lay of the virtual cards). The next part should be more informative.

Essentially what I have is conceptually the same as trying to tell whether a coin is biased. There are some important difetrences like the probability of success is diferent for each trial, but the concept is the same. I have 175 event with a known probability of sucess, this would be like having 175 coin tosses, and you know the expected # of heads is .5.

So how then do you tell if a coin is biased. Well lets say you toss a coin 100 times, record the number of heads, then toss it 100 times again and record the number of heads, you continue to do this an infinate number of times. Now if you plot this "sampling distrobution" what you will get is a normal shaped curve. Most trials will have around 50 heads, but now with this distobution you can quantify how often a given number of heads will occur per trial. So what if you get 60 heads? As it turns out less than 2% of all of those sets of 100 tosses will have 60 or more heads, thus this is a fairly rare ocurence. So as someone said you can't prove anything with statistics but you can quantify its likleyhood. A comon criterion for "significance" or results unlikley to be caused by chance is 5%, meaning something that will ocurr by chance less than 5% of the time. So 60 heads out of 100 is suficiantly unlikley to happen by chance with a fiar coin that you should become suspicious of it's fairness.

Perhaps this does demonstrate somthing about intuitve probability. If I were to say to someone I got 60/100 heads they probably would not see this as all that unlikley, when actually you would get this result a very small % of the time by chance. So much so in fact that you would have to be a "sucker" not to at least be suspicious. For more info on probability and stats cunsult your school library.

Well thats todays lesson, tommorow we will discuss how a bill becomes law. Anyhow, I'm not saying my results are without flaw, it is just a little something that sugests things are as they should be, something to insight furry into those who claim a bias and also to show that I am a dork who like to play with numbers.

[GrannyMae]

One sane voice in a mob of torch-carring Phillistines.

i knew termite season had just started, and i knew you would sniff this thread out like a fallen redwood. however, what's with the philistine comment?
(other than your spelling error)

you are one piece of shiit mr. grand wizard

http://e4u.deltait.com.au/dressed/bek003.gif

[JWise]

[ QUOTE ]
ah good questions. To get this average probability, I computed the probability of a hand winning using twodimes.com, I believe their method is some sort of smulation, e.g. if it was AsQs vs. JcTc prob win = .X, then I averaged these to get the .66, maybe not the ideal way but it seemed reasonable. And you are right there could be some sort of sampling bias, I did try and not choose hands after the fact (e.g bad beats), but before the deal was complete. Of course all I show is that things turn out as they should, and how most would expect them too, so there is no reason to suspect a bias, even though you know how things should be it is just reasuring to play with some numbers and see they work out.

Now as for the person asking how 175 could be enough of a sample. Intuitvley it may not seem like enough, since 175 hands is nothing for gaging your results. Now if I were doing something like seeing if I got AA or KK the right amount of times or how often my set was drawn out on, 175, would be a ridiculously small and meaningless sample. But what I am doing is actually quite different. It would be more like having 175 times you had a set and seeing how often you are outdrawn. And even there how the hand was played would have much to do with the outcome. What I have is 175 hands where the expected outcome is totally dependent on chance (or the lay of the virtual cards). The next part should be more informative.

Essentially what I have is conceptually the same as trying to tell whether a coin is biased. There are some important difetrences like the probability of success is diferent for each trial, but the concept is the same. I have 175 event with a known probability of sucess, this would be like having 175 coin tosses, and you know the expected # of heads is .5.

So how then do you tell if a coin is biased. Well lets say you toss a coin 100 times, record the number of heads, then toss it 100 times again and record the number of heads, you continue to do this an infinate number of times. Now if you plot this "sampling distrobution" what you will get is a normal shaped curve. Most trials will have around 50 heads, but now with this distobution you can quantify how often a given number of heads will occur per trial. So what if you get 60 heads? As it turns out less than 2% of all of those sets of 100 tosses will have 60 or more heads, thus this is a fairly rare ocurence. So as someone said you can't prove anything with statistics but you can quantify its likleyhood. A comon criterion for "significance" or results unlikley to be caused by chance is 5%, meaning something that will ocurr by chance less than 5% of the time. So 60 heads out of 100 is suficiantly unlikley to happen by chance with a fiar coin that you should become suspicious of it's fairness.

Perhaps this does demonstrate somthing about intuitve probability. If I were to say to someone I got 60/100 heads they probably would not see this as all that unlikley, when actually you would get this result a very small % of the time by chance. So much so in fact that you would have to be a "sucker" not to at least be suspicious. For more info on probability and stats cunsult your school library.

Well thats todays lesson, tommorow we will discuss how a bill becomes law. Anyhow, I'm not saying my results are without flaw, it is just a little something that sugests things are as they should be, something to insight furry into those who claim a bias and also to show that I am a dork who like to play with numbers.

[/ QUOTE ]

I never had a problem with your sample size my problem was you stated you were testing to see if the deal is bias towards fish

what if the favorite only won 40% of the time will you then be able to conclude the deal is bias towards the fish?

[Fred Duke]

I'm still just an Associate Wizard.

Don't let that torch go out in the wind.

[Fred Duke]

Looks like solid work. Two Dimes might enumerate all boards yielding the exact odds, not a simulation. Either way your .66 number won't be off by much.