Cheaters in our Midst?

[VivaLaViking]

Presented here is data from a Poker site that declares themselves as the largest. The data is of two conditions, A and B, of a greatly different number of hands played, A << B.
<ul type="square">
Condition A

N Hands 2535
N Pairs 153

Mean Theoretical Pairs 11.4
Mean Observed Pairs 11.7 +2.60%

Mean Theoretical Sets Flopped 1.23
Mean Observed Sets Flopped 1.30 +5.79%

Pair Count Win% Sets Flopped Sets Expected
A 16 87.5 1 1.72
K 11 63.6 2 1.18
Q 5 80.0 1 0.53
J 6 50.0 0 0.64
T 11 54.5 1 1.18
9 12 50.0 1 1.29
8 11 27.2 0 1.18
7 13 53.8 2 1.40
6 9 33.3 2 0.97
5 11 18.1 1 1.18
4 14 14.2 3 1.50
3 15 26.6 2 1.61
2 19 21.0 1 2.04

Condition B

N Hands 18044
N Pairs 1040

Mean Theoretical Pairs 81.6
Mean Observed Pairs 80.0 -2.01%

Mean Theoretical Sets Flopped 8.79
Mean Observed Sets Flopped 6.84 -22.1%

Pair Count Win% Sets Flopped Sets Expected
A 84 86.9 6 9.05
K 82 75.6 7 8.83
Q 79 63.2 4 8.51
J 83 62.6 6 8.94
T 82 47.5 8 8.83
9 73 39.7 5 7.86
8 86 38.3 6 9.26
7 87 25.2 7 9.37
6 77 31.1 8 8.29
5 70 15.7 5 7.54
4 73 17.8 7 7.86
3 85 15.2 9 9.15
2 79 11.3 11 8.51
[/list]
You are invited to comment on this please but I know a site that I will not be playing on this weekend. This evidence is conclusive to me. I appologize but names must be withheld as I am talking with a US Senator about legislation to allow jurisdiction in the US Federal Courts.


I will gladly share the software that I used to generate these conditions but you will need to preserve your hand hitoriy files.

TY

[LetYouDown]

You're basing this judgment on 153 and 1040 examples? I never calculated the Standard Deviations here but I'd say this is well within the realm of reason. Are these only hands that actually went to showdown as well? That would enormously skew the results. There are a ton of factors involved here that need consideration.

[VivaLaViking]

No, they are actual hands played where a flop was seen. It is independant of anything I may or may not have done. Condition B is over 18K hands played showing a 22% lower than expected set only flopping. Now some clearer understandind of yesterdays discusion is obtained and thanks.

[LetYouDown]

18K hands...1040 pairs. The number of hands is essentially irrelevant.

[VivaLaViking]

The number of hands was added to show that it was a reasonably large sample and the software calculates it. This was a copy/paste job. Or are you believing that the
-22% is meaningless? It is possible to flip 100 heads out of 100 trials but I would look for other answers if it happened 'cause thats the type of guy I am.

TY

[LetYouDown]

18K hands is only relevant for your calcuation for the percentage of pocket pairs. This is pretty close, especially considering people obviously fold pairs preflop from time to time.

Once you're calculating the set percentage, 18K goes out the window entirely. You're only dealing with 1040 samples.

Based on your numbers, I come up with about 8.56% flopped sets. This is low, certainly...but I'd bet it's well within the standard deviation for this many hands.

On a complete sidenote...who is cheating?

[VivaLaViking]

Are you asking what site these trials were taken from? If so if I had an email address I'd send you an email. BTW: Condition A was tounament hands and Condition B were cash games. Also, the way hand hisory file are written the flop can be seen if you are in or had folded. This would show up in the Win %. Thanks.

[Tom1975]

Let me get this right, out of 1040 trials, the expected value is 114, and the observed number is 89, and you think think this constitutes strong evidence of impropriety? You then compare this to flipping a coin 100 times and getting 100 heads? Can't you see the basic difference here (1040 trials, 114 expected, 89 observed VS. 100 trials, 50 expected, 100 observed). Hell, in your post yesterday you couldn't even calculate the odds of flopping a set on your own. Take a couple of basic courses in statistics and discrete math and quit wasting our time.

Cheaters in our Midst: No
Idiots in our Midst: Yes

[VivaLaViking]

Tom,

I did not compare it to flipping 100 heads, I simply said it was in the realm of statistical possibility for someone to do this. I further believe that if you run the software on you own hand history files that your results will not vary much from mine.

[elitegimp]

I think you mis-added, buddy... by my count you flopped 89 sets out of 1040 paired hands, so that's 8.56%


edit: my bad, you divide everything by 13 to average it by pair... so you expect 112 flopped sets and only got 89. More commentary coming

edit 2: Okay, with a 10.77% chance of flopping a set, and 1040 attempts, your standard deviation is right around 10 (std deviation = sqrt(n*p*(1-p)) = 9.9995), so being off by 23 flopped sets means you are 2.3 standard deviations away from the mean. This happens roughly 1% of the time, so you're sort of on the border of statistical relevance.

[LetYouDown]

[ QUOTE ]
I further believe that if you run the software on you own hand history files that your results will not vary much from mine.

[/ QUOTE ]
That's the first thing you've said that I've agreed with.

[VivaLaViking]

This was all done by computer but it didn't mis-add. 89 sets / 13 value card is ~= to a mean of 6.84. Thanks

[VivaLaViking]

Thank you for the post. Any ass kicking I get here is a minor percent of what's in store for me so it preps me a little. I wanted to put the standard deviation in but I haven't had time yet. I only put a print routine for some of the pair data. Being on the verge of statistical relevance is like geing a little pregnant. There is other data that I can see but no print routine is enabled, like flopping 2 or three cards with a held suited A. When all the data is viewed cummatively even if statistical significance is not acheaved individually, it tells me to play poker elsewhere.

[mosdef]

[ QUOTE ]
Are you asking what site these trials were taken from? If so if I had an email address I'd send you an email. BTW: Condition A was tounament hands and Condition B were cash games. Also, the way hand hisory file are written the flop can be seen if you are in or had folded. This would show up in the Win %. Thanks.

[/ QUOTE ]

why would the site intentionally flop a reduced number of sets? intuitively it would make more sense to for them to increase the number of good hands to drive up betting and collect more rake.

[elitegimp]

[ QUOTE ]
Being on the verge of statistical relevance is like geing a little pregnant.

[/ QUOTE ]

hehehe. What I meant is that most statisticians consider either 1% or 5% to be the cut-off for "statistically relevent deviation" (at least according to several of my undergrad stat professors). I think you need to collect more data and hope that your deviation from the mean goes up or down (i.e. does not stay the same). And by "more data" I mean another couple hundred hands, not "let's see what happens with the next 5 pocket pairs."

Also, it seems like you are biased towards showing that this online site is rigged. That's a very dangerous mindset to have because you'll look for evidence that agrees with you and ignore evidence that disagrees with you.

[mosdef]

[ QUOTE ]
Being on the verge of statistical relevance is like geing a little pregnant.

[/ QUOTE ]

this is not true. there are measurable degrees of statistical significance, but pregnancy is digital - you are or you are not

[Stephen H]

You really need to include your method of selecting hands to be in your results there. Realize that your selection criteria is going to be hammered the hardest to insure that it doesn't introduce any bias, so if you're hoping to get some idea of what you're going to face, you should probably reveal that now.
You'll also need to at least mention any other factors that may affect results. For example, when AA goes to a flop, is it more likey that another A was dealt to a player than if it doesn't see a flop? (I don't know the answer, and it's probably an interesting discussion point, but that's beside the point) Given issues like that, you really should strive for a much larger sample size; I'd want at least something statistically significant for each rank. 1K/rank, or 13K pairs, would be a good start.
The point being that a few raw numbers in a chart compared against statistical averages isn't really enough to get excited about; and more than that, it's certainly shouldn't be enough to get legislation passed.

[Nottom]

OMG online poker is rigged!!!

I'm surprised it took so long for someone to finally figure this out. I'm gonna go cross-link this to the zoo immediately.