I don't agree with that last statement. Nobody has all hole cards but the poker sites (and Cigital in that case, but those aren't public). The databases I speak of have only the cards that went to showdown, plus the community cards, plus the hole cards of the player gathering the data in some cases. However, with a large enough sample of hands that include 2 or more player's hole cards, you end up with enough hole cards to do some very useful analysis, including many queries that would show skew in the randomness of the deal.

Sure, there are some things you can't analyse with this kind of sample. But you don't have to see hole cards to prove that the community cards are random, or to prove that all-ins win at the expected rates. Both of those things have been analysed already in some large samples without any sign of rigging. And there's a lot more you can tell with this kind of samples.

tk1133, which court is hearing a child molesting case against you again?

zOMG - it's RIGGED!!!!

... but in my favor, to the tune of +.16BB/all-in hand.





Regardless, I'm still not convinced, since this is pretty much the only thing that can be analyzed statistically and excludes other aspects of rigging, such as cold-decks, hole-card readers, board-card predictors, and other forms of cheating such as collusion or bots. Also, since I play tournaments almost exclusively online (to minimize the chances of collusion) whenever it really counts I can never seem to win a pot, such as on the money or FT bubble, or the hand that would put me among the chipleaders.

Although my local poker room has an automated shuffler that seems to deal out big pairs more often than is called for (shudder), I'd rather play there and make decent money.

- Bitterness is joy with a bad attitude

For some reason the images arent coming out above and I get a javascript error when I try to edit, so here they are...

http://yfrog.com/06bbgraphj

http://yfrog.com/06tgraphj

I'm not sure the bolded statement can be made with confidence at the moment. I'm assuming you're referring to IndianaV8's analysis of pre-flop all-ins posted in his research thread. Here are the frequencies of that analysis (quoted from your earlier post)

A cursory glance at this seems to support one of the more popular rigged theories, namely that the underdog in a hand wins more often than he should. The results in all of the underdog ranges are at or near the top of the range. None is in the bottom half of the range. Without knowing anything else about poker, we would expect the averages to be in the middle of the ranges, not at the top. We'd also expect some to be in the bottom half and some in the top.

Your explanation for this:

If a 77/23 matchup is more or less the only thing we can expect in the 20-30% range, shouldn't we expect the result to be close to 23 rather than 29, which is near the very top of the range?

Also, in the 10-20% range, pair vs. pair is not the only common matchup we can expect. Pair vs. two undercards is also not rare (at least at the microstakes, where we can assume most of the hands in this database occur), and that matchup is 84/16, almost exactly in the middle of the range. But the result for the underdog in this range 20%, the top of the range.

I think the problem here is that the ranges are too big. We can only speculate about the matchups in those ranges, and that's not very scientific. Reducing the ranges to, say, 2% spreads instead of the current 10% spreads would eliminate most or all of the bias in the sample.

Another approach might be to sort by hand category rather than equity ranges. Then you'd have something like pair vs. underpair, pair vs. two overcards, pair vs. card of its rank and one undercard, etc. You might even investigate things like pair vs. A + one undercard.

As it stands, though, the study doesn't really tell us all that much. There's too much information hidden in the wide ranges.

true ...

You made some good points.

I'm asking for another run split by 2% ranges, if I can get him to do it. I think at that granularity we can easily fill in the hand categories based on the equity (and do so more carefully).

Thanks

sorry but happens to me too..

always the same answer

You are 100% correct Okiehustler,I wish everyone for once would stop the name calling and insults,if you cant have any input on a positive note keep quite.Lets throw this out there,doesn't the fact that there are concerns being raised by thousands of people about online poker being rigged say something.I have never seen any blogs from people saying boy do they cheat you at the Rio,or boy do they cheat you at Bally's.Just the fact that people are complaining in the numbers that are out there says something.

Yes.

It says that there are thousands of poor poker players with equally poor maths who do not want to take responsibility for their own poor play and instead like to delude themselves that on-line poker is rigged against them.

Rather than simply stop using a service that they believe is cheating them they continue to do so and post brainless whines on poker forums.

Indeed, it says that there are thousands of poor poker players with

Oh, never mind.

People are more likely to be losing players online and will usually have played more hands and had more completely normal runs of bad luck which they cant deal with. Its also much easier to convince yourself that a faceless overseas online company would rip you off, even though being cheated live in one way or another seems much more likely

Ten million smokers can't be wrong!

Nor can many millions of lemmings.

No, wait ...

Number of hands parsed: 102,661,557
Total preflop all-in situations: 535,950 (Once per 191.55 hands).

[Preflop Equity %): #All-ins #Wins Win%
[0.00 - 0.02): 0 0 0.0%
[0.02 - 0.04): 0 0 0.0%
[0.04 - 0.06): 46 2 0.043%
[0.06 - 0.08): 10202 682 0.067%
[0.08 - 0.10): 2071 152 0.073%
[0.10 - 0.12): 1722 197 0.114%
[0.12 - 0.14): 9424 1274 0.135%
[0.14 - 0.16): 3858 608 0.158%
[0.16 - 0.18): 7436 1338 0.18%
[0.18 - 0.20): 93772 17905 0.191%
[0.20 - 0.22): 2168 516 0.238%
[0.22 - 0.24): 2325 547 0.235%
[0.24 - 0.26): 29000 7304 0.252%
[0.26 - 0.28): 26160 6815 0.261%
[0.28 - 0.30): 44814 13266 0.296%
[0.30 - 0.32): 40589 12854 0.317%
[0.32 - 0.34): 23314 7631 0.327%
[0.34 - 0.36): 22147 7530 0.34%
[0.36 - 0.38): 18191 6784 0.373%
[0.38 - 0.40): 18082 7091 0.392%
[0.40 - 0.42): 19580 8061 0.412%
[0.42 - 0.44): 47204 21103 0.447%
[0.44 - 0.46): 52790 24259 0.46%
[0.46 - 0.48): 38580 18403 0.477%
[0.48 - 0.50): 21610 10650 0.493%
[0.50 - 0.52): 23340 11825 0.507%
[0.52 - 0.54): 38566 20171 0.523%
[0.54 - 0.56): 52804 28537 0.54%
[0.56 - 0.58): 47203 26100 0.553%
[0.58 - 0.60): 19581 11520 0.588%
[0.60 - 0.62): 18082 10991 0.608%
[0.62 - 0.64): 18191 11407 0.627%
[0.64 - 0.66): 22147 14617 0.66%
[0.66 - 0.68): 23314 15683 0.673%
[0.68 - 0.70): 40589 27735 0.683%
[0.70 - 0.72): 44814 31548 0.704%
[0.72 - 0.74): 26160 19345 0.739%
[0.74 - 0.76): 29000 21696 0.748%
[0.76 - 0.78): 2325 1778 0.765%
[0.78 - 0.80): 2168 1652 0.762%
[0.80 - 0.82): 93772 75867 0.809%
[0.82 - 0.84): 7436 6098 0.82%
[0.84 - 0.86): 3858 3250 0.842%
[0.86 - 0.88): 9424 8150 0.865%
[0.88 - 0.90): 1722 1525 0.886%
[0.90 - 0.92): 2071 1919 0.927%
[0.92 - 0.94): 10202 9520 0.933%
[0.94 - 0.96): 46 44 0.957%
[0.96 - 0.98): 0 0 0.0%
[0.98 - 1.00): 0 0 0.0%

Weevil - here it is. I marked the three paired ranges that are just slightly outside the expectation. And this is 50 ranges this time.

This one modified the algorithm to determine when at least one player is all-in preflop, so the total number of all-ins shown went up slightly.

So we have three ranges slightly off from expectation:

1. Of 2017 90/10 - 92/8 all-ins, the favorite won roughly 1.7% more than expected (from midpoint).

2. Of 2168 78/22 - 80/20 all-ins, the underdog won roughly 2.8% more than expected (from midpoint).

3. Of 47,204 56/44 - 58/42 all-ins, the underdog won roughly 1.7% more than expected (from midpoint).

I'll have to look at the likelihood of these outcomes, but with the small sample (due to narrow range) I think they are quite likely. I think it's pretty impressive that all the others are inside the ranges.

lol @ agreeing with guy who posted like 2x several months ago. Already establishing yourself as clever fellow who's opinion should be carefully considered.
that's 'cause those folks don't know how to use a computer, which is of course why they play live. but believe me, they're bitching about how there's no way they could be that unlucky at B&M cardrooms too. I think they just blame unlucky chairs and witchcraft.

Some binomial calculations for the three anomalies:

1. The chance for a 92/8 favorite to win at least 1919 of 2071 trials (92.7%), is 14.2%.

2. The chance for a 78/22 underdog to win at least 516 of 2168 trials (23.8%), is 2.4%.

3. The chance for a 56/44 underdog to win at least 21103 of 47204 trials (44.7%) is ~1%.

These are interesting. Weevil?
I've asked Indiana to drill down on these ranges.

I have not taken any statistics courses (my understanding of it is limited to the poker books I've read), so I might be looking at this the wrong way, particularly my second point.

But if I'm correct, there are two (well, three if you count "rigged") explanations for this: (1) The 2% range of equity situations is still two wide and is showing its discrepancies by including too many hands towards the top/bottom ends. (2) If you look at it as one piece of a bigger picture, just like you should look at individual poker hands ("Zomg he hit his one out on the river when we were all-in on the turn.") As you can see, the majority of the hands are within their ranges, and as sample size increases, all should converge to be within those ranges.

I think the word you're looking for is "nerdy."

I agree, and just two 1% or 2% "hits" on this one set of samples isn't all that surprising. I'm more impressed that the other 47 ranges are inside expectation.

The results do look remarkably close to their ranges. The few that are slightly out of whack don't raise any flags with me since in a collection of 50 independent data points, I would expect a few outliers. It would be more surprising if we didn't have a couple of oddballs.

One interesting pattern I noticed was that the underdog in each range is more likely to have a result above the midpoint of the range than below it. Ignoring the first three ranges (the first two have zero occurrences and the third has a sample size of 46 -- not big enough to consider), there are 22 ranges. In 15 of the 22 ranges, the underdog scores above the midpoint.

This isn't a big red flag or anything; it's just interesting. If I were the one doing the analysis, I might take a quick look at it, but I wouldn't spend much energy on it.

All in all, I would say the results from this more finely grained analysis should be pretty reassuring to anyone with a fair mind. I haven't done the math, but just casually browsing through the numbers, nothing looks out of line to me.

Well done, spade and Indiana!

I may try to figure out the matchup types that can fall in each range, and since they are discrete that could explain some of it. There are surely some equity amounts that never occur if you look at a fine enough granularity, or put another way if we used 0.5% equity groups instead of 2%, I have a hunch some more of them (besides the extreme ones) would have zero hands in them. But overall, I agree this is a pretty unremarkable distribution that doesn't show anything to be alarmed about.

And there is no doubt whatsoever that a skew of this tiny amount would never be detected by observation alone, such as we hear in this forum all the time. It also would not alter the site's profit as it isn't enough to affect player behavior (pot size) or win rates or survival times.

omg .. was running so good in 22, in top20 most of the time, than i get this AK mp, i raise got reraised. I think for a while and decide to call. Flop Axx 2 clubs, i check he shoves i call. He had QQ .. Q on the river. Ok, no big deal its only 1/3rd of my stack. Few hands later i get QQ, a small stack shoves, i reraise all in everyone folds, he has 77 good i thought, i have him dominated .. But **** manages to turn the seven .. WTF ?? Ok never mind, its only 200 people left till money and i am still in top600. 4 hands later i get KK in sb then i told to my friend look this is my last hand. Early pos raise, i reraised he shoved and i called. Ofcourse, he has an overpair. So, from being in top20 to going out in less than 30 hands. Not bad ..

It is .. I am still lol . But i am kind of got used to it.

just curious what are the 46 hands in the 4%-6% range... 3 way all ins?