2 Dire

Difficult or impossible to implement such balancing.
The reason is all-ins like 20% (fish) vs 80% (tag) happen much much more often than 80% (fish) vs 20% (tag), or 80% (tag) vs 20% (tag).

In general, fish goes all-in with worse hand than tag. That's why the site slightly increases probability of wining 30%/70% and alike all-ins. This is confirmed by the stats provided by spadebidder.

Tag vs Tag all-ins happen vastly more often than tag-fish all-ins, so it's easy to dampen the aggregate. I have not examined the stats provided as I felt they were meaningless and no site who did decide to rig their games would be foolish enough to not take some sort of effort to balance the aggregates - it's not very difficult. That was also a very simple example, there are much more sophisticated means of aggregate balancing. If there are anomalies even in the aggregates than that's just comical.

Dire,

The problem with your "fish run good" belief structure is that the conditions can always be expanded at will to include more rigged theory variables.

Sure, many believe that fish run good, but then many believe in all sorts of wacky things. Even if you are shown a lot of evidence to the contrary, all that will happen is that this "fish run good" theory will be adjusted to say "well of course Stars will know that we think fish run good so they will let a few run bad to be safe" or "only certain types of fish are allowed to run good"

http://forumserver.twoplustwo.com/54...8/index59.html

shows tons of fish running bad long term and good players running good long term.

The best stats can do is crunch a lot of raw data, it really cannot adapt that well to the selective whimsical beliefs of those who want to say that sinister forces are at work. If the raw data shows that the game is fair, then it becomes the responsibility of those who believe in the mystical to prove their beliefs, otherwise what is the point of any statistical study.

If you do a study of coin flips, and show that a coin is fair, how do you address the person who believes that while the overall results seem normal , what really happened was a rigged coin to come up tails more often was used part of the time, while another that was rigged heads was used to keep the results in line. You make them prove their belief, otherwise it is just wild conjecture.

This is the flaw with nearly all rigged beliefs, they rely on blindly believing unprovable conjecture, while even ignoring (or explaining away) actual real data.

People can certainly believe whatever they like, but eventually to get others to accept these beliefs some proof needs to be shown.



Your assumption is flawed. Eventually a greater breakdown of intervals will be useful, but what likely happened is that within certain intervals most of the data fell on the extreme end of the interval based on common all in situations.

So probably in the [0.54 - 0.56) and [0.44 - 0.46) ones (which are basically the same hands, there were probably a lot of overcard/pair situations that fell on the edge off the intervals. So your assumption of an average is likely off, but this will be corrected if and when they tighten up the intervals further.


Good luck ever finding a statement like this in one of spade's posts. Note how his are all about data and facts without any agenda (he would be happy proving it is rigged or is not rigged).

In contrast, you clearly have an agenda, and while I give you marks for creativity in attempting to use math in your quest, the problem is you are making a lot of bad assumptions, using a lot of irrelevant math, and you have a clear agenda with regards to how you want the data to come out.

As I said earlier, your use of math unintentionally provides a charming contrast when compared to spade's techniques.

Card matchups have discrete equities and the number of matchup types isn't that high. There's no reason to assume these ranges should all be centered. We also didn't count multi-player all-ins or ties. We just need to do it even finer grained than this, and fix some other problems with the analysis that I've identified. But statements that there shouldn't be some small outliers in these 50 ranges are just ridiculous. The sample sizes are relatively small at this granularity.

You have to admit your stats don't prove anything.

1) Both wins and ties must be counted to compare to equity correctly.
2) Actual average equitie within an interval must be provided in order to estimate deviation. No matter how fine you grain intervals as the result is very sensitive to the input equity.

I completely agree. I'm working on perfecting the code.

I told that before, I just dont know how to explain.

First it's not my belief structure. But at the same time, I am a cynic and have no reason to expect integrity from effectively unregulated sites in a legally grey industry. So while I don't quite have a tin foil hat on (in fact, poker is my profession), I wouldn't be the least bit surprised to see certain failures of business ethics occurring. And I happen to love math and statistics, so this is fun for me. And hopefully I help to bring a bit of science to both sides. It's insane nobody mentioned, as I did above, that the 'experiments' conducted in this thread are completely worthless for proving a fair deal, although they could prove an unfair deal.

All that aside, there's a strictly finite limit on the number of practical ways to segregate the data. Player type, profit rate, account date of creation, hands played off the top of my head cover the vast majority. Aside from more direct segregation such as stakes/etc, this seems to cover pretty much everything. Although obviously some of those categories have subcategories, but again everything is finite for the analysis of practical and reasonable categories.

Fine. The belief structure you highlighted, and seemed to suggest is possible.


Always a good idea to be a cynic when it comes to protecting your money, but realistically and historically bad things that happened in the online gaming industry tend to fall in the more mundane. Insiders cheating players due to a badly designed security system. Small poker rooms/casinos not paying out cashouts or honoring their terms and conditions. Cheating from other players in the form of collusion, or fraud based transfers.

You know - simple basic crime. Sadly, it happens all the time. The more people who are vigilant about these things the better.



I would assume the default to start the work is assuming the deal is fair and then try to show how it may not be via statistical means. Trying to prove the deal is fair based on everyone's wild beliefs would be impossible. Some think over cards beats pairs too often, some think pairs beat over cards too often. Proving both might be tricky...

Technically finite numbers, but realistically there is no limit to the creative beliefs people have in this area. Even how one defines "fish" will vary from person to person. What about the deposit curse/boom switch beliefs. The beliefs that people get screwed when finishing a bonus. What about people who create an account but do not play on it for a long time.

Seriously, you could make a thread asking for what "categories" should be tested and it would never end. That is the innate problem with that type of belief structure, it can never be fully proven or disproven, which is how some conspiracy people end up with some fairly creative ideas (one suggested "government mind control" to me recently - good luck disproving that).

We could spend all day trying to disprove all forms of online poker style "mind control," and while that may be a bit entertaining, it is also a huge waste of time that could be better spent studying more common sense and realistic possibilities.

Still, since you are math inclined, why not propose a study that would somehow analyze all of your categories and sub categories. The freaky stats guys here are pretty good at helping fine tune these things (as can be seen in the stats forum)

Monteroy, you do start with the assumption the deal is fair then you test it. If it is fair then there will be no problems. When you do find problems and discrepancies, of which plenty have already been discovered, then you try to examine why those problems might exist.

But it is very time consuming, and in the end meaningless work. The odds of the current main sites (Full Tilt and Stars) still being relevant within two years, and probably even one, are in my opinion extremely low. It would be like now spending hundreds of manhours analyzing paradise poker. Who cares?

This is incorrect. There are technically infinite categorizations. There are, in reality, finite. If somebody finds out that players with 8 characters in their name run better than players with more or less than 8 characters in their name then um... ok.

Well the problem then is you still fall within the "what if" theory without being willing to follow up. I realize you are not the same person as one of the OMG my AA lost twice RIGGZORs crowd, but in the end a "truth is out there" style of what ifs without follow up are not much more then fights of fancy.


Start a thread. Watch the responses you receive. Many will make your tongue in cheek comment seem logical by comparison. That's the flaw of the what if belief structure, test it if you like (much easier to do , just start a thread ).

Good luck even imagining how some of the replies you receive would be tested.

do some monte carlo tests on the all-in confrontations

that will eliminate the clumping issues you guys are discussing

I dont know much about math so i could be wrong but the results look very strange.

If you get your money in bad more often (fish) you do MUCH better than your average equity

If you get your money in good more often (good player) you do much worse than your average equity.

Doesnt this kind of back up the people who say fish win more often than they should.

What am I missing?

i have never taken any statistics course before but these results did surprise me somewhat. granted this may be absolutely worthless because of my lack of knowledge of stats but i was really expecting the numbers to be a lot closer with a 100 million hand history.

[0.20 - 0.22): 2168 516 0.238%

wouldnt this be a pretty large deviation for 100 million hands? i mean we keep telling ourselves that we are running bad and that variance will eventually play itself out. but i would expect this to happen before 100 million hands.

And i bet if there would be anyway to organize data by player types wining rates or something similar deviation would be much bigger.

w/e still rigged

How is it with stack sizes..did shorties win the more / less.

To rigg its not important that preflop all ins win the correct ammount of time.
Its more important to increase the rake.

See, to this person it does not matter if players are fish or not, or when they opened their accounts, or their profit rates. His mystical forces at work involve "setup action hands" in which he has complete faith.

Everyone who thinks it is rigged will have their special "prove this wrong" belief structure. No real way to satisfy them all (or likely any of them), so perhaps for now a brute force proof that the hands dealt are not rigged will be a good start.

If every single hand in the sample was at 22% equity then the chances of it scoring at least 23.8% over this size of sample is 2.38%.

If every single hand in the sample was at 20% equity then chances of it scoring at least 23.8% over this size of sample is .001%.

So somewhere between extremely unlikely and 99.99% [literally, hah!] unlikely.

By the way, this math is not very difficult. It is extremely accurately estimated by just the basic binomial formula. You have n, which is your sample size. In this case n=2168. You need to test for a certain range of 'hits', at least, more than, less than, etc. In this case it makes sense to measure "at least" since the expected value is less than the observed value so we want to see how unlikely this is. In this case the number of observed hits is .238 * 2168 = 516. And now you have all you need - as there are calculators that can do the binomial for you, such as this one here: http://www.stat.tamu.edu/~west/apple...omialdemo.html even though it's written by an aggie, it seems to be not entirely broken.

So now you can go there and enter it yourself:

n = 2168
p = .2 (or whatever probability you want to test)
measuring "at least" 516

Click compute and you have your probability. Now you can test it yourself for whatever other cases you want to inspect.

That group had one of the smaller sample sizes, but I guess an interesting question would be what would the odds of one of those intervals among them all (assuming no other explanation can be given) resulting in this.

Seems if you just focus in on a single one after the fact and draw conclusions (not saying you did) from that in isolation that would be a bit misleading.

I certainly can understand wanting to look at any results like that in greater detail with a larger sample size, but I would imagine that the odds of one of the results being a bit away from expected among all of the intervals, particularly among the smaller samples, is not as extreme.

It is impossible to prove that the hands dealt are not rigged. It is only possible to prove that they are rigged. There is absolutely no reason not to run his theory. If the site is not rigged then every single test you could think of will turn up with a fair distribution. Once you have ran everything you can think of then you can reason that it's extremely likely that the deal is fair - but it can never be proven. There's no such thing as a "brute force proof that the hands dealt aren't rigged" which is why the title of this thread is nonsensical.

And this is a statistical fallacy. You are suggesting a conditional probability calculation (eg - given that this interval was 'normal', what are the odds the next one will be too?) but this is nonsensical. Symmetry aside ( you don't need the sample numbers from .5 and up), the samples are completely independent and so a conditional calculation makes no sense.

It'd be like if you've played the lottery every day for the past 30 years, and were convinced that your odds of winning the lottery the next day were way up there because the odds of you losing every single day for 30 years is just so low! Or the more classic example being you see the roulette table just had 10 blacks in a row so you go put your life saving on red since the odds of 11 blacks in a row is extremely low! In fact it has a probability of somewhere around 1/20th of one percent of happening, but that doesn't mean that a red is now 99.95% likely.

EDIT: Cliff notes: 2.38% is 2.38%.

Yeah yeah, I am getting a bit turned around here and I was 16 tabling when I wrote that (couple bad beats later solved that, damn rigged Stars...)

My main point is what I said earlier, that the possible existence of special forces at work, whether it is action hands, when a person deposits, type of players etc all kind of fall in the same realistically unprovable category.

They are either subjective definitions (player types), whimsical beliefs (ooh scary action hands), or data that is nearly impossible to even obtain in sufficient quantities (when a player created an account or instance).

You dismiss his action hand belief (so do I obviously), but seem open to when a player created an account as a possible factor. I look at that in the same way as action hands in terms of common sense. Both lack any for a huge successful company like Stars to even consider.

However, as I often say, people are always entitled to believe whatever they choose.


Not what I am asking.

What I asked was if you have 40 intervals for instance, what is the odds of one of them falling outside the range.

No idea why you think I suggested it had anything to do with the the other intervals results.

Using your lottery example, if you play every week, the odds of you winning in any specific week is the same, but if you know in advance that you will be playing every week or a year, the odds that you win are certainly greater then your chance in any specific week.

Or a less extreme example. Pick a number between 1 and 100 each day for a month. What are the odds you will get the number right on one of the days (looked at in advance of any picks). Certainly better then 1%, but I am too tired to figure it out, so you can if you like :P. That is what I am asking about the interval results.

Account creation would be logical. It's obviously in the site's best interest to get as many people 'hooked' on poker as they can. It seems reasonable to expect that people who run hot early in their experience with poker will be more easily hooked as they overestimate their ability at the game and can blame future failures on bad luck.

But I don't really think it's poignant to argue why or why not one theory makes sense. We have everything we need to test even the most obscure theories, and if one person thought it a reasonable way for a site to rig the games then somebody else probably thought so as well - and you get closer to satisfying most of everybody, or proving the site rigged depending on the results.