I cannot win on pokerstars. Reasons ??

[sethypooh21]

Rake is like 2-3% in the PS 5/10 6max game, sir.

[Suigin406]

just got done reading this thread as i had mostly dismissed it as whining, but then it got long...

i'm not sure who the bigger tool is, OP or this kid reallygay??

[realjaydub]

[ QUOTE ]
just got done reading this thread as i had mostly dismissed it as whining, but then it got long...

i'm not sure who the bigger tool is, OP or this kid reallygay??

[/ QUOTE ]

The biggest "tool" may be you for using that word

[Suigin406]

^^nah, it's a common word on these forums and aptly describes u reallygay...u insist on starting with people, completely ignored the advice people gave u like adan, nor can u seemingly use the search function...

[Suigin406]

also, plz look at ur name and rethink the tool assessment i gave u

[sputum]

[ QUOTE ]
THE ONLY REASON U CAN'T WIN IS CUZ U DONT HAVE A CHANCE WHEN ITS RIGGED AGAINST YOU COZ YOU IS TOO GOOD. WAKE UP.

[/ QUOTE ]

[washus]

I'm new here. Just wanted to add a few comments about the data posted here. Please keep in mind I'm NOT a poker player.

You'll probably think I am, but I'm no crackpot. I work with PRNG and Monte Carlo simulations (online poker deck shuffling could be classified as one of those).

First of all, thanks for the data! Also, I read "standard deviation" here, which is WAY better than the other posts I've been reading. Also some nice z plots!

The data I'm talking about is that 11k specific pair count.

Anyway, someone said "Standard error is 22.4. 555 is 2.37 standard errors from expectation. A deviation that big would occur ~1.8% of the time by chance. A little less often than hitting a one outer on the river. So far, meh."

At least for me, 2.37 sigma is NOT small. That 1.8% (didnt check it for a normal distro) means that ONLY 1.8% of samples of this size (~11k) would be expected to be this far from the mean! That means that 99.2% would NOT be this far.
That means that I would have to see other 99 sets of data as big as OP's to say, ah, ok, OP's was a fluke.

I suspect that if you all analyze your own sets of data for AA pairs you'll see the same trend. Maybe (probably)less than OP's but consistently higher than expected. Don't forget to check the standard deviation! My own set shows the same trend (its very small, I play for only a week, and unexpectedly not losing, actually winning a little.)

There's a concept called "confidence level" (google it). That's when you put limits to one (mainly both) sides of the distribution and say "here and no further". That 1.8% is beyond the 95% confidence level. So, at least for me, that shows that the probability of the site being rigged is VERY, VERY high. Note that the excess is greater on AA, the "action" hole cards. Also notice (someone's graph above) that the excess tends to raise for "good" pairs.

The normal distribution (and Poisson) can be used an approximation for the binomial distro, which is the distro that actually describes the expected number of pairs dealt.

For the binomial distro, the probability that you have (binomial test) 555 AA or more pairs in 11k events is 0.9% (ok, the factorials are too big for direct computer calculations, but thats a good approx.). That's even wilder! Now I wanna see 99 other sets of 11k that show lower deviations from the binomial hypothesis.

Some other tougths:

1-People say it's impossible to say online poker is rigged. 100% certainty yes. That's your confidence level.... 95% is very good, at least for me.

2-What about the time series (taking the order of the events into account)? Ex: I toss a coin a 100 times. Get about 50 heads so it's ok, right? What if the fist 40 times I tossed the coin I got heads? NOT right! But the average is still perfect. Think about it! One way to take this into account is to count the distance between the events, lets say, the distance d between AA's (if I get AA and the next hand I get AA d=1). I swear the most probable distance is one! the d distro should be [(1-p)^d]*p. If you plot a histogram (monolog) of d it should be a straight line (y is log(#events on bin) ) for a large sample. Also on this note:

2b- People say you need a big sample to say a site is rigged (or suspect with very good reasons). Wrong! Let's say you see 2 or 3 royals on a row. That's only 2 or 3 events. If the probability is very low there you go! Would you play on a site you saw ONLY 2 royals on a row? The binomial distro does not need a very big sample! If you want to approximate by a Poisson or Gaussian... then yes. So to check for excess events on a subset of the data (the first 40 heads I talked about before) you could apply a binomial test on that subsample. I suspect you'll get a big excess on some subsets and a big shortage on others. Hey, the average is ok! Get the probabilities of more (or less) events on those subamples, you'll be amazed. Could that account for being very lucky for a time, and for a hell of a bead beat later? Maybe.... Note again that the hands you're dealt have nothing to do with how well you play.

3-Now for less numbers and more feelings: Party poker changed something this week, I feel. Just go there if you have an account and play as fishy as you can. Please go there and raise 72's pre flop. Keep raising. Play ANYTHING! Voila a 3 of a kind, a flush or maybe a poker.... mostly on the river. I dare you to go there NOW and set aside $2.50 on the micro limit table for this little test. How many BB do you think you'll last? One, two, 3? Don't be amazed if you make money. I know I play badly. I play there and should have lost a lot of money. I feel it doesn't matter how you play anymore. If I play "right" (ok, not like you guys by far) I just get the same average/variance as when I care about outs, poker odds and pot odds. It has been BLATENTLY rigged for a week or so. Thats how I ended here.

[edfurlong]

[ QUOTE ]
Numbers and stuff that some dude made up.

[/ QUOTE ]

lol, ok.

[Josem]

washus' analysis is flawed because it takes the most extreme figure from 13 results.

as i explained earlier in the thread, there's about a 20% chance that the most extreme figure from 13 results is at (or beyond) that point.

[washus]

Ok Josem, if you say so....

But it's still data! Show me the other 99% of AA's that are closer than 2.37 sigma and I will be happy to agree.

Someone posted this link:
http://en.wikipedia.org/wiki/Selection_bias

from wich I quote:
Data

* Rejection of "bad" data on arbitrary grounds, instead of according to previously stated or generally agreed criteria.
* Rejection of "outliers" on statistical grounds that fail to take into account important information that could be derived from wild observations as described by Kruskal.