Point taken (and fixed your post!!)
Getting an overlay on all of those of course will just make you way LESS than what you SHOULD have gotten.

So you have bottom 0.25% over the 750k hands.
And then the next sample of 100k hands or whatever just happens to be in the bottom 2% as well.

Having a 0.25% followed by a 2% is really really slim. And I really doubt this just HAPPENS to be happening to you and am far more inclined to look at whether one's style of play actually CAN make a difference on these calcs.
On it's face it sounds ridiculous since you are just looking at the equity AFTER you've gotten it all-in and are done betting and it does make sense that there should be no skew at all to that. That's why everyone believes those things are so legit. Because it makes logical sense that we can just dismiss the 'evening out' aspect of folded cards after the flop, etc etc.

But I think it's even more ridiculous that you just happen to be in the bottom 2% or whatever pretty much over and over like that.
And also think it's ridiculous that Stars would be tinkering with the cards in just that way.

So of the 3 somewhat ridiculous options I think that my preferred ridiculous option of the pokerEV thing being slightly off (and that one shouldn't accept the results from that as 100% accurate) isn't that far out there.

the steam - You don't understand the graph.
It is not including the times you just steal the blinds with AA.
It only looks at the hands AFTER you get it all-in and whether your equity is running up to par or not. Your post is not relevant at all really.

Sigh.

So many people don't know how the program works and it's so hard to explain it to them.

The program calculates your equity in the pot. Sometimes it is big (when you are way ahead). Sometimes it is small. The program tells you how close you came to reaching your equity.

It doesn't matter how often you get your money in good or bad. It tells you how much you should have won ALREADY TAKING THOSE OTHER FACTORS INTO account.

I hadn't even looked at that graph I was just trying to point out factors why people would think online poker is rigged.

Even though I personally think it isn't

1p0kerboy - I agree that others don't understand and that that is the intention of the program and that it is mostly accurate.
But is it 100% accurate? That's the question and I think it's a fair one.

When myturn is getting stuff like being 5% off where he should be on all-in's (almost positive he said this) then I start to wonder what the hell is going on too.

MicroBOB

There's probably a decent chance that pokerEV isn't 100% accurate. But for sure it's close. As mentioned by Fozzy I think there are a couple of different ways to arrive at these calculations.

I've run really, really bad this year. Whether my results are in the bottom 1% or 2% or 5% it doesn't much matter and it probably won't change anything if we were ever able to accurately pin it down.

I'm not blaming Stars for my bad luck. In fact, I think I'm just the 'lucky one' who ended up in the tiny left corner on the bell curve. Actually I'm outside of the curve because my results are outside of three standard deviations.

If a site were able to manipulate results of the cards once the betting is done I hardly think they would be stupid enough to do it to the extreme.

But the point MTTR was making was that the regular players (and by this I mean the grinders who are pumping out a ton of hands) seem to always perform badly (not ridiculously badly like me) on these calculations as a whole and that seemed a bit weird to him (and me also).

I think me posting my graph took the conversation away from that main point.

I'm not saying that Stars put me in the bottom 1% of possible results. But a little manipulation with a little bad luck could have caused that.

So the new graphs are over 889,353 hands, 21,616 of which were all-in at some point.

if you guys want to prove that a site is rigged, its very easy. simply observe the mtts/sngs at Cake Poker. Track any hand that has exactly two players all in prelop. If it is two overs vs a pocket pair, track which hand wins. For all other all in pre hands with exactly two players (non-coinflips), track if the favorite or underdog wins.

I have done this numerous times in the past 3 months and the underdog has won 76% (150 hands), 78% (265 hands), 71% (176 hands), and 82% (224 hands). I was watching a friend in th $5r last night on cake, tracked 42 hands with this criteria, twice the best hand won.

I havent totaled up all my coinflip situations but the overcards are about a 65-75% favorite so far.

To check if PokerEV works proprerly:

Take a large database (say 10 million hands+)
Order the players by number of hands played and pick the top 1000 or so
Run each of their histories through PokerEV and get their normalized outcome range
Plot all of them on the same graph
Check that the graph is centered around 0 and normally distributed

Fixed again.
Sorry to harp on this semantic thing. But I think this is a more significant possibility than you realize.
You are 3 SD's off ACCORDING TO POKEREV.

I would hazard to say that there are a lot if tight players who are 3 SD's off then is realistically possible if they did the same thing you did and took such a large sample.
If that were true then I really dismiss the possibility that you just happened to be running super-bad just like everyone else because that many people being 3 SD's off or whatever would be WAY unrealistic. Probably like billion to 1 type stuff after you get enough players who are ALL 3 SD's off.

Again, this is all really warped estimating and I don't know if all the tighter regulars are running that far off. But if we did a study and found this to be true then that would sure be interesting in determining the accuracy of this pokerev thing OR determining that Stars really does have it in for their regular players.


And me too. I think it's really noticeable and really weird.

I'm going on the assumption that the Stars cards are fine and that the bit about 'these calculations as a whole' might be incorrect.
He's going on the assumption that the calculations as a whole are 100% correct and thus the cards on Stars might be incorrect.



My point is that 'close' isn't good enough for a large sample where that teeny change makes a bigger and bigger difference the more you hit it over and over again.
For all I know, 99.9% accurate could end up yielding super skewed results after 800k hands or something. I don't know.

This is similar to why it's profitable in the first place. You have a smallish advantage over your less-skilled opponents and you hit that advantage over and over and over again to eventually get to a profit.

After 800k hands this bit about pokerev maybe only being 'close' could become a bigger and bigger error until, after a while, we get end results that are further and further away from correct.

Yes, the difference is ever so slight for individual hand calculations so that it doesn't matter at the time.
Just like our advantage is slight from one hand to the next so that it doesn't really matter at the time (I don't play just one hand with the expectation to actually profit).

I think this could be especially true for a guy like MTTR in a sample of 6-max shortstacking for example where he gets it all-in quite a bit (not sure how much of his sample included shortstacking).
He pushes with 33 on the SB and gets called with A7 by the BB. And the A7 guy seems to suck out so much more than it should on him.
Well, the CO and button folding (as well as MP and UTG) show a pretty high likelihood that they may not have held an ace.
Similarly, the fact that they did fold gives them a higher likelihood of having folded a 3.

This would explain why AQ vs. KK all-ins or something appear to be 'ace magnets'. Because if somebody else had AA then those two wouldn't have gotten it all-in in the first place or would have with completely different percentages against somebody with AA.

For a shortstacker like MTTR I can very easily see how this effect could be even more drastic and up even more.

No.
There's still some bias there.
Every time two players get all-in with AK vs. QQ you absolutely know that every single other player at the table did NOT have AA or KK.

How much such a bias would effect the sample I don't know.
But this sample would not be 100% bias-free.
AK would almost certainly have a slightly better than expected chance of 'sucking out' on the QQ due to the slightly increased chance that there are more A's and K's remaining in the unused cards simply due to the fact that everyone else folded.

After time this sample might skew the results.

No.
This would not work.
You do not know if there is total randomization of the tightness or profitability of these players.

I might think that the very top of the list of 'most hands played' may tend to contain tighter-than-average players of the group.

Maybe not. But you are assuming that there is no correlation between total hands played and whether the top of this group MIGHT be a tad tighter (24-tabling types who play 11/7 for example).

And therein lies my point.
I believe I am finding POTENTIAL (not definite) for bias in a couple other decent ideas to try to make such a determination.

It's really a minefield to try to figure out a TRUE way to make such a determination.

But the Cake Poker SNG one may really be testing whether AK vs. QQ is really a 45/55 proposition (or whatever) or whether the effect of card-removal (or bunching which is a term I hate) can have a significant impact over time.

The 2nd one really could be determining whether some of the tighter players end up skewing the results more than slightly looser players. And I definitely think that assuming that hands played of 1000 players will have completely randomized playing styles is potentially dangerous.

I'm starting to get intrigued. If it turns out there is a bias on Stars we may as well pack up and go home. This is the most trusted site out there.

yes, but a small effect. just ignore the coinflips then. do it for the rest of the hands. my point is the results will be so outrageous that there will be no other explanation.

I still think it's worth doing. Let's say you take all the top players (players playing lots of hands) and you discover that most of them are running below expectation.

Armed with this knowledge you can then take another large sample from another poker site and see if you get the same results. We can then also ask the PokerEV guys to check it out and explain what is happening.

I still think it WILL come out normally distributed though.

BTW I think Holdem Manager can also generate all-in EV numbers. Maybe someone could see if they correlate? If 2 seperate sources give you the same results then it's more likely to be correct.

what 'rest of the hands'?
The bias would have the same problem on the 80/20 hands (turning them into 82/18's or whatever perhaps) as they would have on the 50/50's.
I'm just using AK vs. QQ as ONE example to show that such a bias could be an issue with ALL the hands on a pokerEV graph type study.

This selection bias issue is definitely not just isolated to 50/50 type 'races'.

I'm sorry that my ability to explain this isn't the greatest.
But others are thinking they've found fool-proof ways to show one thing or another and I am finding situations where there is at least the POTENTIAL for bias of the selection and for skewed results.
I'm really pretty dumb at this stuff and am just feeling my way along. But it's clear to me that "close enough" and "should probably balance out in the end" is not even close to being valid.

1p0kerboy is more likely to get it in good with QQ or KK or something because he is a tighter player.
A non-regular who doesn't play as well is more likely to get it in with A8 or something kind of crappy.
In the situations where A8 vs. QQ goes all-in preflop AND everyone else gets out of the way the A8 guy has a greater chance of sucking out with an A because all the other players who got out of the way are less likely to have had an A. Absolutely nobody in the hand had an AA obviously.

This is why a tighter player might show worse all-in results vs. the supposed 'norm' over time. Because he's less likely to be the type to get it in bad with A8o.

If AA was among the cards that any of the others players had then this QQ vs. A8o confrontation would NEVER have been a part of the all-in sample.
It only gets to be a part of the sample when nobody else has AA.
And this is different than dealing A8 vs. QQ in a heads-up game where nobody else is at the table to fold preflop.

you have zero situations in the sample where it's A8 vs. QQ and somebody else with AA decided to get out of the way. None.
It's an incomplete sample. Or rather, it's not a 'random sample' of A8 vs. QQ situations. The A8 is going to win ever so slightly more than it's due even if there's just ONE other player who folded preflop.

The more I think about it the more I am certain that such a bias in the sampling could make a big difference over a monster sample.
And I believe THAT is what 1p0kerboys graph is actually proving.

I don't think it will.
I think winning players are more likely to get it when good. And the percentage of times the underdog catches them will be slightly higher than otherwise expected because the bad players are more likely to play with any ace and to get 2 players all-in means that everyone else has to fold to get to that point.

It might be interesting to look at but there are other conclusions one could draw from the numbers.

It would not surprise me in the least if super-tight players end up with the worse end on all-in equity somehow.

I don't.
They both likely do the calculations in the same way and are making the same small error. Over a short sample it's reasonable and interesting to look at the results to determine if one is getting clobbered on the all-ins or not or whatever.
Over a larger sample the little mistake of it not being able to include the hole-cards sampling adds up.

There are 169 hole-card combos I believe.
2 of those combos should NOT be considered to be the remotest possible for any of the players at the table if they folded pre-flop. Those are AA and KK.
I think that AKs and AKo and QQ and usually JJ are pretty darn unrealistic holdings for any player who folds preflop especially when nobody has entered the pot yet.
So each opponent is given credit for POSSIBLY having any one of the 169 hand combos. But really they should only be given credit for 164.
The 3% chance or whatever the EV thing gives EVERY player of folding AA, KK, QQ or AK is simply not accurate at all.

So giving the other players credit for 100% randomness when they are AT BEST only possibly folding 97% of those cards really adds up.

I believe it was stated that if you have 8 players folding preflop and assumng they all had 1/220 chance of getting AA then the chance that either the SB or BB has AA shoots up to 1/140 or something like that (can't remember the exact number).
This would remain consistent with the chances of an A8 sucking out on a QQ being slightly higher than the 'norm' if everyone else folded.
Whether it gets folded around to the blinds or not doesn't matter. If ANYONE has AA they are going to play.

so if the everyones' results come back that the underdog wins 85% of the time you will just say this is because of the unexposed cards?

underdogs don't win 85% of the time. that's why they're called underdogs.

No.
I'm saying that unexposed cards COULD be a factor.
I'm not smart enough to know for certain whether it would be because of unexposed cards.

I think if we did this then winning players would tend to show alarmingly poor results on the all-in equity graphs.
I think it either has to be one of 4 things:
1. Stars is rigging the games (perhaps against winning players to not win so much)
2. We really did have a 1 in a million situation of ALL these winning players just happening to have horrible all-in equity graph results
3. The way the graph is computed can be skewed over long periods of time perhaps due to the effect of unexposed cards
4. Some other factor that I'm not thinking of that nobody else has mentioned


I am currently in a state where I would tend to bet on #3 if I had to choose but I also admit my own flaws in figuring this stuff out because I'm just kinda dumb on a lot of things.
#4 would be pretty high on my list too.

If we had the magic UB superuser account that could see all the hole-cards I think people would be surprised at the number of times that AK all-in really had a better chance than QQ when we actually get to see what cards had been gotten rid of by everyone else and how the pokerEV equity graph of NOT knowing everyone's hole-cards would be way different than the ACTUAL of graph of knowing everyone's folded cards after one gets to see 800k hands of this stuff.

I would be very interested in seeing some of those simulation programs go to work on this. How about 5 players ALWAYS fold pre-flop at the table no matter what they have and only 2 players play 'normally'...vs. a situation where all 7 players at the table are playing somewhat normally?
I think the graph for the table where all 5 players fold pre-flop would be way different.

Something like that anyway.

Anyway, I still don't think I'm explaining my case very well and would love it if somebody else who understood what the hell I was trying to get at were to take over and try to interpret it a little bit better. Sometimes my thoughts get a little bit confused and turned around all over the place in my brain.

they do at cake mtts/sngs, check it out for yourself