UB/Bet21 Seating Assignments: feeding the conspiracy theorists

Really? When the expected ratio of dealer busting is 28.2% and you are subjected to it occurring 13% doesn’t bother you?

When the percentage of dealers drawing to 20 or 21 should be 25% and you observe a 35% it doesn’t bother you?

True the sample size that I presented is small – I indicated that it was preliminary, but compared to the sample size that you presented when you first noticed the non-random seating phenomena I thought it would at least peek your interest.

That’s not what I was looking at. In addition I don’t understand you argument. For example when the dealer has a 6 as an upcard, we can expect the dealer to bust 43.9% of the time. In this case most people would stand and not hit. However during this recent stretch when the dealer has a 6 showing they only busted 4/13 times (30.7%).

Lets take this for example when the dealer’s upcard is a 3. We know the dealer will bust 37.7% so we stand with our stiff’s right? Well what happens when the dealer only busts 20% of the time. Shouldn’t/wouldn’t that affect the way you played when you have a 12, 15 or 17?

I believe that is directly related to seat position and what cycle the dealer is on. Now unless you believe that you, as an individual, can be targeted and “picked upon” I don’t see the value of tracking other people’s hands and comparison to mine.

But maybe I’m taking a much too simplistic approach to this whole thing?

As an aside do you still believe EBJ rewards the aggressive bettor?

 

long term probs

fgk -

the thing with tracking probabilities - is that many of us have played a lot of tables of online ebj - I am probably close to 1,000 now - with about, say, an average of just 15 hands per table - that's 15,000 hands - so - say you observe a 1 out of 10,000 probability occur over a span of 50 hands - well - the odds that such a long shot probability will happen at least once in your sequence of 15,000 hands is about 78% - so - not long odds at all - when there are so many odd occurances you can be tracking - fishing expeditions always turn up the odd result - and when you look at the real probs - that something will occur somewhere in such a long sequence of tables/hands - that doesn't mean anything -

dealer cards over a short sequence - and anything under hundreds/thousands of hands is a very short sequence - are more likely to be skewed off from the expected probs than align with them - and since all players are playing against the same dealer - it doesn't matter - what matters is if you are playing better than the other players - and how that creates long term results - if there is a long term, very high odds, discrepancy between your hands and the other players' hands that disadvantages you - that is the problem that will hurt you - so that is what you should be watching for - I have tracked such - and have found some very long sequences, with very long odds against, where I got much worse hands than the other players - where I am getting for example 55% of my hands dealt are stiffs - versus about 35% stiffs for the other players on average - but keep tracking - and you'll see the reversal - and over the long run - the hands come out pretty even - I get average hands - except for that dang deficit of bjs - which is still not statistically significant -

and yes - I not only believe that aggression is a plus in playing ebj - but am coming aroubnd to a more aggressive style of play for live tourneys as well - but not off the wall aggression - no all-ins on the first hand - and aggressive betting should be coupled with conservative play - lots of surrenders and few doubles or splits - agressive betting, within reasonable % of bankroll limits, and conservative play wins in ebj -

 

Surely only the recorded hands are of any consequence. The fact that a lot more hands have actually been played (both by the same player and as a whole) just means that a larger sample could have been available; it can't affect the results you expect to see in the small sample.

The key issue is how large a sample is actually needed to assess whether the frequency of certain events is consistent with a fair game. That's where I get out of my mathematical depth, but I would guess it varies according to what event(s) you are actually tracking, and hence the seating assignments could be flagged as non-random based on a fairly small sample but the same may not be true of some of the things Fred is now tracking.

 

perspective on probabilities

London -

the thing with fishing expeditions is that they always turn up something - the question is - "how likely is 'something' to turn up?" -

good procedure, if you are sampling from a very long sequence - is to ask not just 'what are the odds against this happening?', but also explore the possibility that you are just catching an odd occurrance that is highly probable to turn up somewhere in the very long sequence you are sampling from - so a good cut at this is to take the probability that something will NOT happen - then calculate the probabilty that it will not happen ever in series of repeated trials the length of your very long sequence -

thus

probability against a player going 200 hands without a bj is 0.000059167738, pretty long odds, but, if you have played 15,000 hands - the probability that somewhere in those 15,000 hands you have seen AT LEAST one sequence of 200 hands without a bj is 0.58833376, more likely to see such a sequence than not - so the fact that your tracking turned up such a sequence is not necessarily significant - turn up a half dozen such sequences - then you may have something -

or show a linkage between your odd occurrance and something else - such as - the sequence of 200 hands without a bj starts every time my bank balance goes over $X,XXX.XX - that might be meaningful -

on the seating assignments - it was tables that were the universe we are sampling from, and tables are much fewer than hands played - even with 1,000 tables, it was a very remote probability that the sequence of seat assignments I saw would ever occur - so we could conclude that was non-random -

with the hand probabilities we are talking about - we are a long ways away from being able to conclude they are non-random events - we need repeated observations of these occurring, and/or linkinages with other factors - such as account balances -

 

That doesn't make any sense to me. Unless you have recorded those other 14,800 hands and verified that the overall bj frequency is within it's expected range then they may as well not exist. You might just as well say that if you play 200 hands without a BJ it's not significant because 74 other players, each also playing 200 hands, are not being factored into your calculations.

Surely a significant 'something else' here is the decision to record 200 hands. The probability of at least one sequence of 200 containing no bj may be 0.58, but what of the probability of the single sequence which I randomly decided to record containing no bj? Surely that is identical to the probability of any such sequence containing no bj - i.e. 0.000059167738 ?

 

fishing expeditions

London -

it is the difference between 'fishing' and hypotheses testing -

if you are testing a hypotheses - then you are sampling from a large universe, to answer a very specific question, and need to do appropriate significance testing on your results -

but if you are 'fishing' to see what turns up - then you are not approaching the sampling with purpose - and looking for something specific - just looking for 'an oddity' - since there are many possible long odds probabilities that could show up - the odds become pretty good that one of them will - so you need to look at those probabilities quite differently -

 

from non-random seating to cards being dealt unfairly

At first, a few weeks ago, when I noticed I was getting seats 1 or 4 in all my SNGs, I thought, okay...this is part of long-term probability. But when another player spoke up about non-random seating, and I followed with my observation, and then came a dozen others...we KNEW there was something wrong with the software. And when it was fixed the other day, we all KNEW it was fixed. There was no doubt about both things happening, since so many of us reported our findings and KNOWings.

In the last few weeks, I've noticed a degradation in the cards that are initially dealt me, and the hits and double downs I take. I have increasingly received 12-16s, decreasingly received 19s and 20s and BJs. When I hit a stiff, I usually bust...I rarely now hit to a 19-21, while many around me do. I get tons of 17s vs a dealer 10...to the point where I'm surrendering most of these 17s. When I double down, most of the time I get 2-6, with an ace showing up when I double an 11. When I go all in and get a 19...the dealer has 20. When I have a 20...the dealer gets 21. It's almost to the point where I can predict the outcome. Like FGK recently posted, I do better when I bet 500 for each of the first 7 hands.

What follows is something similar to what I've seen a few others have said during some recent games. So, while at first I thought it was an anomaly with me, the fact that it's happening to at least five other players, leads me to think that something is really not right with the cards being dealt. So instead of thinking there may be something wrong, as with the non-random seating, I'm beginning to think that we now KNOW (for sure) that something is up with the dealt cards.

What's most disturbing of all: I have had countless streaks where I may win 1 or 2 hands out of the first 8; 3 out of 16 to 19 hands; and losing streaks of 5-7 hands. I know this is possible in the long run, but in the last two to three weeks, I have had losing streaks like these at least a dozen times. I do not remember this happening in my first five months on the site. Nor have I gotten the reverse to happen -- winning 6 of 8 or even 10 of 16.

I have seen swings against me so many times. Three other players winning their hands and my losing mine. This happens more than the 12% of the time it should happen with just one player.

The only way I've advanced through some tables in the last few weeks is by getting the occasional good hand on an all-in desparation bet, or by other players losing more than me on those dealer 20-21s.

Ken has posted that he has withdrawn all his funds and will no longer play on the site until we get answers. I am close to this point as well. I will try a few more SNGs and tournaments, but if what I've observed recently does not change, there is no point in my playing against a stacked deck. I'll simply play the $25k tny since I've already won a space there, but then that will be it.

 

81-1 odds against has happened four times

I don't know all the odds of losing hands, but I just read this online:
From any point in time, your odds against losing the next six hands (with no pushes) are 81-to-1.
I have lost six hands in a row (heck, at one point I lost 12 in a row) at least four times in the last two weeks.

 

I dont understand the fishing reference.

Tests have been run thru millions of hands dealt to determine how often an Ace should show as the dealers up card in a fairly and random dealt game.

I dont know if its fishing but when Im playing lets say 15 sit and goes today and I note that usually in the first 8 hands its quite common for the dealer to have Ace up at least 3 times...well it gets me saying hummmm...possible sure....and so on for the rest of the rounds...

I think the question is it normal to expect to see Dealer Ace Up 3 to 5 times in the first 8 hands...? I think it happens pretty often (beyond the trial run percents) The fact that it happens so often makes it notable but unstudied...

I dont think we need to look at every game dealt to get averages Dealer ace up ...a one in 13 chance of getting the ace....so dont you have to double that to assure that that ace is UP?...Im not looking at or talking about "back door" dealer BJ...only ace up.

I still dont thing any of these devs from the norm affect any individual player in a meaningful way....except that maybe its an advantage to understand that the dealer is gonna have strong cards ...maybe stronger than normal percentages that one would expect to see....

If the dealer shows an Ace up it is terrible for everyone...not any individual...

Now if the dealer shows this strong card more often than normal then it is still bad for everyone ...but the only real effect is to dwindle BR's and threrfore possibly end the game sooner....

Why do that you ask??? The sooner the game ends the quicker the next one starts...I pay a rake for every game I play....Its better if I pay 5 rakes per hour than if I only pay 2.

Who knows if this is really even happening .....?

But Id like to see some independant test results that show ....

AN ace being dealt to the dealer once every 13 th card....or what ever percent is normal in a random dealt million hand trial.

Seating.....I agree ...something changed.....I am now being moved around ...I get 1 thru 7 .....I played aprox 35 games today and watched that quite closely...

As far as the deal...I dont notice any change since day one....I have always suspected adjustable dealer levels...IE easy, medium , harder , killer.... but that doesnt affect my win rate in any way...only the length of the game maybe...

I think its still a great game....I suspect there are undiscovered bugs but I certainly dont suspect anything neferious ....I dont suspect any particular player gets dealt an advantageous hand on any sort or regular basis...

I do sometimes wonder if the game rewards "dumb' play....I watch the dumb moves being rewarded over and over and start to say hummmmmmmm..like tonights donkey special....guy on my left splits 3's against dealer ten up....round one!!!! 7,000 on each hand .....21 and 19 to the dealers 18....lol....

Yes he eventially lost with a scores ranging from 120,000 down to zero fianlly...but he lasted almost to the final hand...

I also played a game today where the first 6 dealer hands were BJ,BJ,BJ,21,BJ,20....thats odd ....No?????

 

I had 3 BJs in a row and witnessed another player get 3 in a row (different tables) at UBT online, did I see something rare?

being a mediocre programmer I know it would be easy enough to write the program in a manner the cards a player receives could be based on any number of variables, player rating, how the player plays the cards, how the player bets, how often he plays and so on, but it would have to be very elaborate to target specific players while others are left to random chance, especially when it comes to the deal....

funny but before I got into playing the game while in college one of my pearl101 (thats a server side programming language) class assignment was to create and run an online blackjack game and slot machine on a school server.

on the subject of random number generators, you can't really manipulate them for an expected outcome easily, they either are or aren't. once you manipulate a RNG it will always produce the same result, unless you embed another generator within a generator so on and so on, so if someone feels a seating assignment is being manipulated a specific pattern will eventually unfold..

since there is no "real" shoe to examine the only way to see if any of these online establishments are on the level as far as the deal goes they would have to allow programmers access to the source code, but even then unless someone was babysitting the server the ability to verify that was indeed the software running on the server would be difficult...

as far as seating assignments I wouldn't bet the bank a RNG is in charge there, seating could be assigned according to other variables including player rankings? once again, have to look at the source code....

 

whoever wrote those odds never played at the ironhorse casino... :laugh:

 

non-mathematician

FWIW RKuczeks posts about hypothesis testing and "fishing" makes a lot of sense to me. The way I understand it, what he is saying is that just because we happen to witness/record an improbable event occurring, for example a spell of 200 hands without a BJ, it doesn't necessarily mean that we have seen anything unusual or unexpected.

Lets take TedinNaples 6 hand losing streak as an example, which as he correctly says has a 1 in 82 chance of happening. If I watch him play and at the precise point I start watching he hits his losing streak I can't yet say I have seen anything unexpected even though there is only a 1.2% chance of it happening. If I continue watching him for another 486 hands (I'm assuming that a 1:82 chance of a six hand losing streak equates to 1 in 492 hands. Am I right or should I expect to see that streak in 82 hands??) then I may conclude that the streak I witnessed at the start was in fact not unusual or unexpected at all. If Ted has another streak then I may get suspicious but even then if he only has 2 streaks in 986 hands then that is to be expected. I suppose the question is how many hands must we track before we can say with any confidence that something is amiss?

Another way of looking at it is that if we assume a hand rate of 40/hour then Ted would expect to see that 6 hand streak at least once in a 12 hour session.

I'm probably missing some point here and talking out of my fundament but... ...meh!

Unless something has a 0.00000% chance of occurring we shouldn't expect it not to occur.

Cheers

Reachy Babbage

 

Gone Fishing

That makes more sense (if I'm reading it right). The key issue is not so much the sample size, but how many separate variables you could arbitrarily choose to examine. The more there are, the more likely it is that one of them might appear out of whack; to continue the metaphor, your net will always contain some odd sea creature among all the fish you were expecting to catch.

So, once you've identified a particular variable (e.g., the frequency of BJs) that you are suspicious of, further sampling in which that is the only thing you are examining is in order.

Is that a reasonable summary?

 

wow

For someone who has been playing scheduled tourneys and just started sng (hey I finally cashed on a scheduled tourn after 2 months) - I am amazed - I noticed last week that I was in seat 5 a LOT - thought it was coincidence. But if there is a conspiracy - who does it benefit?

 

I meant but why would the casino bother? OR I guess you all are saying they WANT certain players to win - OMG - I am so naive...ready to go to the real casino and play

 

It's unlikely that the non-random seating problem is a conspiracy, much more likely that it's a programming glitch that has turned into something more simply due to inattention by the folks responsible for website maintenance.

 

right question

in testing probabilities, the starting point is making sure you are asking the right question, and also testing for the significance of your result. In example - in my sample of fifty seating assignments, where only two seats were in the 5,6, & 7 positions, that was a 0.0000000005116498 probability, but, there are 200 possible combinations of four out of seven seats, so the probability that one of those combinations would come up with such an improbable seating assignment is 1-(1-0.0000000005116498)^200, where ^ is the symbol for an exponent - or the probability that I would find such a low probability result for SOME four seat combination is 0.00000010233, still pretty remote - but much more likely than the first. Given that I have played about 1,000 tables, a good guess as to the liklihood of such a remote possibility coming up AT LEAST ONCE in those 1,000 tables is 1-(1-0.00000010233)^950, or 0.0000972088, about 1 out of 10,000; remote enough that this is very likely non-random -

the question here is not 'what are the odds against 48 out of 50 seating assignments being in seats 1 through 4?', but 'what are the odds against 48 out of 50 seating assignments being in SOME combination of four out of seven seats, and, what are the odds that such an assignment could happen in the sequence of tables I have played at least once?' -

in short sequences, probabilities are more likely to skew off than reflect the expected results, and 10,000 hands is a very short sequence. The probabilities we work with are the odds that infinitely long sequences converge to, unless you are looking at maybe 100,000,000 hands - don't expect to see them approach exactness -

what is happening here, is that with all the tables we have played, it would be highly unusual if some very remote probabilities did not occur, but as experienced players we notice when they do, and react to that, thinking, 'this is really odd' - it actually would be very odd if we didn't see these sequences occur some of the time -

 

More Raw Data

These were the numbers from this past, and final, weekend of play at Bet21.com

Raw numbers only:

Hands 452

Dealers cards only for this post:

Aces - 32 (7.1%0
Dealer BJ - 23 (5.1%)
Dealer Bust – 107 (23.7%)
Dealer 19-21 – 182 (40.3%)

Expect values:
Aces - 7.2%
BJ - 4.75%
Busting - 28.2%
Drawing 19-21 – 38.6%

I also tracked MY first 2 cards and those results are dismal. The distribution of these numbers is also very skewed. For example one game had the dealer drawing 19-21 71.4% of the hands while in another it was only 12.5%.

There was also a difference between SNG and MTT. The sample size was too small to determine differences between SNG table limits but I suspect there was a skewering effect also. More at a later time

PS Rando - I didn't get you the sheet on purpose - I'll elaborate on it later. Sorry