Blackjack q - continuous shufflers - can they be beaten?

[Poshua]

Congratulations on your good luck. You cannot systematically beat a CSM.

As for what you observed, see:

http://en.wikipedia.org/wiki/Confirmation_bias
http://en.wikipedia.org/wiki/Availability_heuristic

Edit to add: what was your system for bet size management? It is easy, with bet size manipulation, to create a situation where you'll have small losses at most table sessions and large wins at a few. The house edge is unchanged and you will still lose over the long term.

[denks]

Thanks for the input. My big concern is that I happened to hit an extremely 'lucky' streak (I hate that word - I believe in probability, not luck) and just happened to sit at tables where the cards were set in my favour and in the long run things will even out.

To give an indication on the simulation Im completing, It will be 0 - 52 cards in the discard tray (some of the dealers go up to 2 decks consistently in the discard tray), with 1 - 11 cards in a buffer to be dealt. Any cards inside the machine will be randomly shuffled each round. I am not sure the exact model of CSM used, it consists of a large wheel inside which rotates with roughly 20 - 30 slots for the card to be dropped randomly into in a first in first out manner. It appears these slots are then dropped into the buffer to be dealt (hence why Im counting on up to 10 cards in the buffer). I will add this slot functionality in the next version of the simulator I write using 30 slots.

So far the simulations have supported the above that a CSM cannot be beaten, I am going to see if there is any difference in number of boxes playing or seating position.

As an aside, based on several million games of simulation per betting strategy I have performed over the past month in what appears to be roughly a 1% likelihood - my lucky month I guess [img]/images/graemlins/smile.gif[/img]

Edit: As I made more than the average amount off the BJ tables meaning I had a better than average run it is quite likely that I did indeed observe the above rather than it being cognitive bias. However that is not to say that in the long term my observations will hold - it is likely that what I observed will balance out in the end and prove to be an exception rather than the rule. Will keep you posted as to the results of the simulations. Curious on thoughts - is 1 mill simulations enough per situation?

[SheetWise]

[ QUOTE ]
My big concern is that I happened to hit an extremely 'lucky' streak (I hate that word - I believe in probability, not luck)

[/ QUOTE ]

If you hate the word luck -- then just think of it as a fortunate fluctuation. [img]/images/graemlins/wink.gif[/img]

[denks]

Thats exactly what I was thinking of it as [img]/images/graemlins/smile.gif[/img] Either I was onto something, or I had hit a very profitable statistical outlier. My main focus is actually in trying to work out the mechanics behind if these outliers can be predicted in any way. Theoretically winning and losing streaks should follow a bell shaped curve, with the y-axis being 0. My personal experience was that the outer edges of the bell shape seem to occur marginally more frequently than is expected in a perfectly random deal - both winning and losing streaks. This is definitely anecdotal still as I have not been able to reproduce this in a simulation, hence why I was to try to model as exactly as possible the CSM in use. Was wondering if anyone here has any experience or knowledge with 'clumping' - what exactly is it? And is it applicable to CSMs?

[Photoc]

[ QUOTE ]
and just happened to sit at tables where the cards were set in my favour and in the long run things will even out.

[/ QUOTE ]

No...they won't. You will lose long run, not even out. It's not possible to beat a CSM using basic strategy and counting into them has proven impossible as 99% of the games I've seen load the cards after each played hand.

[denks]

I should have been more precise about my above statement - by 'even out' I meant 'return to normal' (I am very aware of the house edge, out my way it is approx 0.4%). The dealers at my local in many cases deal out between 1 and 2 decks before placing back into the shuffler. Not all of them but a large enough number. I still not convinced a 2 deck penetration will provide a good enough count into 6 decks though am def running a few sims on it. In the process of doing other things so my model of the shuffler is taking a bit but should shortly have a very close model of how the cards are dealt, including a moderately accurate representation of the shuffler. The only part I can't figure out is whether the shuffler dumps an entire tray into the buffer or only pulls out single cards from a randomly picked tray. I do not hear the machine working while the dealer is dealing which lends me to believe that the tray is not continuously spinning each deal and thus lean towards the first - that a tray is dropped in each time it is required. Nobody happens to know which shuffler this would be (a large circle inside with approx 30 trays that rotates)? Or where I could get info on how it works so I can accurately model it?

[SheetWise]

[ QUOTE ]
I still not convinced a 2 deck penetration will provide a good enough count into 6 decks ...

[/ QUOTE ]

You would need a 12+ running (using a single level count) to make your bet. The frequency of that situation will be so low, the min/max ratio required for your top bet will (in all probability) push you over table limits. At the very least, the spread you need will draw a lot of unwanted attention. I think the zero memory game needs to offer you more than -.004 to beat a 2 deck penetration into a 6 deck game (and my simulations agree with me [or the other way around]) [img]/images/graemlins/wink.gif[/img]