Diaconis et al: Analysis of Casino Shelf Shuffling Machines

#4
Both ASMs and CSMs are likely to suffer similar dificiencies.
There is no doubt that Ace-Key Tracking works with ASMs (some if not all) and increasingly
we have empirical evidence that keying Aces also can be viable on (some if not all) CSMs. zg
 

Gamblor

Well-Known Member
#8
Pfft, seems like a stupid shuffle to me. Good news is they are bigger morons then I thought.

Thanks for the post Ken, might give me some food for thought on my local shuffle machines.

'In fact, the statisticians have reported that the the president of the company responded "We are not pleased with your conclusions, but we believe them and that's what we hired you for."' :laugh::laugh: That's what you get when you let the intern think up the shuffling algorithm.
 

Gamblor

Well-Known Member
#10
zengrifter said:
Both ASMs and CSMs are likely to suffer similar dificiencies.
Agreed. The most effective shuffle is the simplest algorithmically, I have 3 lines code to do it. Good news is that there are physical and mechanical limitations that would make it very difficult to execute this shuffling routine mechanically/physically.

I am sure that all shuffle machines suffer from this limitation, and I would conjecture that all shuffle machines are probably not that dissimilar to what is described in this article.
 

Sonny

Well-Known Member
#12
As with many AP topics, we should be very careful about what information we post publicly here. Things like this are best shared privately or (my favorite) not at all. ;)

-Sonny-
 

blackriver

Well-Known Member
#14
one strength they mention but don't elaborate on is that after the elevator dumps all the cards each shelf has about 1/10th of the deck AND "the ten pile are now assembled into one pile IN RANDOM ORDER."






Ps, I like that on the first page at the bottom it says that one of three authors was supported by the national institute of health. No wonder they haven't found cheaper medical solutions (kidding on the political critique, please don't hijack)
 

Gamblor

Well-Known Member
#15
blackriver said:
one strength they mention but don't elaborate on is that after the elevator dumps all the cards each shelf has about 1/10th of the deck AND "the ten pile are now assembled into one pile IN RANDOM ORDER."


Ps, I like that on the first page at the bottom it says that one of three authors was supported by the national institute of health. No wonder they haven't found cheaper medical solutions (kidding on the political critique, please don't hijack)
I assume the algorithm picks one of the the 10 shelves randomly, puts it ate the bottom of the pile. Then it picks one of the nine remaining shelves, and places it on top of that pile, etc., etc.,

The reason the health industry and field doesn't find cost effective medical solutions is the same reason the energy industry won't go out of its way to find cheap energy solutions :)
 

blackriver

Well-Known Member
#16
Gamblor said:
I assume the algorithm picks one of the the 10 shelves randomly, puts it ate the bottom of the pile. Then it picks one of the nine remaining shelves, and places it on top of that pile, etc., etc
great job, thank you


The reason the health industry and field doesn't find cost effective medical solutions is the same reason the energy industry won't go out of its way to find cheap energy solutions :)
because the energy industry is busy studying how to beat blackjack?
(again jk)
 

MangoJ

Well-Known Member
#17
Gamblor said:
I assume the algorithm picks one of the the 10 shelves randomly, puts it ate the bottom of the pile. Then it picks one of the nine remaining shelves, and places it on top of that pile, etc., etc.,
Thats not the way the shuffler works. The shuffle algorithm is well-explained in the paper: For each card, the algorithm picks 1 of the 10 shelves at random, and then places the cards in the shelf on top or at the bottom - also at random. None of the shelfes are excluded in any process (well, unless there are mechanical limits).

Stating the algorithm in much simpler words: You have 20 shelves, where all cards are dispersed always on the top at a random shelve. Then the cards in every odd shelve are reversed, and all shelves combined to a single pile.

If you would number the original deck with 1-52, you would come up with 10 rising sequences of cards interspaced by 10 falling sequences. That's why the authors are able to guess 8 cards instead of 4.
 

blackriver

Well-Known Member
#18
MangoJ said:
Thats not the way the shuffler works.
We're talking about what exactly happens after the cards are put on the shelves
The shuffle algorithm is well-explained in the paper.
All the more reason to expect the guy whos illustration is more clear than the article itself would have at lest grasped what the article says clearly and repeats at least twice
 

MangoJ

Well-Known Member
#19
You're right, I missed the part you were talking about how the shelves where combined into the final stage.

Actually the details of that combination process doesn't matter, because each shelve is chosen at random in the first stage. They could even combine the shelves in a fixed order, it won't change the "quality" of the algorithm.


I don't get what the designer's problems are with such machines. Obviously they have a microcontroller unit on board. They should do a Fisher-Yates shuffle to produce out a truely random sequence in memory, and then simply use their mechanical devices to perform the physical card swapping and reordering of cards necessary to bring the cards into that random sequence.
 

blackriver

Well-Known Member
#20
It matters significantly. If the cards are combined as out seems, the way gamblor describes then it just prevents a lot of forms for sequencing.

If they were piled 1 card at a time randomly then the cards would be effectively shuffled twice which would be sufficient. They don't elaborate though and if you didn't know better and only had that sentence out of context it could (and maybe should if read technically and literally) be interpreted as if this second and dangerously sufficient shuffle was what was happening
 
Top