shuffling / programming/ algorithm

Hola: algorithm

I want to do a question for those of you that are doing SIM software.
can you explain me the algorithm used to sim the shuffing procedure of the dealer???

I do this question because aparently this issue is the "point", when compare the sim on Counting, shuffle tracking, and finally progression (sorry about the word).

Can the procedures used in soft programming replicate the procedure used by the dealer in shuffling???? and, Obtain the same patterns (or whatever) that the dealer???

I am programming a simulator, and I don't want to make a mistake in this item, because i believe that this is "THE POINT".

TNKS in advance

saludos

Most simulators will use a random shuffle for several reasons:

• It's easy to code.


• It's fast and efficient.


• Simulations of real-world shuffles have already been done, and do not produce significantly different results.

I'm not aware of any simulators that support shuffle-tracking. If you succeed, it could be a valuable tool.

>Can the procedures used in soft programming replicate the procedure used by the dealer in shuffling???? and, Obtain the same patterns (or whatever) that the dealer???<

Are you asking the algorythm of the "computer random shuffle?"

Are you asking how to program a riffle? How "technical" do you want to get on the riffle? IOW, do you want to account for the fact that the cards tend to drop at different rates depending on how deep you are into the two "picks?" Or, are you willing to assume, as I did 25 years ago when I was doing what you are trying to do now, that the drop rates vary so much from dealer to dealer, and from the start of a dealer's shift to the end of the same dealer's shift, than an overall average drop rate is the best way to go?

Here is the drop rate I ended up using:

int rnd3( )
{...int tmp;

....if ( !( tmp = ( rnd( ) & 15 ) ) )
........return 3;
....else if ( !( tmp & 3 ) )
........return 2;
....else return 1;
}

Of the 16 possible binary values for "tmp":
0000 returns 3;
0100, 1000 and 1100 return 2;
and all the rest return 1.

Thus the average "drop" size becomes 1.3125 cards, which is almost exactly what my research showed was the average drop size of a flesh-and-blood dealer riffling two half-deck picks using a "fine" shuffle.

Also, I do not use the "rnd( )" function -- the period is too short. Instead, I use the "r250( )" function which was written up in Dr. Dobbs Journal, and which is available on the internet. Try a Google search on "r250 +programming" or "r250 +random." The period for this RNG is 2^250 - 1, which is more than adequate for any program you would ever live long enough to be able to run.

Thank you for the answer.
I under estimate the item, so your answer has been very usefull to me. I will re-think wath i am doing.

Aren't the least significant bits the worst ones to use with rnd()?

I'm not familiar with this product. I thought it was just like CVShuffle, but I see it boasts an "automatic simulation feature".

Can you define the shuffle and how you intend to track it, then run a billion hands and see your expectation? That's what I thought marcelo1248 meant by a simulator that supports shuffle-tracking, and what I'd consider a valuable tool.

So don't use rnd( ). If you refer to the earlier post, you'll see where I use and recommend the r250( ) algorithm.

In either case, we're talking about the number of cards dropped from each pick in a riffle. With the great variations in drop rates from dealer to dealer, and even with the same dealer at different times during the shift, the *average* (especially when looking at the aggregated results over hundreds of millions of shoes) is FAR more important than the *sequence*. IOW, the sequences will all average out in time.

Also, it's interesting to watch in debug mode, as the values of 2 often cluster together. By happy coincidence, this is also what happens with flesh-and-blood dealers, as true riffles tend to be coarser at times and finer at others, usually depending on how deep they are into the pick.