[futurebasic] Re: [FB] Re: Random File List

Message: < previous - next > : Reply : Subscribe : Cleanse
Home   : October 1999 : Group Archive : Group : All Groups

From: Joe Lewis Wilkins <PepeToo@...>
Date: Mon, 04 Oct 1999 00:39:46 -0700
Jay Reeve wrote:

> >Yours is a brilliantly
> >fast and simple recipe for "random"
> >shuffling a deck of cards.
> >
> >However I cannot quite convince myself that you
> >have produced a perfectly shuffled deck.
> >Have you an argument to prove that it is?
> >
> >The order 1,...,320 does play a definite r^ole
> >in the shuffling so why not some r^ole in the result?
> Hi Larry,
> I couldn't get this to go through privately, so I'm sending it to the
> list. Perhaps it will interest someone else, too.
> First, the (admittedly brilliant) recipe is not mine. I don't recall
> where I learned it, but I used a much more complex procedure until I
> discovered this somewhere.
> There's nothing I can offer as concrete "proof," but here are some
> arguments and illustrations that may persuade you.
> Take six cards, say the Ace thru 6 of spades, and lay them out in order.
> Assign an index number (1-6) to each position. Roll a die 6 times. On the
> first roll, swap the card in the position shown on the die with the card
> in position 1. (If the die shows a 1, don't move anything.) On second
> roll, swap the position indicated by the die with  position 2, and so
> forth for the six cards.
> You'll note that the card chosen for swapping is not always the card that
> started out in that position, but that each position in the deck is
> filled by a card chosen completely at random. The fact that the same card
> could conceiveably end up in its original position, either by being
> "swapped" only with itself, or by a series of swaps into other positions,
> further illustrates the randomness.
> When you're done, think about how the position of any card could have
> been predicted without knowing the dice rolls. It can't. The card that is
> put in each position is chosen randomly from the entire deck. Whether it
> stays there or gets moved again is also decided randomly. It's actually
> far more random than the physical (standard) shuffling of a deck of
> cards, because the initial positions do have some effect on a standard
> shuffle.
> The only effect the sequence has on the procedure is to make sure each
> card gets chosen randomly. That's why the example Jonathan offered was
> ambiguous. It had no way of establishing when all of the positions in the
> deck had been addressed.
> As another exercise, run the short demo below. You can add your own
> routines if you like, to store the results and subject them to
> statistical analyses for randomness. (I assume such analyses exist--I'm
> not familiar with them.) You'll find the results appear intuitively to be
> random. Try it with only 2 cards, or with 9. If you see a suspect
> pattern, run it again and see if it recurs.
> Remark out the FN makeDeck line in the main loop and run again. It
> doesn't matter whether you start with a "new" deck, or the one you ended
> up with on the last shuffle.
> This fascinates me, so I could go on, but I think that's 'nuff for now.
> Let me know if your doubts persist.
>  0"0
>  =J= a  y
>   "
> =======================
> _cardsInDeck = 6                          '2 to 9 best for this demo
> _trials      = 50
> DIM gDeck(_cardsInDeck)
> LOCAL FN printDeck
>   POKE @deck$,_cardsInDeck
>   FOR c = 1 TO _cardsInDeck
>     POKE @deck$+c, _"0"+gDeck(c)
>   NEXT
>   PRINT ,deck$
> LOCAL FN makeDeck
>   FOR c = 1 TO _cardsInDeck
>     gDeck(c) = c
>   NEXT
> LOCAL FN shuffleDeck
>   FOR c = 1 TO _cardsInDeck
>     SWAP gDeck(c), gDeck(RND(_cardsInDeck))
>   NEXT
> WINDOW 1,"Shuffle test",(0,0)-(200,610)
> FN makeDeck
> FOR t = 1 TO _trials
>   FN makeDeck        'Remove this line & try again
>   FN shuffleDeck
>   FN printDeck
> PRINT:PRINT ,"Click to end."
> DO

Jay, I find no fault with anything you've said; other than the fact that the
RND Function always provides the same sequence of output - as advertised, it
is a psuedo-random number generator. Supposedly, a seed value, that would
alter this "same random sequence", can be introduced, which I have tried and
found it "appears" to have very little, if any, effect.

Mind you, I have done no scientific "testing"; merely noting the frequency
with which certain questions in my games appear to be selected in "about" the
same order. Since, the reappearance of the same questions does not bother me
(repetition is one of the best methods of learning), I have not pursued this.
But if "true" randomness were really important, I think I would really have to
check out this function more completely.

Joe Wilkins