RF>   What you said above proves my point. A routine that does not
RF> statistically average out over time so that every outcome appears
RF> an equal number of times is better than a predictable routine where
RF> each outcome comes out an equal number of times over time. Think
RF> about that.
DW>Well, this is a strange point! The conventional definition of "random"
DW>includes the concept that non-degenerate outcomes should be equally
DW>probable, which means that they should come up equally frequently over a
DW>long term. (The business about "non-degenerate" has to do with
    Keep in mind that the study of statistics was created to try to
   understand and find order where there supposedly isn't any order.
   It isn't an exact science and should not be used or thought of as such.
   By imposing this definition that all possible outcomes should come
   out an equal number of times, and since you said that is what you found
   with your routine, you should be concerned not overjoyed. Remember,
   we want the most random routine, not the routine that produces a
   pattern of outcomes more quickly over time.
    What your argument seems to be is that if your routine, for the
   most part, produces orderly "textbook" outcomes over time, it is better.
    I'm saying that if my routine does not or takes longer before any
   pattern is seen it may just mean that my routine is better in that
   no pattern is seen over time or my routine produces random outcomes
   for a longer period of time before any pattern is seen. Not that I
   believe that. I suspect that both routines over time are very likely
   equal in their ability to produce random outcomes.
DW>indistinguishable outcomes that can be reached in several ways. It's not
DW>an important factor in our discussion.) You are suggesting that it is
DW>somehow better for this equi-probability *not* to occur. Hmmm.... I
    You should make certain that you run both routines more than
   600 times and see if that changes the results. Maybe one routine
   produces more random outcomes before a pattern begins to develop.
   Also keep in mind that BASIC's randomness is not the best possible.
    As far as this thread goes you would have to post the complete code
   that you are using to test the routines so that I can run them here
   and you should also post your findings. I've learned that anyone can
   claim anything and so I need to be able to duplicate your findings
   before I ever agree with you that your routine is more random.
DW>guess you wouldn't take this so far as to say that the best "randomizer"
DW>is one that always produces the exact same order every time, for example
DW>the same order as the one in which the characters (or other things) were
DW>originally in! Such a routine would have the advantage of being
DW>extremely easy to write!
   No. Random means that any arrangement of outcomes is possible.
DW>There are plenty of realistic applications in which any (known)
DW>deviation from equi-probability could have serious consequences. For
DW>example, the pay-outs in various gambling games are calculated so the
DW>house makes a long term profit, which makes casinos profitable. If
DW>equi-probability were not to occur in whatever "randomizing" mechanism
DW>is involved, and if the players were to know which outcomes were most
DW>probable, they could make a consistent profit, and drive the house into
DW>bankruptcy.
   Having written several complete card game BBS doors (at least one
   which was published with a book on BBSs) I know a little about
   programming card games using QB\PDS.
    In the case of Video Poker, the card deck is shuffled between each
   round of poker so each new deal is dealt from a fresh deck of cards.
   "After the draw, win or lose, the machine deals the next hand from
   a full deck." - Video Poker Mania! by Dwight & Louise Crevelt.
   In my Blackjack game I opted to deal thru one complete deck. I did
   that so that no one could use the game with real money since card
   counters can defeat the house by counting cards. I wanted the game
   to be only entertainment.
DW>There are some legal consequence, too. For example, cars are supposed to
DW>be picked at random for certain kinds of spot-checks by police - for the
DW>drivers' alcohol levels, etc.. If it could be shown that the randomizer
DW>picks some cars more often than others, there would be legal cause to
DW>nullify huge numbers of convictions. This could be chaotic!
   I doubt that happens in the real world. Police, like everyone else,
   will use their experience when doing their job.
DW>Anyway, if you are agreeing that your character shuffling routine *does*
DW>produce some outcomes more often than others, defying the conventional
DW>idea of randomness, and that mine does not, then I think we are in
DW>sufficient agreement for us to stop cluttering this conference with this
DW>discussion. Feel free to send me private netmail, if you want, at
DW>1:250/710.
DW>                             dow
   As I see it you are claiming victory by saying that your random routine
   follows "the conventional idea of randomness" where each outcome comes
   out a fairly equal number of times over a period of time.
    I'm saying that it is wrong to believe that there should be a pattern
   in random outcomes. That there is a pattern (over time) with any random
   generator man can create is due to our inability to create a perfect
   random generator.
    This conference is very slow so I see no problem with this little
  thread.
-Robert
 * OLX 2.1 TD * C-C-C-C-Caffeine, a programmers best f-f-f-friend ...
--- PCBoard (R) v15.3/M 10
---------------