Again, I appear to have been skipped by a posting, but I wanted to comment
on this statement by JG.
----- Original Message -----
From: "jgw321" <jgw321_at_gmail.com>
To: <fc-solve-discuss_at_yahoogroups.com>
Sent: Wednesday, April 01, 2009 9:29 AM
Subject: Re: Release 2.20.0 is available now with an installer for Windows
> The player was later extended to some huge number
> of games (I forget how many) but to my knowledge it has not been proven that
> they are all unique.
>
The game # is a seed to a random number generator that has a cycle of 2**33.
This means that all sequences of 52 numbers, given a different seed, will be
unique. Thus, every deal that has a different game number (up to 2**33-1)
is dealt from a unique sequence of 33-bit numbers. However, only the 16
least significant bits are actually used to select the next card from the deck.
So the chances of a deal repeating itself over the 2**33 cycle, is the chance
that a sequence of 52 numbers repeats itself in the lower 16 bits. This would
indicate a pretty poor random number generator, and I don't believe that this
one is that bad. Basically, it advances from one 33-bit seed to the next by
multiplying by 214013, adding 2531011, and essentially taking the low
order 33 bits of the result as the next seed. ------ I hope I have this right!
-Gary Campbell
> --
> JG
>
Received on Wed Apr 01 2009 - 14:55:07 IDT