[104899] in Cypherpunks
Re: Random array
daemon@ATHENA.MIT.EDU (Greg Rose)
Fri Oct 30 16:37:52 1998
Date: Fri, 30 Oct 1998 10:05:33 +1000
To: Coderpunks <coderpunks@toad.com>
From: Greg Rose <ggr@Qualcomm.com>
In-Reply-To: <3639ED05.2BDEE0F@acm.org>
Reply-To: Greg Rose <ggr@Qualcomm.com>
At 08:44 30/10/98 -0800, Jim Gillogly wrote:
>[bad idea deleted...]
>So the six possible permutations of abc have this probability:
>abc: 5/27
>acb: 5/27
>bac: 4/27
>bca: 4/27
>cab: 4/27
>cba: 5/27
>
>This is a subtle effect and wouldn't be noticeable for 256 elements,
Actually, there is a paper somewhere which I saw once, and have been
totally unable to find ever since, which proves that as the number of
elements increases, the discrepancy in probabilities for this incorrect
algorithm *increases*. By that, I mean that the ratio of probabilities of
the most probable outcomes and the least probable, increases. This is a
counterintuitive result, which I can't prove myself, although it looked
like a correct proof at the time. If anyone knows the proof (or the paper)
I'd really appreciate it. This is one of my pet subjects.
>Please God don't let this mutate into the Monty Hall argument!
I think I'm going to regret asking, but... what is the Monty Hall argument?
Or did you mean the Monty Python argument ("I came here for a good
argument!" "No you didn't... you came here for an argument.")? Feel free to
answer off the list if that would be better.
Greg.
Greg Rose INTERNET: ggr@Qualcomm.com
Qualcomm Australia VOICE: +61-2-9181-4851 FAX: +61-2-9181-5470
Suite 410, Birkenhead Point, http://people.qualcomm.com/ggr/
Drummoyne NSW 2047 232B EC8F 44C6 C853 D68F E107 E6BF CD2F 1081 A37C
