Author: Dann Corbit
Date: 18:24:01 12/30/99
Go up one level in this thread
On December 30, 1999 at 21:05:17, stuart taylor wrote:
>I think that if any program can score atleast 50% against all opponents,
>within four games, say each one having black and white twice one with QP
>type opening and one with KP type opening-then that program is by far the
>best of all of them.If you like, make it 6 or 8 games with the basic
>different style openings for each-as both white and black.
> Such a program, against its top class collegues of today I think would
>qualify as the ideal program, with no real weaknesses, and could acheive
>almost all the (sound) sacrificial masterpeices-those based on tactics,
>and many of those based on (even long -term) possitional understanding-
>which programs so far have not been at all reliable at seeing. It would
>be extremely virtual human intelligence, and its games-very instructive.
>It would likely be more worthy than deeper blue.
> I can't agree with the opinion that he who is a "greater" player or program
>should be seen as
>a matter of opinion. It can be proven using the accurate criteria. The
>intrinsic greatness of the players personality (or the ingenuity and
>insight of the programmer or player) is quite another matter-I agree.
> True, comparing playing strength isn't always so clear cut. But
>when the gap is large, of cause it's patently clear.Therefore when small
>it also exists.
>Any comments?
Yes. It does not work. The best program can lose ten times in a row. It is
unlikely, but it can happen. The only answers you can get by playing programs
{or humans for that matter} against each other are arguments from probability.
Lots of times, we see SSDF results where program A wins the first 4 games and
everyone starts proclaiming that program a giant killer. Then it loses 8 in a
row. That's why the SSDF runs hundreds of games. Otherwise, it's just a shot
in the dark.
Intuition about strength of programs does not work.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.