Author: Alastair Scott
Date: 08:13:38 09/21/03
Go up one level in this thread
On September 21, 2003 at 09:00:16, Uri Blass wrote: >I looked at the game and I do not expect humans to play moves like Ne5 and I >include weak humans that I can beat without a queen but not without queen and 2 >rooks. > >They make tactical mistakes but not this kind of mistake. > >If I see something that plays Ne5 I can be almost sure that it is a machine(the >only other option is that it is a beginner). > >Uri I think this is a "great unsolved problem" - nothing I have _ever_ seen plays a convincingly weak game. Kasparov Chess is a particularly instructive example. The (notional) grading range is 600-2300. At 2300 you most certainly get Ruffian hot and strong; at anything below that obviously absurd moves are made, at increasing frequency as the "grade" goes lower. The result is that the actual dropoff in strength is exponential, not linear, and the only sensible option is to play at 2300! I suspect that the algorithm used is something ridiculous like "with probability 1-(x-600)/1700 of the time play the worst move in the search rather than the best". Alastair
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.