Author: Bob Durrett
Date: 17:55:50 12/09/03
Go up one level in this thread
On December 09, 2003 at 12:32:15, Robert Hyatt wrote: <snip> >I don't see how this matters since the engine doesn't understand the concept >of "my opponent might make a mistake." <snip> Hmmm. That is an interesting idea! One thing humans do when losing is sometimes to play using a [programmable] strategy of trying to create problems for the opponent and playing to increase the probability of the opponent making a mistake. This would be a good idea if conventional play were perceived to have a low probability of yeilding the draw. A typical decision made by human GMs playing in bad positions is to "complicate the game" in the hopes of increasing liklihood of the opponent going astray. This is just one example of selecting moves on the basis of perceived probabilities. Another example is putting the opponent into a severe bind. One does this not because a clear win is forseen but because one feels, based on past experience, that "an opponent in a bind is likely to not see the hard to find one-and-only defense." Putting the opponent into difficult positions may be 99% of GM chess. Some may argue this point, but I see chess at the top levels as being very much a game of probabilities. I don't know what kind of program algorithms and code would be needed, but the idea is interesting. Perhaps some programmer will make a stab at this. Generally, one must have a "model" of the opponent stored in memory [or imbedded in the code itself]. This model could be expressed statistically or in probabilistic terms. The idea of having and using a model of the opponent is a key idea! Bob D.
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.