Author: Dave Gomboc
Date: 16:29:02 06/16/99
Go up one level in this thread
On June 16, 1999 at 19:05:29, KarinsDad wrote: >On June 16, 1999 at 18:44:25, Dann Corbit wrote: > >[snip] >>Let's just call it "Markov Process Variations." >>That way it will sound a lot more scientific and important. >>It is often the case that when people say "luck" what they really mean is that >>there is an element of probability involved. Personally, I don't believe in >>luck, but I do believe in probability. > >Good plan. > >I believe in luck in random circumstances such as a large meteor striking the >atmosphere and it is big enough so that a marble sized fragment hits the ground >500 feet away from me. Boy, was I lucky that it did not strike closer. It is an >extremely improbable event in the first place, but there is no normal way to >determine when and where it could happen, hence if you have a close call, you >are lucky that the improbable event did not occur (or if it does occur, you are >unlucky). > >But in computer chess (for the most part), things are deterministic at a >micro-level. Otherwise, you could not play the same variation against a computer >program and identically repeat a game (I am sure that some programmers attempt >to make their programs non-deterministic in some ways). Things are not >deterministic at a macro-level (i.e. you do not control who you will play and in >which round, etc.), but although the macro level affects the outcome of the >tournament, it does not prevent a strong program from defeating all comers. So, >I do not believe in luck as such in a computer chess tournament. > >KarinsDad :) A lot of the micro-level things are subject to fluctuation too. Ignoring the typical example of parallel search, you can still find it because so-and-so set their evaluation term to this weight, but if they had not found a particular test position from a game they read about in the New York Times, it would have been this other weight instead, and so this move would not have been played, and it wouldn't have won. Lucky me. :-) Okay, I'm not actually arguing that position assessment is chaotic or anything, but there is a certain amount of stochasticism in it that must be recognized. It is what allows deeper searches to improve estimates of the position. And the way we generate static evaluations is an extremely ad hoc process. Dave
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.