Author: Dann Corbit
Date: 11:03:49 02/12/04
Go up one level in this thread
On February 12, 2004 at 13:49:38, Bob Durrett wrote: > >The short answer: "From some smart people." > >It seems to me that chess algorithms are like scientific theories. They come >and they go. First was the theory that the Earth was flat and then round. >Algorithms are like that too. Chess algorithms must be no exception. Chess algorithms are somewhat like that. When you only utilize a small portion of the earth's surface, a flat model will do nicely. After all, with a 7000 mile diameter sphere, a 100 mile radius has very little bulge to it. Later, people notice that the earth really is a sphere and change their models. This works fine for centuries, and then satellites and other objects notice that the earth is not really a sphere at all but an ellipsoid. An ellipsoidal model works pretty well for a while, but then the earthquake calculations don't work quite right. So some careful measurements show that the earth is really an eccentric ellipsoid. >There is something higher than an algorithm. Worship of a given algorithm seems >terribly dumb to me. I have never met an algorithm worshipper. I think you made it up. I have seen people use alpha-beta variants over minimax, since it examines sqrt(N) nodes instead of N nodes. >Better would be to "worship" the creators of the >algorithms, but than seems dumb too. Worship of the algorithm creators is pretty stupid as well. However, they do deserve a lot of respect because: 1. They spent the time to solve a problem that others found difficult 2. They bothered to explain their idea to other people. >There has to be something at a higher level. Maybe that "something" is >unconstrained creative human thought by someone capable of seeing "the big >picture." It's called math. There are things higher than that too, but math will do for now. >Surely all chess programmers wish to be that something at a higher level. It is >only human. I have no desire to become mathematics, although mathematical systems underlie everything I do, say and think. That's because mathematics is the queen of science. >Of course, there will always be those mindless "algorithm >worshippers." : ) Who are these people? >It has been said that "If you cannot create good computer code, you should not >post bulletins at the Computer Chess Club bulletin board." Obviously that's >wrong. Lots of people say lots of things that are wrong. I don't see any relevance. >It may be time to start looking for new algorithms. It's always time to start looking for new algorithms. The hard part is finding them. I invented a disk based sorting algorithm that is far faster than replacement selection, but it took me a decade to do it. > After all, alpha and beta >are only the first two letters in the Greek alphabet. What about the rest? If you understood how the algorithm worked, you would be standing in admiration of it (as opposed to worship). Making light of the algorithm is silly to me. The current alternative is minimax. >It's time for a new generation of chess software which can both play at a higher >level [planning, etc.] and do new things. It will come. Knowledge expands exponentially over time. >Just my two cents worth. > >Bob D. You seem to want to criticize current practice. However, your criticism is not constructive criticism for several reasons. First of all, you have not identified a problem. "I don't like alpha-beta" is not a problem. Second, you have not offered any sort of alternative idea except "Do something better." Do you really imagine that nobody is looking? Do you further imagine that if we could find something better we would not use it? There are one hundred people in this forum at least who have thought long and hard about alpha-beta. So far, most improvements have been incremental. Here is the nature of the problem: How do you distill an exponential amount of information into something that is not exponential? Imagine a game where each player can make 10 different moves only. After 6 plies, that is one million moves minimax would have to examine. Alpha beta would examine about one thousand. That's pretty good. Now do you see why people rave about it? If we could find something where the nodes examined grew only polynomially (which is still pretty fast) chess would be solved overnight.
This page took 0.01 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.