Author: Len Eisner
Date: 12:45:56 07/19/98
Go up one level in this thread
On July 19, 1998 at 14:35:22, Don Dailey wrote: >On July 19, 1998 at 13:42:02, john c cook wrote: > >>i have played fritz5 vs rebel and junior and hiarcs6 and it did come out >>on top but can it be number #1 with people. >> look like rebel9 may be #1 with people fritz5 record is not as good as >>rebel9 vs humen am i right or not it was hard for me to belive fritz5 >> as #1 i like rebel9 better > >That's an interesting question. I used to argue that slow positional >programs should be better against humans and fast tactical ones should >be better against other computers (relative to their total chess >playing strengths.) But it has not necessarily turned out to be >that way. > >This might be another example of how a reasonable idea may not be >the correct one, I don't know. The idea that positional programs >should be better against people is based on the idea that positional >play is the computers greatest weakness and that this is how humans >beat them. But no one ever seems to consider that tactics is the >humans greatest weakness against computers and this is how computers >beat them! I don't know why no one ever thinks of this but it is >possibly because of our ego-centric nature. We seem to think that >computers should conform to the human style of playing chess, probably >because we are humans! > >A chess master once told me not to worry about my opponents >too much and play the kind of chess I play best. He said a common >mistake for weaker players when facing stronger ones was to change >their games out of respect for the masters. Most common was that >people would avoid tactics even when it was their strength. > >Then later a tennis pro told me the same thing. He said I should >do what I do best, and my tennis game benefitting a lot from that >advice. I was free to concentrate on my own strengths. > >If this advice was extended to computer chess programs, the advice >would be write programs that maximize the things that chess programs >do best. The point many people miss is that fast programs can >still know quite a bit about chess, the same as I can always improve >on my tennis weaknesses. But it would not make sense to give up >your greatest strengths to do this. > >I don't know what the answer to your question is. Unfortunately >this is a hard question to answer as the data is sparse. I do >remember a program a few years ago that had a strong reputation >as being especially good against people relative to other chess >programs. Against other programs it was good, but not nearly >as impressive. One characteristic of this program was that it >was extremely fast. This was the Novag constellation series >of programs. It was the first program to get an "official" >USCF rating of over 2000 against people. > >I know that Fritz is pretty popular with the stronger players >too. I don't know if it's because of the tactics or because >of the overall chess strength. Don't make any assumptions >yet about which kind of program is better against people. I >really don't think there is a known basis for determining this >although I am quite sure you will get strong opinions on both >sides of the coin. > >- Don I don’t think you can separate tactics and positional play. In the 19th century people believed you could. They thought great players like Morphy could "create" brilliant combinations in any position. Today, we know this is not true. We know that combinations flow from positional advantage. Steinitz was the first to demonstrate the relationship between the two. So the questions is this: if computers play weak positional chess, how can they create positions against strong players that contain winning combinations? Was Stenitz wrong? - Len
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.