Author: Uri Blass
Date: 15:30:19 04/20/02
Go up one level in this thread
On April 20, 2002 at 17:27:09, Russell Reagan wrote: >On April 20, 2002 at 16:26:13, Mike Hood wrote: > >[snip] >>Russell, although you're contradicting my basic tenet (ie that computer programs >>don't use strategy, they just use brilliant tactics), our viewpoints aren't so >>far apart. Using your metaphor, the grandmaster is the tortoise and the computer >>program is the hare. The computer program can analyse 100000 positions while the >>grandmaster is analysing a single position. And yet the grandmaster can still >>win. How? The grandmaster makes up for his severe loss of speed by using >>"strategy". He can't outrun his opponent, so he throws the ball at the last >>moment. This is possible at the moment, but one day computers will become so >>fast that not even strategy will save him. As rightly claimed in a parallel >>thread on this topic, tactics -> infinite = strategy > >I think it is possible for computers to make use of positional ideas and to >create a plan in a way similar to humans, but I also think that while they >_could_ do that, they would probably play weaker chess than they do with their >current method. > >Usually strategies (plans) stem from positional ideas. IE "He has a holes at f6 >and h6 so I will move this piece here, then this piece there, then bring up my >knight and I will have a good attack against his king." That is a plan, a >strategy. And what is the point of this strategy? The end goal is to "have a >good attack against his king". In other words, the end goal is tactics. > >Think about other areas of science. We don't understand every little special >case in science. We have physics equations that only work for constant >acceleration (and more accurate ones too). Because we don't understand every >small special case in much of science, we develop general principles to help our >understanding of a particular science, and chess is no different. Take >predicting the weather. If we knew and could measure every possible variable >from the surface of the sun to earth (and probably more would be needed), then >perhaps we could predict with great accuracy what the weather would be like here >on earth at any given moment. We could say "it's going to rain today at 1:43pm >and it will last for 23 minutes and 17 seconds." Since the science of predicting >weather is too complicated for us to currently understand, we form general >principles to help us predict weather. The weather man says, "there is a cool >front moving in here so it will be colder in two days than it is now, and it >might rain tomorrow", but he isn't really sure about any of it. Weather men use >hueristics like chess programs do. Little shortcuts and generalities that get us >started in the right direction, but can't provide a 100% accurate answer. > >Chess is no different from other sciences in this respect. In chess we have the >same generalities that other sciences develop to help us understand complex >positions. We call these generalities positional factors, and they get refined >and refined over time. Eventually as these general theories we have about >positional factors becomes more and more refined, someone will notice something >new. That's how theories work. Someone makes a theory, and if it works for the >most part, it's accepted. Then that theory becomes the generally accepted theory >for a while. Eventually there are too many special cases that the theory doesn't >handle correctly, so a new theory is formed that can handle all of the special >cases. It would be nice to think that eventually in chess we will refine our >theories on positional play so much that we unlock a "secret key" to playing >chess, but I don't think that's going to happen. > >I think chess is purely tactical, and while we have strategy to help our >tactical ideas succeed better, chess really boils down to tactics. When >computers are finally fast enough to "solve" chess, then we can take any >position and say for a fact what the best move is. The down side to that will be >that the reasoning for that move being the best move will be billions and >billions of variations that lead to mate or a draw, so while we might be able to >say what the best move is, we might not have any understanding of why it works. >Imagine trying to understand a 120 move combination :) > >Russell It is not interesting. It is clear that if the computer is fast enough it can solve chess. The real question is what to do when you have not computer that is fast enough. The case of 120 move combination already happens in tablebases thanks to people like nalimov who use a lot of computer time to investigate these endgame and I do not understand the purpose. I could understand building tablebases for a private program in order to win a competition but I do not see a reason to waste a lot of hours of computer time only to give tablebases for free. Helping chess players? I doubt if it helps much. What do chess players earn from it except frustration when they see a mate in 200 in some KRB vs KBN when even after hours of analysis they cannot understand the idea of the moves? Uri
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.