Author: KarinsDad
Date: 12:04:34 04/14/99
Go up one level in this thread
On April 14, 1999 at 13:09:44, Bruce Moreland wrote: [snip] > >People often talk about the ending as a place to add sophisticated knowledge, >but I've never heard an example. > >bruce There is a story where 2 famous GMs (I forget which two, I will have to look it up) were analyzing an endgame and Capablanca walked over and instantly made a comment such as "Oh, this is the continuation.". Neither of the other two GMs had even seen the continuation and when they questioned Capablanca on how he had known it, he replied something like "You think, I know." or some such. If I can find the exact GMs, comments, and ending, I will post it. The fact is that some people have a much greater grasp of endgame knowledge (Capablanca was one of them) and it would seem that some of this knowledge can be used. Once you take simple endings and make them more complex, a computer is well suited to handle those complexities. The difficulty lies in the human arena of creating an algorithm which is specific enough to pick the best (or nearly best) moves, but generic enough to handle many different types of endings. For example, if you ask a beginning chess player how to mate with KR vs. K, he usually does not know (most of us probably forget the days when that problem would have been difficult for us). But once the concept of forcing the lone king back is understood, it is even easy to write an algorithm to do this. If you have a sophisticated enough algorithm, you could even add "result info" into the system so that you did not need tablebases. For example, if I know that a given ending is a win for one side, as a human, I have a much greater chance of discoverying that win than if I did not have this information. I once played a game where I had forced the enemy king out into the open and I KNEW there had to be a way to mate him. It took me 15 minutes to find it, but because my intuition told me that there had to be a win there, I kept searching until I found it (my opponent resigned immediately after I made the first move since he was higher rated and had seen the continuation way before I did). I think that it may be possible to program a computer to just know the results of the tablebases (and hence you could eventually add in larger tablebases) without knowing the exact mechanism to that result (i.e. the program can figure it out on it's own once it KNOWS what the expected result is). This could drastically decrease the size of the tablebases since the moves could be calculated on the fly. Just some ideas. KarinsDad :)
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.