Author: Peter Kappler
Date: 21:55:33 06/08/99
Go up one level in this thread
On June 09, 1999 at 00:21:30, KarinsDad wrote: >On June 08, 1999 at 23:50:14, Peter Kappler wrote: > >[snip] >> >>There is no question that Garry makes mistakes in his games - I didn't mean to >>imply that he is anywhere close to playing perfect chess. >> >>The key question is how wide is the drawing margin in chess? There are >>countless material-down endings that are drawn, and we all have seen games where >>one side has a nagging edge for the entire game but it isn't quite enough to >>convert to a win. > >However, we do not know whether the "nagging edge" advantages wouldn't convert >to easy wins with a perfect tablebase. There are many endings which the best >endgame GMs in the world thought were wins or draws (after a lot of analysis) >and the endgame tables proved that the result was different than expected. If >this could occur repeatedly with 4 and 5 piece endgames, why would you think >that 32 piece tablebases would not immediately slam even the most subtle of >mistakes (let alone multiple subtle mistakes made throughout an entire game)? > I think it really just depends on the magnitude of the mistake. Also it is not clear to me that having more pieces on the board necessarily implies that a mistake will be more severely punished. I think just the opposite might be true... >Just due to the SHEER complexity of chess, it seems that no human could ever >compete against a perfect tablebase. I estimated Kasparov would at *best* draw one game out of ten. That isn't exactly a fantastic perforance. Out of the other nine games, I think maybe half would be interesting struggles where Garry would go down slowly, and the other half would see Garry getting blown off the board. > I guess the reason this seems obvious to me >(regardless of whether it is correct) is that I have a theory that when humans >play against each other, they accidentally fall into inferior or superior >positions. In other words, a computer may see that a given move has a tactical >mistake 14 ply down, but a human does not see it. He looks down 10 ply and sees >no problem and in fact, may think that he is winning. A few moves later, his >opponent suddenly finds himself in a winning position and does not even realize >how he got there. I agree that this happens frequently at the amateur level, but I feel pretty certain that this is quite rare at the Super-GM level. Computers have this same problem. Their evaluation says that >they are up a pawn and within 5 moves for each side, the winning side is >suddenly losing. > >With a perfect tablebase looking down 150 ply, I can conceive that even opening >moves which are perfectly valid or strong right now could be proven to be >flawed. > >Does this make sense? > Absolutely - we just disagree on how lopsided the contest would be. It does make for an interesting discussion! :-) --Peter >KarinsDad :)
This page took 0.02 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.