Author: Robert Hyatt
Date: 07:33:47 05/15/99
Go up one level in this thread
On May 15, 1999 at 04:58:51, Francis Monkman wrote: > >On May 15, 1999 at 00:51:49, Robert Hyatt wrote: > > >>That's not my point. I have watched LGG play many games on ICC. I have seen >>it have an eval of +1.5 when it is losing, or +1.5 when material is dead even >>and it has no huge positional edge. So what is convincing about an eval of >>+.23? It obviously doesn't see that this is 'winning'. Which means it it >>playing that move for some "other" reason. Most likely totally unrelated >>to what happens assuming this is a tactical win... > >I disagree, and LGG's analysis line bears me out. I quite agree that LGG appears >to get >unreasonably optimistic about its chances without due cause, but this is a clear >case of >its finding what *I believe* to be the winning line (and not, I assure you Bob, >played 'randomly'), for reasons far from 'unrelated' to that line, (and for that >reason, its actual >assessment at such an early stage of the search is far less important than that >it finds >the move). I assure you, I have no wish to waste anyone's time, or to put up >material that >'just happened to work' -- in fact I'm putting these up to be 'knocked down', if >that turns out >to be possible. I'll post the GMs' verdict(s) as soon as I get them. > >All best >Francis Here's the classic 'experiment'. Find _any_ 'brilliant move' you want. Try the position on your favorite program and it most likely won't find it. But keep trying on _all_ other programs... and it is very likely one will find it, for the wrong reason (assuming it is not a queen-sac that loses so much material that you either solve the tactical issues or you can't possibly make the move.) One good example was the move h5 in the DB vs Kasparov match, the move Kasparov said "No computer will play this move...". He was proven wrong. Because after a lot of testing, at least two old versions of Crafty played that move (for the wrong reasons no doubt) and also Junior 5 played it. None of the programs had an eval showing that there was a sound reason for playing the move. It just happened to randomly pop up as the best move. I consider a program to solve a tactical 'win' when (a) they produce a PV that shows that they understood why the move was important and (b) the score shows that the program understood how good (or bad) the move was. But just picking the "right" move is something that can happen for lots of reasons, totally unrelated to playing strength.
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.