Author: Robert Hyatt
Date: 18:34:20 05/22/99
Go up one level in this thread
On May 21, 1999 at 09:36:38, Francis Monkman wrote: >On May 21, 1999 at 02:28:16, pete wrote: > >>Here the winboard engine crashed but it was clear from the previous move lgg >>intended c4 which is a clear win. But if you check the lines with lgg you will >>see : from move 3 to 6 for example lgg judges the position as completely even; >>then it buries its rook on a8 and for many moves has no idea to form a winning >>plan; in fact it was a much closer match than I expected :) >> >>Pete > >Buries a rook?? Musta been dead anyway... > >You say: "for many moves has no idea to form a winning >plan", and this in a sense proves my point "Computers, >however strong, are not really playing chess", which is >essentially about planning. I'm not surprised that Hiarcs, >on playing the pos, finds the harder way to die (this line >is given in Smyslov's notes, although he has 4. (really >22.) ...Ke7 Rxh7 [which is better after 4. ...Ke6 as well, >and then, really, Black has no 'useful moves' to make -- FM] >Ke6 Rf8+- (Sm). At this point computers show only a slight >advantage to White. (I'm quite unsurprised that LGG 2.0 >fares no better (maybe worse) than other machines in this >kind of endgame position.) Whereas I'm quite sure that GMs >would see the position even before Nxf6! as 'won for White'. > >PS If Bob comes along and says "Ah, but LGG 2.0 chose the >move Nxf6! for the 'wrong reason'", I'd point him to the other >guy (sorry, can't remember who!), who said "short of mate, >there can never be a right reason", with which I tend to agree. >Also fair to say that LGG 2.0 predicted the text line for as >long as I watched it, but as I say, Smyslov has foreseen all >eventualities... > >PPS Bob, Crafty 16.6 did *exactly* what you were talking about, >the very next time I powered it up. It chose a move I played >'because I liked it' (while there was an objectively stronger >move to be played) and clung to it (with !!) for 10 minutes -- >which has endeared it to me greatly! (I'll send you the pos, >if you'd like, you'll have to laugh...) > >Francis Always glad to supply amusement. :) But don't miss my point with LGG (or any other program.) If you can produce a position with a move that is a tactically winning move, and a program plays that move without seeing that it is tactically winning, then _that_ is dangerous and probably bad. And that is what I mean by "right move, wrong reason." As if it doesn't see that it is winning, it also doesn't recognize when it might be losing... and that is a 'coin toss'. For non-tactical moves, anything is ok, but if a position has a tactical killer, then (IMHO) a program has to see why it is a tactical killer before playing it... or else it is just 'lucky'. And we all have luck at times. And die for the lack of it at other times.
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.