Author: Robert Hyatt
Date: 07:34:22 08/29/98
Go up one level in this thread
On August 29, 1998 at 06:55:39, Serge Desmarais wrote: >On August 28, 1998 at 23:47:30, Robert Hyatt wrote: > > >> >> >>the real problem is to be a pawn down, with a nice lead in development, and >>then suddenly find a way to win a pawn but give up the positional advantage in >>doing so... then it ends up left with a ragged position, maybe equal material, >>and loses the endgame in a slow grind... >> >>Or, if you remember the DB vs Kasparov game 6, an IM tried that position against >>a couple of the top commercial programs last year, with the IM playing black, >>and he won all games easily. Because the programs would invariably find a way >>to "cash in" and get a little material back, but black ends up a piece ahead >>and the win is easy if white's attack fizzles. This is what makes playing such >>openings so very difficult.. As a human, when attacking, after tossing a pawn >>or even a piece, I'm not looking to win a pawn (or the piece) *back*... I am >>looking for much more... something programs often don't grasp... >> >>But one day, maybe they will.. the thing that helps is great search depth, so >>that it can "keep the win of material in sight" but still carry on the attacking >>plan... I've seen much better results as depth has increased. In the old days, >>doing 6-7 plies, such problems were serious, because that's not much horizon to >>both win material *and* keep up the attack. At today's 10-12 plies, the >>programs can do better. But not as good as they will at 15-16-18 plies in the >>future... > > > That is understandable. But I am under the impression that more and more PC >programs are becomming less materialistic. For example, Fritz 5 often gives >pawns. For example, if a pawn it threatened of capture and defending it would >mean bringing the pieces back behind the lines and having a somewhat cramped >position, it would often let go the pawn and to keep active pieces play and try >to make threats directly at the opposing king. Of course, it too had to suffer >and slowly die in a long ending a few times. But sometimes it works and it would >win, or at least obtain a draw because of sufficient compensations. > > > One major problem I found that happens in several programs is in endings with >all pawns on the same side while being a piece down (say K+R+3P vs K+R+B+2P). >And when the opponent would offer a rook trade, they would go for it!. Fritz, >did it while at 17 plies. What is strange is that almost IMMEDIATELY after the >trade, the eval drops damatically below what it was a lot deeper before the >trade! I have seen similar things in Genius too (trading queens to remain with >K+pawns vs K+pawn(s)+B! With a piece down, the opponent can take the opposition >at will with a B or N's move! As for PC programs, they are now much better than >me at finding counterplay and evaluating compensations for material invested, it >seems. (I am just about a 1800-1900 USCF player - rated over 1680 here in Quebec >and as I know one must add between 100 to 200 points to translate our ratings in >USCF). I admit I have much more trouble playing against tactical and aggressive >programs like Fritz, Crafty etc. than playing quiet ones, like Genius, agaisnt >which I am able to draw from time to time at 40/2 (when I succeed in bringing >positions with opposite colour bishops etc.). But against the tactical fast >searchers (aggressive?) it is tough to even pass the 30th move without being at >a disadvantage! > > >Serge Desmarais many programs don't implement the "if ahead, trade pieces but *not* pawns, and if behind, trade pawns but *not* pieces". For reasons that are unknown. So there is plenty of opportunity for swindles in that regard...
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.