Author: Stephen A. Boak
Date: 21:27:46 01/13/03
Go up one level in this thread
On January 13, 2003 at 18:18:44, John Merlino wrote: > >Kg8 might be the best move, but it appears that 56...Kf5 blew a won position. >Before that move, Chessmaster 9000 shows a score of about -2.2 for Black. After >56...Kf5 the score is immediately 0.0.... > You may be right that a win was missed by Kasparov (I didn't look at the specific position at all). On the other hand, I've seen many endgame positions where programs show one side is up 2 pawns but the ending is a known draw with accurate play by the defender. I had an adjournment once at my weekly club, with lots of time afterward to analyze before the game got resumed--perhaps several weeks. I had been in a difficult position (I had Rook & two minor pieces vs two Rooks, but I was down some pawns (outside types) that might eventually queen & win the game). I figured out a forced exchange line (over the board, commencing with my sealed move) that traded my two minor pieces for one of my opponent's rooks, and thereby forced the game into a R+3P vs R+P endgame that would hopefully reach (without too much trouble): (White) R + f & h pawns vs (Black) R where the defender (me!) was able to get his king in front of the pawns. I went home after the adjournment to see if my exchange decision would indeed hold the game (reach the hoped for draw). After doing my endgame refresher homework with the famous Reuben Fine book, and Fritz 6 with endgame tablebases, I knew it was going to be a draw--since I had practiced & practiced the lines & techniques. However, until the program was close enough (in my case, close enough was about 21 ply or so) in position to see a tablebase draw within available analysis time, it blithely continued to score the position as +2 for my opponent, who had the extra material. I held the draw when the game resumed. However, I admit I had an uneasy moment or two during the play of the session. There was some time pressure (2nd time control, I guess) and I thought I had slipped up & hadn't played an optimum move and that I had perhaps made a losing move. Not a very comforting feeling! I analyzed the actual finish play at home later, and it turned out that I had never actually waivered from a drawable position. So why had I felt so uncomfortable? I probably had simply entered a not-well-practiced position where very careful (i.e. accurate) play was still necessary to hold the draw, in which I detected (in time--before moving hastily) that there were some relatively easy to overlook possibilities to lose the game with a sloppy move. The realization that I could have gone wrong with a hasty move or two was very disconcerting during the actual play. I certainly sensed the possibility of making a fatal mistake under the time pressure and felt lucky to have not gotten so confused as to make a fatal blunder. LESSONS (oft repeated): 1. Never blindly trust a computer's output--especially in complicated strategic positions or endgames where the comp can't see far enough to get into the endgame tablebases. It often misevaluates the position, and often can & does go wrong because of that. The computer shows its limitations in many kinds of positions--these are just some of them--where moderate strength human judgement could easily be better. 2. Humans can often evaluate [some positions] better than a comp--even non-expert players. For example, I'm only an A-player (approx. 1800 USCF these days), but I knew the comp was wrong in my adjournment example. Of course, knowing that didn't make the game any easier to play, over the board, with a live human opponent trying to pose some difficult questions for me to answer. 3. Just because a position is theoretically drawn doesn't mean a mistake can't be made, or that it can't become complicated enough in practice to go astray! It is hard to practice & memorize *all* the possible lines that can result from a position of 6 or more pieces on the board. 4. Good endgame technique is being able to handle any & all of the possible positions that arise from a starting point--or at least the ability to steer the game into the subset of lines (& positions) that you are able to handle without significant error. --Steve
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.