Author: Don Dailey
Date: 17:38:54 12/25/97
Go up one level in this thread
On December 25, 1997 at 19:01:43, Steve Worley wrote: >Don Dailey's endgame book investigation is interesting and verifies >a well known fact. In endgames, pushing pawns is not overly common. >As he said, a pawn push is often preceded by 20 moves of jockeying of >the other pieces. > >However, this does NOT mean that passed pawn push extentions are not >important. The purpose of extensions is not as a prediction of whether >a move is good or not; it is an attempt to identify moves that are >"critical" and make sure that if the move causes problems the side >to move can deal with them or not. > >In the case of passed pawn pushing, I suspect that the extensions are >important because they help with the non-pawn jockeying which is so >common. The purpose of this jockeying is to get into a situation where >you are finally able to push the pawn. By using pushed-pawn extensions, >you identify those situations where it's finally safe to push. > >Thus, the benefit of pushed pawn extensions is not that it allows you to >push faster, it's actually enhancing your "jockeying" to make sure >you're >setting up (or defending against) that eventual pawn push later. > >Think of check extensions. You don't do them because checks are always >great moves, you do them to make sure that when a check occurs, the side >in check is able to deal with it effectively. > >-Steve I acknowledge the truth of what you are saying. You may have noticed I was careful not to make the claim passed pawn extensions would not help the program understand this position. I would like to point out that probably ANY EXTENSION of any kind will make a program see something sooner in terms of iteration depth but it gets real complicated when you consider TIME which is the crucial thing in computer chess. But after looking at the book again I discover (not surprisingly) that there was a lot more going on than just pushing a single b pawn down the board. This whole maneuver revolved around "fixing" the black king side pawns. The book says that this is required to force a win, otherwise it is a draw. White's move 4.h5! is given an exclamation because it threatens another maneuver 5.Rg5 Kf7 6.Rf5+ and so on. None of these plans involve a single push of a passed pawn and now I believe extending passers in this position would probably hurt the program. I am NOT saying this is a bad thing in every position please understand. And it was said earlier that the whole idea (actually only idea) of endings is to promote pawns. But this is a subgoal that only sometimes is relevant and may not be relevant at all in many endgames even when a passer exists. In this ending it was relevant, but only after a completely different sub-goal was achieved. For a while it was of little or no relevance. The real goal is to checkmate the opponents king, and promoting a pawn must be considered of sencondary or less importance depending on the position in question. But enough of this nonsense. I am curious about which passed pawn extensions these other programs are using and want to see if they will help Cilkchess play better chess. I never said it made your program weaker to have some passed pawn extensions did I? I have this feeling I made an offhand comment about passed pawn extensions being difficult to program and by so doing offended every programmer who uses them. Excuse me!!! At least I now know who the ones are that use them aggressively! Let's see if we can figure out who is doing attack extensions and see what they are doing with them: Attack extensions are really dubious and you should not be using them in your program. If you do, you must surely be a bad person. -- Don
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.