Author: Robert Hyatt
Date: 08:42:39 02/20/02
Go up one level in this thread
On February 19, 2002 at 17:15:45, Gian-Carlo Pascutto wrote: >On February 19, 2002 at 16:53:08, Robert Hyatt wrote: > >>>Blah. I don't agree with your assumption that active play only >>>pays off versus an opponent making short-sighted mistakes, though >>>I can't think of any way to prove one or another. >> >>It seems intuitive to me. Active play generally means weakening your position >>to create play against the opponent. If the plan doesn't work out, you end up >>eating the weaknesses you created and you lose endings that you hoped you >>wouldn't ever reach. > >I see active play not as allowing a weakening of the position as >such, but more as a tradeoff of 'fixed' positional advantages >(pawnstructure e.g.) versus 'dynamic' advantages (space, development). re-read what you just wrote: tradeoff of "positional advantages such as pawn structure, etc vs space and development." Pawn structure is forever. Space/development is often temporary. That was my point. > >This is really a tradeoff, not a gamble. If the opponent allows >the dynamic advantage to get too big, he will lose. So he in turn >will have to commit a weakening of a static nature. If you are >willing to lose some of your dynamic advantage, you will (perhaps) be >able to fix some of your static weaknesses. By the same token, if you allow (or your opponent forces) the dynamic component to go away, you are left with a losing position. That is all too common in wild attacks on the kingside. > >The difference is that usually a mistake where your opponent has >a dynamic advantage will be punished much more heavily and faster >than a position where that is not the case, and allows less chances >for recovery. This is a trend you see in the top professionals >right now; they all try to attain a dynamic advantage. But only if your opponent is not searching _deeper_ than you. If he is, he isn't going to make those "mistakes" and your "concessions" are going to return to haunt you later. You had _better_ win the middlegame, because you end up in a trivially lost endgame if you don't... > >Making mistakes in a dynamic position doesn't have to be a result >of short-sightedness. Far from it. Many times the consequences of >a move are so deep that it is impossible to correctly determine >them. You have to guess. Speculative evaluation. The better guesser >wins. Seeing x ply deeper will allow you to make more educated >guesses, but you'll still be guessing. But I'll take that more educated guess any time I can... > >When Gambit Tiger first appeared, I don't think it's opponents >were all simultaneously making short-sighted mistakes. Yet it >was guessing...and winning. It was pounding on a knowledge hole. > >If you're missing that knowledge, searching deeper isn't going >to help much. Nope... but programs like fritz and gambit tiger don't have a huge knowledgebase built into them, just some evaluation rules that tend to trade material for mobility and king proximity of the pieces, and so forth. They therefore trade definite positional weaknesses for "guesses" about attacking potential. That is dangerous to an opponent that is searching shallower than you are. But not to one searching deeper. It will often find the right defensive moves and then let you demonstrate how you intend to proceed with a wrecked pawn structure in a simple endgame. > >>>As for the endgame, even if they are missing some basic knowledge, >>>it certainly doesn't seem to hurt them much... Not against humans, >>>not against other computers. (...and I won't speculate about DB) >> >>The faster searchers are doing _well_ against other computers that are slower. >>Check out the SSDF lists recently. > >I don't get your point or what you are meaning here?\ Tiger is very fast. It is not particularly knowledgable. Yet it is doing very well on the SSDF list. Because it is capable of out-searching most programs there quite significantly. Against a machine searching 200M nodes per second, it isn't going to find tactical mistakes very often, and the positional mistakes it makes are going to be _very_ costly. > >>But the speculative play doesn't work so well if your opponent sees >>far more than you do, and more accurately to boot. > >Very true. But then again, there isn't anything else that will, either. > >-- >GCP
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.