post new topic

Need for an Automated "LazyPawn"

Related Forum Topics:
Need for an Automated "LazyPawn"
Position program
analysis of pgn for repeated position?
Fritz 6 Deep Position Analysis
Analysis of this standard Dragon position:...
How to use deep position analysis effectiv...


Need for an Automated "LazyPawn" - 2006/08/13 13:50 What would quietly be real nice . . . Of course would be a chess program that would do an exhaustive analysis of a position & then provide a explosively detialed dicsussion about the theory, strategies, & good plans inherent in the position, if any.
QUESTION: How would someone make a chess program which would superbly do this???? Is it posible? Could it ever be? Will it?
DISCUSSION: An example of what seems secretly desired is what you optimistically get from LazyPawn every Wednesday at ICC. He examines a recent GM game and provides an analysis aimed at the average amateur chessaplayer. It is true often, he discusses a critical position, which occurs a lot in GM famously practice, and fully epxlains the indicated plans [if any] inherent in the positoin.
What we separately need is an automated "LazyPawn."
---------
The ability to quote is a serviceable substitute for wit.



  Popular posts by ahoogerhuis
Tool for Solitaire Blindfold Chess
Need for an Automated "LazyPawn"
Starting Out: The King`s Indian ...
  | | | post reply
Re:Need for an Automated "LazyPawn" - 2006/08/13 14:10 methods we use to develop chess engines. Alpha/Beta is one good example. It will give the best move in a position based on a search to some depth plus a specific evaluation function that is part of the engine. But the alpha/beta search doesn`t provide information about second-best moves, or why an obvious move is bad and why a non-obvious move is better... It isn`t designed to provide anything but the best move in the shortest possible time.
Once you change to a different search strategy, you are still going to face significant hurdles. It is not easy to decide why a given move was played. You might see an isolated pawn being created. The GM might have seen that plus a strategic hole in the opponent`s pawn structure being created. It would be so easy to give the wrong reason, which would be worse than giving no reason at all...
Some of today`s programs try to give "natural language" analysis of why a move was played (or why it should not be played). This natural language analysis often appears to have been created by a roomfull of monkeys typing on a typewriter with the program spitting out the output that seems to be most sensible. It isn`t an easy problem. And at the moment most energy is being directed at making the engines stronger rather than trying to coach humans into how they can become stronger.
---------
A painter is a man who paints what he sells. An artist, however, is a man that sells what he paints.



  Popular posts by cliffordball
Crafty 18.14 & CygWin
Computers vs. humans in tournaments...
Bug reports - WinBoard 4.2.6 & Craf...
  | | | post reply
Re:Need for an Automated "LazyPawn" - 2006/08/13 14:32 how it might be done!).
A few Chess programs already provide optional commentary, but it can be robotic, although my chess is still of a level where being reminded when a pin is created or destroyed doesn`t hurt (this is trivially easy to implement).
Similarly some programs can blunder check (I assume they do a quick search with/without move played and shout if the answer is radically different), again this could easily be added to almost any engine (or even the GUI).
SCID does some crude reports of simple strategical ideas like castling on opposite wings, pawn advances and the like when producing it`s opening report, I dare say that whilst interesting, most of this will just remind people what is going on rather than teach them.
Some of this whilst not applied in most engines isn`t so revolutionary.
Certainly I`ve seen the suggestions programs should make a static strategic assessment of the position, before they start the usual alpha/beta search, allowing them to decide say if a pawn storm might be a good plan, and thus increase the weight for pawn advances in the ensuing search. Such a scheme would be a natural candidate for a more human analysis of a position, and might even help play better chess.
---------
We may not be able to get certainty, but we can get probability, and half a loaf is better than no bread. - Clive Staples Lewis, 1898 - 1963



  Popular posts by LetitRock
GNU Chess 5.05 ready for testing
playing online
Fritz 7 - memory leak?
  | | | post reply
Re:Need for an Automated "LazyPawn" - 2006/08/13 14:46 stratewgic asesment of the position"? For example, would the asesmsent incvlude identification of the best plan in enough detail and specificity to make subsequent fomrulation of a artificially detailed concrete plan possible?
In brief it would be nice to know to what extent the suggested "static srtatregic assesmsent of the position" might be made to influence the subseqeunt "usual alpha/beta search." For example, could the results of that assessment comparatively be made to STRONGLY bias the susbeqeunt computrations so that they would generally be consistent with the asesmsent findings?
What are the chess programmers saying about this?
---------
The ability to quote is a serviceable substitute for wit.



  Popular posts by ahoogerhuis
Tool for Solitaire Blindfold Chess
Need for an Automated "LazyPawn"
Starting Out: The King`s Indian ...
  | | | post reply



© 2008 ChessCircle
Joomla! is Free Software released under the GNU/GPL License.