Author: Vincent Diepeveen
Date: 08:44:49 12/27/02
Go up one level in this thread
On December 27, 2002 at 05:49:28, Gerd Isenberg wrote: >On December 26, 2002 at 15:05:19, Vincent Diepeveen wrote: > >>On December 25, 2002 at 06:21:23, Gerd Isenberg wrote: >> >>>On December 24, 2002 at 19:38:27, Vincent Diepeveen wrote: >>> >>>>On December 24, 2002 at 03:40:24, Bruce Moreland wrote: >>>> >>>>>A backward pawn has the following attributes: >>>>> >>>>>1) It cannot be defended by a pawn. >>>>>2) If it advances, it will be captured by an enemy pawn. >>>>>3) It is now, or can advance to become, the base of a pawn chain. >>>>> >>>>>The classic case is black pawns d6, e5, white pawn e4. >>>>> >>>>>The pawn doesn't have to be on an open file. >>>>> >>>>>I argue that the pawn cannot be a member of a duo, >>>> >>>>I disagree. Some pawns can be member of a duo and backward. >>>> >>>>For example white Rb1,c5 >>>>black b7,c7 Kc8 >>>> >>>>b7 is backward. c5 is not. It is isolated. >>> >>> >>>Hi Vincent, >>> >>>That's interesting. >>>I thought backwardness is independent of pieces (per definition) and could >>>therefore been calculated without considering pieces and stored in the >>>PawnHash-Table?! >> >>This is the major problem of most scientist in computerchess. They >>see one time in their life a definition of something and then use >>that till they are old and grey. > >Or until you elucidates them. regrettably the english-dutch dictionary is downstairs because we have australian guests, but i guess i understand the meaning :) >One amazing thing with backward pawns is, that everybody comes up with sample >positions, but Bas initial questions still remains open, geay area. I must have missed the postings from Bas but i am not sure i would have answerred it anyway (nothing personal Bas!) :) >My current approach don't considers b7 as backward, because of duo, but as a >target for a rook on a halfopen file. >Anyway, the approach Bas mentioned works even well with your definition, if one >considers piece attacks in the domination bitboards. that will be pretty fast with bitboards. Adding up how many attackers and defenders you have whereas i just do a single AND from my attacktable in order to find out how many attackers there are and what kind of attackers ;) >> >>However evaluation is a big grey area. > >Yes for sure. > >> >>For a human b7 in the example is backward. Of course a major problem >>from chess literature versus evaluation in a chessprogram is the classical >>case where in human chess there are only 2 types of bishops. A good one and >>a bad one. >> >>In my chessprogram there are dozens of bishops though so i ran out very >>quickly out of names and invented new ones. >> >>However bishop evaluation is a peanut compared to pawn structure code. >>This is a clear example of that. >> >>Bruce sees it as the result of tactical pressure that b7 is backwards. >> >>That is of course true, but it is a backward pawn from a pawnduo. >> >>Whether you advance c7 to c6 or not. b7 keeps backward. When i play away >>the rook, then b7 is not backwards in DIEP's evaluation but still a little. >> > >OK, but with pawn on c6, b7 is really or better statically backward. It depends upon the position simply. >>>What is the exact reason whether c5 is not backward. >> >>as i said: c5 is isolated pawn. Not a backward pawn. >>c5 is a very strong pawn here. > > >But also isolated pawns may be backward. Isn't it necessary to distinguish >between isolated that may push forward or not? >- - - - - - - - >- - - - - - - - >- - - - - - - - >- - - - B - - - >B - - - - - - B >W - - W - - B - >- W - - - - W - >- - - - - - - - > >Here b2, d3 and h4 are backward, but not e5 because it's more advanced than it's >counterpart d3. b2 is backwards. h4 is backwards. d3 is isolated. e5 is isolated. I do not see d3 in diep as backwards here. I do not see it myself as backwards either. If there would be a pawn on e3 or e4 it sure would be yes. of course there is too many pawns exchanged here to look the pawn structure like a middlegame. However i hope you realize that the further an isolated pawn like e5 goes the easier i could win it for white. getting d3 is very hard though. So i do not understand your definition that d3 is weaker than e5 from that viewpoint. On the other hand if you do not have very good mobility in your program, this is a way to say kind of: "because the pawns of black are further than from white, black has more mobility so let's give the penalty for the pawn structure instead of black bonus for mobility; the netto result is the same evaluation". Bit primitif said but i hope you can understand what i mean. >> >>Again something to go wrong easily. I can remember so many games of >>DIEP at the auto232 players of Jan in the past where a strong pawn >>was by means of tactics very quickly a weak pawn and then the pawn >>was lost and the game some moves later too. >> >>So c5 is a very strong isolated pawn here. > >But i guess it's so strong because of the interaction with b7 and c7/c6. And it >crossed already the boarder, a fact where backwardness may become an advantage, >specially if there is only one (dynamic) backward guard. >The black pawn on c7 disguises the "backwardness" of c5. The weak or dynamic >backward pawn on b7 makes c5 not backward. > >Isn't that the fundamental idea behind minority attacks? Even if the stop square minority attacks is something completely different. >(c6 here) is dominated by black, the remaining pawn structure after bxc becomes >a positional monster (isolated double pawn c7/c6). > >> >>I wouldn't possibly know how you could put backwards pawns in a pawntable. >>Everything is related in chess to the other pieces on the board. In principle >>nothing is independant evaluated in DIEP. >> > >Ok, i use several eval passes, the first one rather statically with only a few >piece interactions, preparing additional data for further passes, where piece >interation is the main issue. > >Regards, >Gerd > > >>I have the pawntable for terms basically which i didn't improve yet too >>well. If i would, the terms would consider things i cannot hash. >> >>For this reason a year or 2 ago i have thrown out the bishop table in >>DIEP. There was not a single pattern left that i could hash independantly >>from bishop+pawns. >> >>It won't be too long before i also get rid of my pawntable. >> >>Already for passed pawns i cannot hash anything anymore. >> >>All i can hash very well is the entire evaluation of a board position, >>because the nullmove and transpositions cause a lot of times that something >>evaluated for white, then i can use for black to move. >> >>Best regards, >>Vincent >> >>>1. no candidate >>>2. if two opponent pawns have backward-distance, >>> the most advanced is not backward. >>>3. because it's isolated. >>>Regards, >>>Gerd
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.