shiv wrote:Maybe a computer is useful to delve into the depths of the problem. However, for any super GM, Bf6 is an obvious move, and calculating further in this line is a piece of cake for most GMs. Even players far weaker than GM strength can see Bf6 almost instantly. The reason this position confuses computers is because both Bf6 and a move like Kf1 lead to a decisive advantage.
I am not discrediting this position, which is a very interesting (and good) one, but I feel that people may not realize how strong super GMs really are. For an example of depth of GM level analysis, have a look at
http://www.chesscafe.com/dvoretsky/dvoretsky.htm
Well I don't know if all top GMs today would play Bf6 in a tournament game if they had never seen Aljechin's game before. Aljechin would, but I don't think all the Grandmasters in the present time's Top Ten are of the same calibre as Aljechin in this type of position. I think there are less than ten.
As for Dvoretsky article: it speaks for itself that we have already done an exhaustive analysis of all the material he presents
Well, that was just a coincidence, I saw his column some weeks ago, it certainly is interesting stuff. What I found was that with less than perfect programs you run into the fact that as long as all moves seem to draw, computers will not make the same distinction as super GMs and super-analysts like Dvoretsky do. Just like you say they make less distinction between Bf6 and Kf1, maybe Qf1 in Aljechin - Sterk if they find winning lines after these moves. For instance according to Mark Dvoretsky, up to a possible 47.f3! everything may still be a draw in the analysis of Kashtanov – Inarkiev and, without guidelines, computers may play "strange" moves because they see no distinctions between the draws. Only if they see any losing sequences on the horizon suddenly they become superb defenders.
Unless the position is too difficult and saves are over the horizon; with deep analysis I was not able to reproduce a move like 47. f3! or the standard plan recommended by Dvoretsky of 45. f3 followed by 46. g4. With Rybka 3 on a fast computer it is very probably easier to find these moves. But 47. f3 was not found, so without guidelines for a strategy here the computer will then probably lose the position even when looking very deep, I was analyzing this for several hours. Assuming 47. f3! is really the only move that gives White a draw at this point.
Earlier on, instead of 44. Kg2 Blueberry at the time, a few weeks ago, preferred 44. Ke2 and this seems to draw just as easily.
Only later in the analysis when White's position is further compromised I started to find more of Dvoretsky's moves, I believe Blueberry found 47... Kd6 after a long search but Ernesto Inarkiev's criticised 47... Ra2?! seemed to give a similar advantage. I would have to do the analysis again with new versions of Blueberry and with better endgame routines in it or stronger programs, EGTB support 6 men if I had them or could use them, to make sure 47... Kd6 wins but 47... Ra2 does maybe not.
That is as far as I got I believe, the article is certainly a good source for testpositions and I hope to return to it later with some new rules for MAT_KRPawnsKRPawns and MAT_KRRPawnsKRRPawns!
What I had so far is really only some very basic bonuses for Rooks behind Passed Pawns,
Code: Select all
if (mat_info->recog == MAT_KRPawnsKRPawns){ Rookendgame = true;
} else { Rookendgame = false;
}
if (mat_info->recog == MAT_KRRPawnsKRRPawns){ Four_Rooksendgame = true;
} else { Four_Rooksendgame = false;
}
if (Rookendgame || Four_Rooksendgame){
rook_one = board->piece[att][0];
rook_two = board->piece[def][0];
if (COLOUR_IS_BLACK(att)) {
rook_one = SQUARE_RANK_MIRROR(rook_one);
rook_two = SQUARE_RANK_MIRROR(rook_two);
}
file = SQUARE_FILE(sq);
rank = PAWN_RANK(sq,att);
if ((SQUARE_FILE(rook_one) == file)&&(SQUARE_RANK(rook_one) < rank)) delta += 40;
if ((SQUARE_FILE(rook_two) == file)&&(SQUARE_RANK(rook_two) < rank)) delta -= 40;
if ((SQUARE_FILE(rook_two) == file)&&(SQUARE_RANK(rook_two) > rank)) delta -= 20;
}
if (Four_Rooksendgame){
rook_three = board->piece[att][1];
rook_four = board->piece[def][1];
if (COLOUR_IS_BLACK(att)) {
rook_three = SQUARE_RANK_MIRROR(rook_three);
rook_four = SQUARE_RANK_MIRROR(rook_four);
}
if ((SQUARE_FILE(rook_three) == file)&&(SQUARE_RANK(rook_three) < rank)){
delta += 40;
if ((SQUARE_FILE(rook_one) == file)&&(SQUARE_RANK(rook_one) < rank)) delta -= 10;
}
if ((SQUARE_FILE(rook_four) == file)&&(SQUARE_RANK(rook_four) < rank)){
delta -= 40;
if ((SQUARE_FILE(rook_two) == file)&&(SQUARE_RANK(rook_two) < rank)) delta += 10;
}
if ((SQUARE_FILE(rook_four) == file)&&(SQUARE_RANK(rook_four) > rank)) delta -= 20;
}
nothing that really is at the level Dvoretsky would like to see
Maybe, probably, Rybka is already far beyond this stage, but Mark does not say what tools he used for his analysis, probably mainly for checking holes in his analysis and that of his pupils.
Regards, Eelco
