Uri Blass wrote:
Larry kaufman never said that the latest rybka positionally is at 2400 level.
Maybe he said that Rybka's evaluation is at that level but it is clearly a different claim.
I am almost sure he said that, and anyway, it is more meaningful to say that than that eval is 2400, how do you quantize the Elo level of the eval?
You simply do not understand the reason that computers are so strong positional players.
I think we have different notions of "positional" here. Read below.
It is because of the search and not because of the evaluation.
Deep search help chess programs to find strong positional moves
because the deep search help them to translate positional advantage that their evaluation does not understand to positional understanding that their evaluation understands.
The computer may see by deep search that it can cause weaknesses in the pawn structure of the opponent.
strong GM may be not able to see it but may consider the move as strong positional move for reasons that the computer does not understand but it is not important because the result is the same.
Both GM and computer play the same strong positional moves.
The computer may even find (thanks to deep search) some strong positional moves that the strong human does not find so the total result is that rybka is stronger than humans positionally.
Fruit2.1 and Strelka are also strong positional players and the evaluation of them is clearly unimportant for the discussion.
They are weaker positionally than rybka3 and part of the reason is that rybka3 searches deeper.
Uri
As a resume, you are simply saying that 20 half-moves search is already a positional understanding, as you said that eval of Fruit 2.1 and Strelka is irrelevant. I disagree. I am talking of positional understanding in the meaning that from middle middlegame a strong GM can see what kind of endgame will result, or by pushing a pawn kingside he knows that the queenside backward pawn will promote. That things are well over the horizon for engines and they have no clue about that. Whenever a serious game appears somewhere, I like to listen to the opinions of strong GM's, for this I am visiting several sites which give real-time comments given by GM's on the game. Their comments are very illuminating, no engine even remotely can help me to understand the quirks and paths that GM sees.
Short bits of commentaries from GM Peter Svidler from
www.chesspro.ru on a Topalov-Kramnik game.
12.Nxd7. 12.Nxg6 would have been completely unreasonable - this bishop is the only piece that Black has to worry about, and White certainly does not want to trade it.
17.a5. Very optimistic - this pawn is more of a weakness than an asset. It is not obvious what to suggest instead, however.
21...Rc6! ...but after this move White sank into thought - Black is just in time. From now on he will have an option of trading one pair of rooks wherever he wants, and with only one rook on the board White's control of the d8 square is rendered useless.
22...Rgc8 23.g4 Bc5. A very logical decision - if White takes on c5 Black is clearly slightly better, due to the fact the one of his rooks will forever be tied down to the defence of the pawn on a5. However, this move allows White to solve all his problems tactically. Black could have waited another move: 23...Be8 - and White would have to find 24.Rad1 Rd6 25.Rxd6 Bxd6 26.Kd3!, preparing to meet 26...Bc5 with 27.Bxc5+ Rxc5 28.b4, to ensure he will not have to suffer too much.
25.Rd7+ Kf8 26.axb6 Rxb6 27.R1d6! Rxd6 28.Rxd6 Rc6! Black could make a draw by other means, for instance 28...Be8!? 29.Rxe6 (29.Bxe6 Rc6) 29...Bf7 30.Rb6 Bxc4+ 31.bxc4 Rxc4 32.Rxb7, and the handshake is not far away - but the text is also very forcing.
I could give you more and more of that, even more relevant stuff, stuff about which engines have no clue. You argument that 20 half-moves are already positional knowledge just contradicts my understanding of positional knowledge.
We can resolve our dispute if a top GM will be allowed to take back on a >0.3 blunder and 3 times at any time against Rybka on any hardware. I can bet that the top GM will win in a 6 games match.
Kai
