The Importance of Move Orders?
Wednesday, March 3, 2010 at 3:10PM A good opening book will discuss move order finesses. It's important to know that in a certain variation it's better to play A first and only then B, because in that way you cut out certain of the opponent's options. Well and good, but what about this, from Victor Bologan's new book on the King's Indian?
[After the moves 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Be2 0-0 6.Bg5 Na6 7.Qc2, Bologan writes]
7...h6
White's position is slightly better after 7...c5 8.d5 h6 9.Bf4.
8.Be3
Or 8.Bf4 c5 9.d5 [further analysis ensues, all of which is supposed to be fine for Black]
Of course, the position is exactly the same in the two cases. Both times Black plays h6 and c5, and both times White plays Bf4 and d5. But if Black plays ...h6 first, then White is better, and if he plays ...c5 first, then White's not! It's for subtleties like this that I get opening books, and Bologan's ability to suss out such fine points makes it clear what a vast gulf there is between super-GMs and the rest of us.
(Seriously, though: this gaffe notwithstanding, I'm a fan of Bologan's writings in general and the book in particular. But I do wonder what he was thinking when he wrote that.)
Reader Comments (6)
What is the difference between fine for black and white having a slight advantage? It is darn close, and generally when I am black and white has a small advantage out of the opening, i would call my position/play fine.
But point taken regarding the inconsistency, I wonder how much proofing these type of books go through.
Here's a nontrivial experiment. Enter the two transposing move sequences into your favorite chess-engine interface. Load your favorite engine, give it just 1 core if you're on a multi-core computer (the parameter dialogue may say "Threads" or "CPUs" or "Processors"), and put it into analysis mode. If your engine has a multi-PV setting, set it for (say) 10 PV's. Jump to the position after White's 9th in each case, and clear the hash table before beginning an open-ended ("infinite") analysis...
See if you get different results, for some lines at certain search depths, between the two "identical" positions. You may have to wait a long time to see any difference. Repeat each, and see if the differences reproduce. (The point of using just 1 core is that owing to changing OS memory schedules between cores, runs even with completely deterministic engines generally do not reproduce.)
If you do, here is the most likely explanation. The standard FEN/EPD notation for a chess position includes the # of half-moves since the last capture or Pawn move. Some engines include a term for this value in their evaluation function in case the 50-Move Rule is a factor, and/or include this value in the hash signature of a position. In one line the value is 0 because White has just played 9.d5, while in the other with 9.Bf4 it is 1. Maybe just maybe, this slight difference can cause a hash collision that propagates all the way to the final evaluation of some line at some depth, which you can observe by saving the output in the engine-analysis window. I can reproduce such a case with the chess engine Toga II beta4.5c, where a difference between 0 or 2 in this value causes a 6.5-pawn upsurge in the evaluation of a drawn position, but that is in an endgame not an opening.
Yes of course- if I can summarise Mr Regan's post, strictly speaking the positions are NOT "exactly the same" as it would take a different number of moves in either case to reach a 50-move rule draw. Not that that should have affected Mr Bologan's judgement of the position...
If I ever start a band (or a chess team) I think I'll call it 'Hash Collision'.
Dennis, check this out...
Check this out:
In one section the line 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 0–0 6.Nge2 c5 7.Be3 is given. 7.d5 is the main move, but after 7.Be3 (which a long note) 7...Nc6 the are two oddities. First, 8.dxc5 dxc5 9.Qxd8 Rxd8 10.Bxc5 is given with some analysis and game fragments. However, this position is a simple transposition to the line 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 0–0 6.Be3 c5 7.dxc5 dxc5 8.Qxd8 Rxd8 9.Bxc5 Nc6 10.Nge2, which is covered in much greater detail later on using completely different game fragments.
Even stranger is that in the same line (1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 0–0 6.Nge2 c5 7.Be3 Nc6 ) after 8.d5, the author(s?!) say that the main move 8...Ne5 "has been severely criticized by theory and quite deservedly at that. We recommend another retreat to the edge of the board - 8...Na5!?" Huh? Of course, when this position is covered via the move order 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 0–0 6.Be3 c5 7.Nge2 Nc6 8.d5, the main move 8...Ne5 is given without comment, followed by several pages of analysis! Of course 8...Ne5 is the move, and perhaps 8...Na5 could have been given as an interesting and less explored alternative. But come on, does anyone really believe that this was Bologan's doing? 2700 players do not miss such obvious transpositions.
David,
If it is not Bologan's doing does that imply he had parts of the book ghost written? I can't believe an editor would make large changes in the structure and content of a 2700s GM analysis.
Brian,
That it's partially ghost-written doesn't seem impossible to me. Kasparov used Plisetsky, and I've heard on several occasions that Khalifman's books are team efforts. A little disappointing, but certainly possible. But it's also possible that Bologan did do it, and at a certain point got sick and tired of the project and was careless about some transpositions.
By the way, do editors of chess books do anything? I find it very easy to find errors of all sorts (facts, grammar, easily checkable analysis, etc.) in chess books, without really looking for them, so maybe that's just a title given to the person who turns a ChessBase or other file into the format preferred by the publisher. Or - and this strikes me as vile - an "editor" is someone who takes a classic book and changes the original text (invariably without giving the original, and sometimes not even acknowledging that any change was made) to "improve" it. It's one thing to make changes, in consultation with the author, before the initial publication (THAT is what an editor is supposed to do); it's another thing altogether to falsify history!