Max Lange attack

Max Lange attack - computer analysis and some thoughts to go with it


1.e4 e5 2.Nf3 Nc6 3.d4 exd4 4.Bc4 Nf6 5.O-O Bc5 6.e5 d5 7.exf6 dxc4
8.fxg7 Rg8 9.Bg5 f6?! 10.Bh6 (this variation, with f6 rather than Be7 was first looked into by the american Horowitz)

The position after 10.Bh6

We will start off with the analyst Horowitz himself speaking:

"A resume of the position here discloses various pluses and minuses for White and Black: White's king is secure, an important point; 2) the White Pawn at g7 is a belligerent threat: it must hurdle only one rank to reach the eighth; 3) Black's "King is insecrure; 4) Black dominates the greater portion of the center. Because Black commands the center and, too, because White's pawn at g7 is tenuous, White must exploit and capitlize his advantages before Black achieves security. Given sufficient time Black will drive the White Bishop from h6, pick off the g-pawn, and rake white on the open g-file. Black must therefore strive to develop his light-squared Bishop and Queen, clear his first rank for Queenside castling, check advances toward his king, and, finally, institute a counter-attack."

The usual way for black to play, and indeed the line that seems both safest and best on the way to equality is:

10 ...Bg4 11. h3 Bh5 12. Qe2+ Qe7! 13. Qxc4 0-0-0 14. Nbd2 Bf7 15. Qd3 Bg6 16. Qc4. This line will not be considered here, for the variation that is safest and best is not always the one that generates the most interest in us.

Black creates more difficulties for himself by playing 10...Qd5 instead of 10...Bg4. Play could continue (and indeed will, since it reaches the position we are interested in here):

11. Re1+ Ne7 12. Nxd4!

The line I'd like to investigate a little further here, or rather, to investigate with the "help" of a computer is this other line, 10...Qd5 (see following diagram). It seems that older engines evaluate the position very differently from newer ones, giving black a solid lead. This I haven't checked myself, but it is an interesting fact, if it is true. If it's not true, it's not a fact, and hence not interesting just at this moment.

The position after 10...Qd5

If we follow Horowitz, which seems very reasonable, with 11. Re1+ Ne7, we reach a very interesting position. Arguably there are better moves than Ne7 for black. Perhaps Be6 would be better, or maybe even Kf7. But, it is not the task at hand to find the better move just here, but rather to look into the way the computer handles the position arising after black plays 11...Ne7.

The position after 11...Ne7

Here Horowitz finds a very dynamic move, at once upsetting the position and creating further complications. 12. Nxd4! This isn’t exactly counter intuitive, but it is a very hard move to find. When we look at it a little closer, we find that 12...Bxd4 13. Nc3 Qc5 14. Ne4 Qe5 15. c3 Bb6 is very troublesome for black, with the fantastic 16. Ng3! to add to the sorrow. This is a line that is more or less forced (the only way to diverge may be the bishop sacrifice at f2 on move 13, which we'll talk a little more about later).

While looking at the position before Horowitz’ move Nxd4, I set Rybka (beta 13d) to analyze it. I left it over night, the analysis reaching a good 20 hours. In all this time it did not think that Nxd4 was the strongest continuation, but rather preferred Nc3. For quite a while during the start of the analysis it preferred Nh4, which also seems to be a good move. The move Nc3, which seems very easy to find for a human, is long considered to be less-than-optimal, before coming out on top after a while, after playing underdog to the likes of Nh5, Qd2 and b4. It seems that Nxd4 is hardly evaluated - and when looking at the evaluations while forcing the moves, it's no wonder.

When forcing Nxd4, it (atleast for a good while) concurs that Bxd4 is correct, followed by 13. Nc3 Qc5 14. Ne4 Qe5 15. c3 Bb6, and then finally 16. Ng3. While carefully studying the evaluations with each move, it is not before white's move Ng3 the engine seems to understand that white is very much better now. From about +1.50 for every move up to Ng3, to +2.50 after.

The position after 16.Ng3

The computer analysis of this position:

FEN: r1b1k1r1/ppp1n1Pp/1b3p1B/4q3/2p5/2P3N1/PP3PPP/R2QR1K1 b q - 2 16

Rybka v1.01 Beta 13d.w32:
3 00:00 80 81.920 +0,38 16. ... Qb5 17.Qf3 Kf7
4 00:00 136 139.264 +0,69 16. ... Qb5 17.Qf3 Kf7
4 00:00 368 376.832 +0,59 16. ... Qc5 17.Qf3 Kf7
5 00:00 672 40.478 +0,44 16. ... Qc5 17.Qf3 Kf7
6 00:00 1.688 36.010 +0,49 16. ... Qc5 17.Qf3 Kf7
7 00:00 6.360 29.602 +0,38 16. ... Qc5 17.Qf3 Kf7 18.Rad1 Qc6
8 00:01 35.848 31.294 +1,01 16. ... Qc5 17.b4 Qxf2+ 18.Kh1 Bf5 19.Re2
9 00:02 90.120 39.086 +2,28 16. ... Qc5 17.b4 Qxf2+
9 00:03 112.184 37.689 +1,36 16. ... Qb5 17.a4 Qc5 18.Ne4 Qf5 19.a5
10 00:05 175.064 36.069 +1,86 16. ... Qb5 17.a4 Qa5 18.Qe2 Qc5 19.a5 Qxf2+ 20.Qxf2 Bxf2+
11 00:08 294.712 37.865 +1,63 16. ... Qb5 17.a4 Qa5 18.Qe2 Qc5 19.a5 Qxf2+ 20.Qxf2 Bxf2+ 21.Kxf2 Kf7 22.Re4
12 00:25 929.392 37.876 +2,50 16. ... Qb5 17.a4 Bxf2+ 18.Kxf2 Qc5+ 19.Kf1 Kf7 20.b4 cxb3 21.Qxb3+ Nd5 22.Ne4 Qc6
13 01:02 2.273.112 37.570 +2,50 16. ... Qb5 17.a4 Qxb2 18.Qh5+ Kd8 19.Rad1+
14 03:39 8.953.976 41.783 +2,87 16. ... Qb5 17.a4 Qxb2 18.Qh5+ Kd8
14 05:17 14.347.440 46.321 +2,60 16. ... Bxf2+ 17.Kxf2 Qc5+ 18.Kf1 Kf7
15 07:24 19.971.432 45.974 +2,47 16. ... Bxf2+ 17.Kxf2 Qc5+ 18.Kf1 Kf7 19.Re4 Ng6
16 18:39 47.702.848 43.656 +2,54 16. ... Bxf2+ 17.Kxf2 Qc5+


There is also a suggestion that after 12...Bxd4 13. Nc3 black could play 13...Bxf2+ 14. Kxf2 Qxd1 15. Raxd1 Kf7 immediately. This is evaluated to +1.45. This seems the more sensible line of play, since after 16. Ng3 in the main variation we are looking at it suggests Bxf2+ anyway, but with a worse game.

Rybka's analysis for this line:

FEN: r1b1k1r1/ppp1n1Pp/5p1B/3q4/2p5/2N5/PPP2bPP/R2QR1K1 w q - 8 14

Rybka v1.01 Beta 13d.w32:
3 00:00 96 98.304 +1,73 14.Kxf2
4 00:00 184 11.083 +1,73 14.Kxf2
5 00:00 552 33.249 +1,73 14.Kxf2 Qxd1
6 00:00 1.256 38.974 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
7 00:00 2.832 60.416 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
8 00:00 6.120 65.967 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
9 00:00 12.688 68.743 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
10 00:00 26.720 72.769 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
11 00:00 55.648 74.294 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
12 00:01 116.736 72.800 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
13 00:03 239.160 70.903 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
14 00:07 532.152 69.187 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
15 00:16 1.078.016 68.255 +1,73 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
16 00:38 2.569.936 67.583 +1,45 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5
17 01:44 6.541.728 63.881 +1,45 14.Kxf2 Qxd1 15.Raxd1 Kf7 16.Nb5 Bf5


The point of all this is the fact that the computer seems to use a very long time finding the best continuations in this murky landscape, while the human eye quickly - or at least quicker - finds the better moves. The true difficulty is finding Nxd4 which seems at least as strong as Nc3 (and of course, in human play, the psychological factor weighs heavy on the game to follow after such a sacrifice). As mentioned earlier, the computer didn't even consider Nxd4 as a top four candidate after 20 hours of brute-forcing it's way through the jungle of moves in this position. Very strange indeed.

The Max Lange attack is a very exciting opening with lots of opportunities and action on the board. I haven't really played it myself, but I might well give it a go, since I greatly enjoy complications and turmoil in a game.


For further reading I recommend:
Kenilworth Chessclub on the Horowitz variation
New Jersey State Chess Federation on the Max Lange attack

For games and a short bio on Max Lange, there is the always excellent Chessgames.com