Test Position: Karjakin-Kramnik (analysis)

[Albert Silver]

Karjakin won a spectacular game that started with a knight offer on g5. Kramnik did not take but the question remained whether he could. He could but it would have been extremely delicate. Here is one line that would go bad on him, though engines take a long time just seeing some of the continuations, so it may not be findable in a reasonable time.

[D]r3r1k1/pp1qbpp1/2pp2b1/4n1P1/2P3P1/4BP2/PPPQB3/1K1R3R w - - 0 18

The engines I used choose 18.Rdg1 but this leads to nothing. 18.Bd4! is the move.

[meschle]

Rybka 4 multipv confirms this - but not by a great margin


Analysis by Rybka 4 x64: depth 17

1. = (0.19): 1.Bd4 Bd8 2.f4 Be4 3.Bxe5[] Bxh1 4.Rxh1 Qe6 5.Bd3 dxe5[] 6.Bh7+[] Kf8 7.Bf5[] Ba5[] 8.Qxa5[] Qd6[] 9.c5 Qd8 10.Qe1 Qd5 11.g6 Rad8 12.a4 f6 13.g5 fxg5 14.Qg1
2. = (-0.04): 1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6[] Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rf3 9.Re1 a5 10.Bf5 a4 11.a3 Rf8 12.Rd1 Be5 13.c5

[Dirt]

Karjakin won a spectacular game that started with a knight offer on g5.

Apparently not computer preparation, either.

[Albert Silver]

Rybka 4 multipv confirms this - but not by a great margin


Analysis by Rybka 4 x64: depth 17

1. = (0.19): 1.Bd4 Bd8 2.f4 Be4 3.Bxe5[] Bxh1 4.Rxh1 Qe6 5.Bd3 dxe5[] 6.Bh7+[] Kf8 7.Bf5[] Ba5[] 8.Qxa5[] Qd6[] 9.c5 Qd8 10.Qe1 Qd5 11.g6 Rad8 12.a4 f6 13.g5 fxg5 14.Qg1
2. = (-0.04): 1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6[] Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rf3 9.Re1 a5 10.Bf5 a4 11.a3 Rf8 12.Rd1 Be5 13.c5

After 1.Bd4 Bd8 2.f4 Be4

[D]r2br1k1/pp1q1pp1/2pp4/4n1P1/2PBbPP1/8/PPPQB3/1K1R3R w - - 0 20

20.Bxe5 Bxg5 21.Bxg7[] Kxg7 22.Rh5 Re5 23.Rxg5+ Rxg5 24.fxg5 Re8 25.Qc3+ Kg8 26.Bf3 Bxf3 27.Qxf3 Re5 28.Qf4 Qe7 29.b3 Kg7 30.Kb2 a6 31.Qd4 d5 32.cxd5
+/- (1.18) Depth: 18 00:02:36 25504kN

[Albert Silver]

Karjakin won a spectacular game that started with a knight offer on g5.

Apparently not computer preparation, either.

I can believe it. I think Bf5 looks lousy, so how could Karjakin expect it?

- move before was Be6, so this is another tempo with bishop.
- bishop is placed in line of fire of pawns

consider that black has had to suffer through several nasty kingside attacks, just this year: Topalov-Gelfand, Ivanchuk-Gelfand, Nakamura-Kramnik, and one wonders how he justified the first two points.

[M ANSARI]

There is absolutely no way on earth that Karjakin did not heavily analyze his initial knight sac with heavy computer usage by him and by his seconds ... simply no way. He would have looked into many different lines where things can quickly go bad for black, and really it was an extremely difficult position for black to play knowing that your opponent most likely has the position analyzed to the nth degree with a 3200 ELO engine. Kramnik did not take on g5 even though he probably thought that was best simply to avoid the lines his opponent might have analyzed heavily. Also he played Bf5 in a very sharp position, moving the bishop twice and losing a valuable tempo, again most likely to avoid the heavily analyzed lines and go back to playing the player rather than the heavy computer analysis. He might have had a few chances to draw as Karjakin did not play the strongest continuation, but wow ... what an incredibly difficult game to play for black. Kramnik did the same thing against Nakamura, but in the case of Naka he did not have the guts to burn his bridges and go for the win ... Karjakin certainly did and deserves a heap of credit for having such courage and crushing one of the strongest players in the history of chess.

This only goes to show how weak even the strongest humans are against today's top engines. They simply cannot handle such complicated games on the fly without computer assistance.

[jdart]

Arasan 12.2, analysis, 4000 seconds/move:

"Karjakin-Kramnik, 2010 (analysis)" bm Bd4
result: Bd4 score: -0.16 ++ solved in 2828.85 sec. (8422.53M node
s)
Bd4 f6 Rh2 d5 gxf6 Bxf6 cxd5 Qxd5 g5 Nxf3 Bxf3 Qxd4 Qxd4 Bxd4 Rxd4 Re1+ Bd1 Rg1
Rhd2 Rxg5 Rd7 b6 R2d4
result(2): Rdg1 score: -0.63 ** not solved in 4000.08 secs. (
11869.20M nodes)
Rdg1 Qe6 f4 Be4 fxe5 Bxh1 Rxh1 g6 exd6 Bxd6 Rh3 Bc5 Qc3 Qe5 Bxc5 Qxc3 bxc3 Rxe2
Bd4 f6 gxf6 b6
result(3): Rh2 score: -0.16 ** not solved in 4000.15 secs. (
11634.15M nodes)
Rh2 f6 Bd4 d5 gxf6 Bxf6 cxd5 Qxd5 g5 Nxf3 Bxf3 Qxd4 Qxd4 Bxd4 Rxd4 Re1+ Rd1 Rxd1
+ Bxd1 Re8 Rh1 Re5 Rg1 Be4 Re1

[peter]

After 1.Bd4 Bd8 2.f4 Be4

[D]r2br1k1/pp1q1pp1/2pp4/4n1P1/2PBbPP1/8/PPPQB3/1K1R3R w - - 0 20

20.Bxe5 Bxg5 21.Bxg7[] Kxg7 22.Rh5 Re5 23.Rxg5+ Rxg5 24.fxg5 Re8 25.Qc3+ Kg8 26.Bf3 Bxf3 27.Qxf3 Re5 28.Qf4 Qe7 29.b3 Kg7 30.Kb2 a6 31.Qd4 d5 32.cxd5
+/- (1.18) Depth: 18 00:02:36 25504kN

Good test position!
Cause your main line of R4 above seems a little optimistic with 20... Bxg5, to me 20... Bxh1 looks better (3... in my notation), I made a little analysis of my own with the main line 20. f5 (3. f5):

[Event "Silver took N sac"]
[Site "?"]
[Date "2010.11.14"]
[Round "?"]
[White "Karjakin"]
[Black "Kramnik"]
[Result "*"]
[Annotator "Martan,Peter"]
[SetUp "1"]
[FEN "r3r1k1/pp1qbpp1/2pp2b1/4n1P1/2P3P1/4BP2/PPPQB3/1K1R3R w - - 0 1"]
[PlyCount "27"]
[EventDate "2010.??.??"]

1. Bd4 (1. Rdg1 $6 f5 (1... Qe6 $6 2. f4 Be4 3. Rh5 Bxg5 4. fxe5 (4. fxg5 b6 5.
Rh2 Nxc4 6. Bxc4 Qxc4) 4... Bxe3 5. Qxe3 Bg6 6. Rg5 Qf6 7. Rh1 dxe5 8. Qh3 Kf8
9. Qa3+ Kg8) 2. gxf6 Bxf6 (2... gxf6 $2 3. f4) 3. f4 Nf7 4. f5 Qe7 5. fxg6 Qxe3
6. gxf7+ Kxf7) 1... f6 (1... Qe6 $6 2. f4 Be4 3. f5 (3. Bxe5 $6 Bxh1 4. Rxh1
Bxg5)) (1... Bd8 2. f4 Bxg5 (2... Nxg4 $6 3. Bxg4 Qxg4 4. Qh2) (2... Be4 3.
Bxe5 Bxg5 (3... Bxh1 $6 4. Rxh1 Bxg5 (4... Qe6 $2 5. Qd3 $1) 5. Qd3) 4. Bxg7)
3. fxg5 Be4 4. Rh3 Rad8 5. b3 b6 6. Bxe5 Rxe5 7. Bf3 Bxf3 8. Qh2 Kf8 9. Rxf3
Qe6 10. Qh8+) 2. gxf6 Bxf6 3. g5 Bd8 (3... c5 $2 4. Bc3 Bd8 5. f4 Bxg5 (5...
Nf7 $6 6. Bd3 Re4 (6... Bf5 7. Qh2) 7. Rhe1 Rxe1 8. Rxe1 Nh8 9. Rh1 Bxg5 10.
fxg5) 6. fxg5) 4. f4 Bf5 5. fxe5 dxe5 6. Bd3 exd4 7. Qh2 Bxd3 8. cxd3 Qf5 9.
Qh8+ Kf7 10. Qh5+ Qg6 11. Rhf1+ Bf6 12. Qh4 Rh8 13. Qf4 Rhf8 14. gxf6 *

[Taner Altinsoy]

Houdini finds the right move but thinks black is still a bit better. AMD quad9850 1gb hash

Analysis by Houdini 1.03a x64 POPCNT 4_CPU:

1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 6/19 00:00:00 4kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 7/35 00:00:00 4kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 8/35 00:00:00 5kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 9/35 00:00:00 7kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 10/35 00:00:00 9kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rf8 9.Rd1 Rf3 10.c3 Be5 11.Bf5 Re8 12.Rdf1 Rh8 13.Rxf3 Rxh1+ 14.Kc2 Rh2+ 15.Kb3
= (-0.11) Depth: 11/35 00:00:00 14kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd1 Rf8 9.g5 Be5 10.Rf1+ Ke7 11.Rxf8 Kxf8 12.c3 Ke7 13.Rh7
= (-0.11) Depth: 12/35 00:00:00 38kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.gxf7+ Kxf7 7.Qxe3 Rxe3 8.Bd3 Rae8 9.a3 Rf3 10.g5 Be5 11.Rf1 Rxf1+ 12.Rxf1+ Ke6 13.Bf5+ Ke7 14.Bg6 Rf8 15.Rxf8 Kxf8 16.c3 Ke7 17.Kc2
= (-0.13) Depth: 13/35 00:00:00 75kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd1 Rae8 9.g5 Bd8 10.Rf1+ Kg8 11.Bh5 Rf8 12.Rd1 Bc7
= (-0.11) Depth: 14/35 00:00:00 123kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd1 Ke7 9.g5 Be5 10.Rf1 Re8 11.c3 Rg3 12.Bh5 Rf8 13.Rxf8 Kxf8
= (-0.10) Depth: 15/35 00:00:00 676kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 16/35 00:00:00 678kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 17/35 00:00:00 928kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Rae8 9.a3 Ke7 10.Bg6 Rf8 11.Rf1 a6 12.Rh5 Re2 13.Bd3 Rd2 14.Re1+ Kd7 15.c5 dxc5 16.Rxc5 Re8 17.Rxe8 Kxe8
= (-0.10) Depth: 18/36 00:00:00 1345kN
1.Rdg1 f5 2.gxf6 Bxf6 3.f4 Nf7 4.f5 Qe7 5.fxg6 Qxe3 6.Qxe3 Rxe3 7.gxf7+ Kxf7 8.Bd3 Ke7 9.Bf5 Re8 10.Rd1 Re2 11.Bg6 Rf8 12.Bf5 Be5 13.c3 Bf4 14.Rhe1 Rxe1 15.Rxe1+ Kf6 16.Kc2 Be5
= (-0.10) Depth: 19/38 00:00:01 2888kN
1.Rdg1 Qe6 2.f4 Be4 3.fxe5 Bxh1 4.Rxh1 g6 5.Bd1 Kg7 6.exd6 Bxd6 7.Qf2 Be5 8.b3 b6 9.c5 b5 10.Bf3 Rh8 11.Rh6 Rad8 12.Qg2 Bc3 13.Bc1 Qe5 14.Bxc6 Qxc5
=/+ (-0.27) Depth: 20/50 00:00:22 64562kN
1.Rdg1 Qe6 2.f4 Be4 3.fxe5 Bxh1 4.Rxh1 g6 5.Bd1 Kg7 6.exd6 Bxd6 7.Qf2 Be5 8.b3 b6 9.c5 b5 10.Bf3 Rh8 11.Rh6 Rad8 12.Qg2 Bc3 13.Bc1 Qe5 14.Bxc6 Qxc5
= (-0.21 !) Depth: 21/50 00:00:25 73952kN
1.Rdg1 Qe6 2.f4 Be4 3.fxe5 Bxh1 4.Rxh1 g6 5.Bd1 Kg7 6.exd6 Bxd6 7.Qf2 Be5 8.b3 b6 9.c5 b5 10.Bf3 Rh8 11.Rh6 Rac8 12.Qg2 Qd7 13.Be4 Kg8 14.Qf3 Bg7 15.Rxh8+ Bxh8
=/+ (-0.29) Depth: 21/50 00:00:43 121mN
1.Bd4 f6 2.gxf6 Bxf6 3.g5 Bd8 4.f4 Be4 5.Bxe5 Qe6 6.Bh5 Bxh1 7.Rxh1 dxe5 8.Bxe8 Qxe8 9.Qh2 Kf7 10.fxe5 Qe6 11.Qf2+ Ke7 12.Qc5+ Kd7 13.Rh7 Qg6 14.Qd4+ Kc7 15.Qd3 Qxg5 16.Qd6+ Kb6 17.Qd4+ Ka6 18.e6 Qf5
= (-0.09 !) Depth: 21/66 00:01:58 332mN

(, Microsoft 14.11.2010)

[peter]

Houdini finds the right move but thinks black is still a bit better.

...

1.Bd4 f6 2.gxf6 Bxf6 3.g5 Bd8 4.f4 Be4 5.Bxe5 Qe6 6.Bh5 Bxh1 7.Rxh1 dxe5 8.Bxe8 Qxe8 9.Qh2 Kf7 10.fxe5 Qe6 11.Qf2+ Ke7 12.Qc5+ Kd7 13.Rh7 Qg6 14.Qd4+ Kc7 15.Qd3 Qxg5 16.Qd6+ Kb6 17.Qd4+ Ka6 18.e6 Qf5
= (-0.09 !) Depth: 21/66 00:01:58 332mN

If you show 6. Bxg7! instead of the 6. Bh5?! in your main line of Houdini above to it, you'll see its eval climb.