The Doomsday of Stockfish DD

[Lyudmil Tsvetkov]

This post will be somewhat off topic, but, as it has a bearing on the human chances to play successfully in a match against a top engine, I will try to see if it is possible to easily draw any game against a top engine by deliberately trying to close the game.

Below 2 games I just played. The aim was simply to close the position to ensure a draw, regardless of whether the moves were not optimal, or even inferior, passive, ugly and so on. I did not play the games to the end, as the outcome was already very evident.

[pgn][PlyCount "82"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.12.04"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish DD 64 SSE4.2"]
[Result "1/2-1/2"]
[ECO "A00"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "41"]
[MLFlags "000100"]

{512MB, Dell XPS 4Cores} 1. c3 {[%emt 0:00:00]} 1... e5 {[%emt 0:00:04]} 2. d3
{[%emt 0:00:01]} 2... d5 {[%emt 0:00:09]} 3. Nf3 {[%emt 0:00:02]} 3... Bd6
{[%emt 0:00:03]} 4. g3 {[%emt 0:00:01]} 4... Nf6 {[%emt 0:00:03]} 5. Bg2
{[%emt 0:00:01]} 5... O-O {[%emt 0:00:03]} 6. O-O {[%emt 0:00:02]} 6... c6
{[%emt 0:00:18]} 7. Nbd2 {[%emt 0: 00:03]} 7... b6 {[%emt 0:00:18]} 8. Qc2
{[%emt 0:00:03]} 8... Re8 {[%emt 0:00:04]} 9. e4 {[%emt 0:00:02]} 9... Nbd7
{[%emt 0:00:05]} 10. h3 {[%emt 0:00:06]} 10... a5 {[%emt 0:00: 05]} 11. Re1
{[%emt 0:00:09]} 11... Ba6 {[%emt 0:00:08]} 12. Nf1 {[%emt 0:00:07]} 12... Nc5
{[%emt 0:00:04]} 13. c4 {[%emt 0:00:00]} 13... d4 {[%emt 0:00:08]} 14. a4
{[%emt 0: 00:07]} 14... Nfd7 {[%emt 0:00:05]} 15. b3 {[%emt 0:00:14]} 15... Bc8
{[%emt 0:00:08]} 16. g4 {[%emt 0:00:06]} 16... Na6 {[%emt 0:00:03]} 17. Ng3
{[%emt 0:00:06]} 17... Ndc5 {[%emt 0:00:00]} 18. Rd1 {[%emt 0:00:09]} 18... Nb4
{[%emt 0:00:04]} 19. Qb1 {[%emt 0:00:03]} 19... g6 {[%emt 0:00:00]} 20. Bg5
{[%emt 0:00:00]} 20... f6 {[%emt 0:00:03]} 21. Bh6 {[%emt 0:00:17]} 21... Ra7
{[%emt 0:00:00]} 22. Kh1 {[%emt 0:00:00]} 22... Rf7 {[%emt 0:00: 04]} 23. Rd2
{[%emt 0:00:11]} 23... Qd7 {[%emt 0:00:03]} 24. Qd1 {[%emt 0:00:08]} 24... Bb7
{[%emt 0:00:03]} 25. Kg1 {[%emt 0:00:09]} 25... Ba6 {[%emt 0:00:07]} 26. Kh1
{[%emt 0:00:03]} 26... Bc8 {[%emt 0:00:03]} 27. Rb1 {[%emt 0:00:19]} 27... Ne6
{[%emt 0:00:04]} 28. Qf1 {[%emt 0:00:00]} 28... Bb7 {[%emt 0:00:03]} 29. Kg1
{[%emt 0:00:02]} 29... Nf4 {[%emt 0:00:02]} 30. Bxf4 {[%emt 0:00:19]} 30... exf4
{[%emt 0:00:03]} 31. Ne2 {[%emt 0:00:02]} 31... c5 {[%emt 0:00:01]} 32. Nh2
{[%emt 0:00:12]} 32... Rg7 {[%emt 0:00:12]} 33. f3 {[%emt 0:00:05]} 33... g5
{[%emt 0:00:03]} 34. Kf2 {[%emt 0:00:00]} 34... Nc6 {[%emt 0: 00:03]} 35. Rdd1
{[%emt 0:00:11]} 35... Qf7 {[%emt 0:00:03]} 36. Qe1 {[%emt 0:00:03]} 36... Qg6
{[%emt 0:00:04]} 37. Qd2 {[%emt 0:00:02]} 37... Ne5 {[%emt 0:00:03]} 38. Rh1
{[%emt 0:00:06]} 38... Qh6 {[%emt 0:00:02]} 39. Kf1 {[%emt 0:00:10]} 39... Qh4
{[%emt 0:00: 06]} 40. Rd1 {[%emt 0:00:06]} 40... Ng6 {[%emt 0:00:04]} 41. Qe1
{[%emt 0:00:03]} 41... Qh6 {[%emt 0:00:01]} 1/2-1/2

[PlyCount "73"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.12.04"]
[White "Stockfish DD 64 SSE4.2"]
[Black "Tsvetkov, Lyudmil"]
[Result "1/2-1/2"]
[ECO "C01"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "36"]
[MLFlags "000100"]

{512MB, Dell XPS 4Cores} 1. e4 {[%emt 0:00:04]} 1... d6 {[%emt 0:00:08]} 2. Nf3
{[%emt 0:00:04]} 2... e6 {[%emt 0:00:06]} 3. d4 {[%emt 0:00:09]} 3... Nc6
{[%emt 0:00:07]} 4. c4 {[%emt 0:00:06]} 4... Nf6 {[%emt 0:00:02]} 5. Nc3
{[%emt 0:00:04]} 5... Be7 {[%emt 0:00:01]} 6. Be2 {[%emt 0:00:04]} 6... O-O
{[%emt 0:00:03]} 7. O-O {[%emt 0:00:01]} 7... e5 {[%emt 0:00:00]} 8. d5
{[%emt 0:00:03]} 8... Nb8 {[%emt 0:00:02]} 9. Ne1 {[%emt 0:00:03]} 9... c5
{[%emt 0:00:02]} 10. Nc2 {[%emt 0:00:05]} 10... Ne8 {[%emt 0:00:06]} 11. Ne3
{[%emt 0:00:04]} 11... g5 {[%emt 0:00:11]} 12. a3 {[%emt 0:00:09]} 12... a5
{[%emt 0: 00:02]} 13. Bd2 {[%emt 0:00:05]} 13... Na6 {[%emt 0:00:06]} 14. Bd3
{[%emt 0:00:08]} 14... Ng7 {[%emt 0:00:06]} 15. b3 {[%emt 0:00:01]} 15... b6
{[%emt 0:00:02]} 16. Qf3 {[%emt 0:00:06]} 16... f6 {[%emt 0:00:03]} 17. Rfe1
{[%emt 0:00:08]} 17... Bd7 {[%emt 0:00: 09]} 18. Qe2 {[%emt 0:00:00]} 18... h5
{[%emt 0:00:14]} 19. Na4 {[%emt 0:00:09]} 19... Rb8 {[%emt 0:00:07]} 20. Rac1
{[%emt 0:00:00]} 20... Qe8 {[%emt 0:00:13]} 21. Nc3 {[%emt 0:00:00]} 21... Nc7
{[%emt 0:00:13]} 22. Rb1 {[%emt 0:00:08]} 22... Qg6 {[%emt 0:00:00]} 23. Rf1
{[%emt 0:00:05]} 23... h4 {[%emt 0:00:04]} 24. g4 {[%emt 0:00:07]} 24... Ra8
{[%emt 0:00:19]} 25. Rfc1 {[%emt 0:00:06]} 25... Rfb8 {[%emt 0:00:03]} 26. h3
{[%emt 0:00:04]} 26... Qe8 {[%emt 0:00:05]} 27. Qd1 {[%emt 0:00:02]} 27... Kf7
{[%emt 0:00:09]} 28. Be2 {[%emt 0:00:04]} 28... Ra7 {[%emt 0:00:10]} 29. a4
{[%emt 0:00:04]} 29... Raa8 {[%emt 0:00:04]} 30. Qf1 {[%emt 0:00:04]} 30... Ra7
{[%emt 0:00:02]} 31. Rd1 {[%emt 0: 00:04]} 31... Raa8 {[%emt 0:00:05]} 32. Nb5
{[%emt 0:00:03]} 32... Nxb5 {[%emt 0:00:13]} 33. cxb5 {[%emt 0:00:03]} 33... Bd8
{[%emt 0:00:03]} 34. Rbc1 {[%emt 0:00:04]} 34... Rb7 {[%emt 0:00:03]} 35. Bc3
{[%emt 0:00:04]} 35... Be7 {[%emt 0:00:04]} 36. Bd3 {[%emt 0: 00:04]} 36... Qb8
{[%emt 0:00:02]} 37. Qe2 {[%emt 0:00:03]} 1/2-1/2
[/pgn]

[d]4r1k1/1b4rp/1p1b1pnq/p1p3p1/P1PpPpP1/1P1P1P1P/4N1BN/3RQK1R w - - 0 42

[d]rq6/1r1bbkn1/1p1p1p2/pPpPp1p1/P3P1Pp/1PBBN2P/4QP2/2RR2K1 b - - 0 37

It seems that chess is simply a straightforward draw that can not be avoided if at least one mistake has not been made. The symmetrical pawn structure of both sides, consisting of 8 pawns, fully guarantees that outcome in spite of the white turn to move. But the reality is that no one cares about dull draws or routine plans prepared beforehand. Therefore, after I play some drawish moulds or routine plans, ensuring myself some points, I quickly switch back to unchartered waters, featuring all kinds of positions, where I might well lose against Stockfish, but will at least learn something in the process.

[carldaman]

I thought Stockfish got much since version 4 better, but maybe it just got marginally better. Well, this was just on a funny note. Seriously, I really think Stockfish has improved, it has deeper understanding of closed positions, but obviously this does not make it immune to defeat.

Thanks for posting these entertaining and enlightening games, Lyudmil!

I can attest to the fact of SF's improvement in closed positions since v4. In my special test suite, it no longer plays the silly Qd1-h5-d1 maneuver I posted about a while back, and actually recognizes Black's attacking potential. Further evidence of this is its winning quickly and in crushing fashion over Komodo 6 on the black side of the same B32 variation.

Edit: here's the link to the old post featuring the dubious Qd1-h5-d1

http://www.talkchess.com/forum/viewtopi ... hlight=bug

(Games to follow when I get home)

Regards,
CL

[Lyudmil Tsvetkov]

Hi Lyudmil,

I always enjoy your posts, especially these ones where SF loses

I am not a good chess player but I also think that using good anti-computer strategy and allowing human to think as long as he wants this can lead still today to some draw or even some win against any engine.

Engines are very strong, but are not tuned for dealing with these kind of positions, simply because do not occur a lot in practice, in engine vs engine matches, that is the main tool developers use for testing new ideas and tuning their engines.

OTH hard-coding some kind of knowledge to deal with these positions, apart from the fact that is far from trivial, it is also absolutely not guaranteed to be an improvement in the average case. And the average case is what the engine is tested against when you run thousands of games, as is routinely done today as part of engine development.

Thanks for posting your games and for clearly annotate them.

Marco

Hi Marco.

And I always enjoy your sense of humour.

Engines are not tuned to deal with similar kinds of positions, because not many engines try to implement them, and there are even fewer with such implemented knowledge to test again. But to tell you the truth, from the few games I played, I am left with the clear impression that Stockfish has laready achieved quite a lot from Stockfish 4 to be able to much better deal with similar positions. I posted only the games I won, but there were also games in such positions where I was outplayed where previously I would not have been. There is a sample of positions in a single pattern where Stockfish succeeds and another sample where it somehow fails. Obviously, other eval and search terms decide when the engine succeeds or fails.

You are absolutely right, if a good idea fails, it is either bad, or not tuned with the whole.

Thanks for the nice words and I wish you and the entire team even bigger success in the future. Stockfish provides joy to many hearts.

[carldaman]

Lydumil,
I've always thought grandmasters would be able to draw almost at will against top engines, if not win their share of games in the way you have, by using 'dull' but effective strategies as shown here, but somehow this hasn't materialized in general practice. Your games [of course, not just the games but your thinking that goes with it], along with those of Dr. Deeb's, etc., are a rare but welcome exception.

This seems very puzzling to me -- it is as if some form of group-think hypnosis has taken hold of the top echelon of the chess community (namely, the titled players) so that they cannot play to their true potential against these top engines. In fact, there are hardly any published recent games that I'm aware between top human players and very strong engines. Again, it is as if everyone has collectively thrown in the proverbial towel, and they will not bother to exert themselves against the 'machines', which turn out to be quite beatable after all.

CL

[Lyudmil Tsvetkov]

Thanks for posting these entertaining and enlightening games, Lyudmil!

I can attest to the fact of SF's improvement in closed positions since v4. In my special test suite, it no longer plays the silly Qd1-h5-d1 maneuver I posted about a while back, and actually recognizes Black's attacking potential. Further evidence of this is its winning quickly and in crushing fashion over Komodo 6 on the black side of the same B32 variation.

Edit: here's the link to the old post featuring the dubious Qd1-h5-d1

http://www.talkchess.com/forum/viewtopi ... hlight=bug

(Games to follow when I get home)

Regards,
CL

Hi Carl.
I will be waiting impatiently to see your games. You might post them here for a clearer perspective. (anyway, the thread is already a bit off topic, but probably not)

[carldaman]

Hi Lyudmil,

I always enjoy your posts, especially these ones where SF loses

I am not a good chess player but I also think that using good anti-computer strategy and allowing human to think as long as he wants this can lead still today to some draw or even some win against any engine.

Engines are very strong, but are not tuned for dealing with these kind of positions, simply because do not occur a lot in practice, in engine vs engine matches, that is the main tool developers use for testing new ideas and tuning their engines.

OTH hard-coding some kind of knowledge to deal with these positions, apart from the fact that is far from trivial, it is also absolutely not guaranteed to be an improvement in the average case. And the average case is what the engine is tested against when you run thousands of games, as is routinely done today as part of engine development.

Thanks for posting your games and for clearly annotate them.

Marco

Hi Marco,

Recent improvements in Stockfish are very encouraging, especially in the way closed positions are played. This clearly shows that it is possible to improve play in this relatively weak area AND at the same also increase its overall strength/rating.

Regards,
CL

[Lyudmil Tsvetkov]

Nice anti-computer strategy gameplay

I prefer to beat them using my well stumbled opening lines regards,
Dr.D

It is important to beat them, to show that humans are not inferior.

I appreciate anyone who plays the computer chess engines and shows some guts

Since the year 2002 I've played 1011 games at long time controls which were 20 minutes + 20 seconds increment then I incresed the time controls to 40 minutes + 20 seconds increment....

I switch off the chess engine thinking lines,no takebacks and the only advantage that I apply for myself is giving myself infinite time control....

I have an elite database with my best games annotated by the top chess engines and commented by myself....

I play the following opening systems mainly:
_With White I play the Larsen-Nimzovitch attack,a marvellous opening system with a lot of potential....

Against the Sicilian I play the Grandprix Attack with a lot of attacking schemes resulting in a beautiful fireworks over the chess borad....

_With Black I play the Pirc Defence against 1. e4 along with the Berlin defence against the Ruy Lopez and the Scandinavian Defence from time to time....

Against 1.d4 I play mainly the KID and the Old Indian Defence which I admire for it's remarkable formations and the Chigorin defence though rarely....

Oh,I play on an official FIDE chessborad as if there is a human in front of me regards,
Dr.D

P.S.Lately,I set up a chess game against some of the oldies but goodies like this one here played recently

[pgn][Event "____12G Match 02"]
[Site "?"]
[Date "2013.11.28"]
[Round "?"]
[White "Dr.Deeb"]
[Black "ChessGenius v1.003"]
[Result "1-0"]
[ECO "A01"]
[Annotator "Deep Rybka 4.1 x64 (20m)"]
[PlyCount "104"]
[Source "Chess Informant"]

{A01: Nimzowitsch-Larsen Opening} 1. b3 e5 2. Bb2 Nc6 3. e3 Nf6 4. Bb5 d6 5.
Ne2 Bd7 6. O-O a6 7. Bxc6 Bxc6 8. d4 e4 (8... Qe7 9. Nd2 O-O-O 10. a4 Qe6 11.
c4 Qg4 12. e4 h5 13. d5 Bd7 14. c5 h4 15. f3 Qg5 16. Nc4 Kb8 17. Qd3 h3 18. g3
dxc5 19. Nxe5 Be8 20. Nc4 Bd7 21. e5 Ne8 22. Qc3 f6 23. e6 {
Revelator (2504)-Simplicissimus (2477) playchess.com INT 2007 1-0 (60)}) 9. c4
d5 10. Ba3 $146 (10. Nbc3 Bd6 11. Ng3 h5 12. cxd5 Bb5 13. Nxb5 axb5 14. Qe2 Qd7
15. f3 Bxg3 16. hxg3 Qxd5 17. Rfc1 c6 18. Rc5 Qd6 19. Re5+ Kf8 20. a4 b4 21.
Rf1 Rh6 22. Qc4 Kg8 23. fxe4 Rg6 24. Rf4 Rxg3 {
Wolf,R-Haist,W (2225) Bad Urach 1986 1/2-1/2 (42)}) 10... Bxa3 11. Nxa3 O-O 12.
Nb1 Qe7 13. Ng3 Bd7 14. Nc3 c6 {Secures b5} 15. c5 {This push gains space} Bg4
{Black threatens to win material: Bg4xd1} 16. Qe1 Nh5 17. h3 Nxg3 18. fxg3 {
White has new doubled pawns: g2+g3} Be6 19. Ne2 Qg5 {
Black threatens to win material: Qg5xe3} 20. Nf4 Bc8 21. b4 h6 22. Kh2 Bd7 23.
Qe2 a5 24. a3 Rfe8 25. Ra2 Bc8 (25... Reb8 26. b5 $11) 26. Rfa1 (26. b5 a4 $14)
26... Qe7 (26... Bd7 27. g4 $11) 27. Qh5 (27. b5 Qd7 $14) 27... Rf8 28. Qd1 (
28. b5 Bd7 $14) 28... Be6 29. Qd2 (29. b5 cxb5 30. Rb1 b4 $14) 29... Ra7 (29...
g5 30. Ne2 $11) 30. Qb2 (30. b5 Raa8 $14) 30... Rfa8 (30... g5 31. Ne2 $14) 31.
Qd2 (31. b5 Re8 $14) 31... Bd7 (31... g5 32. Nxe6 Qxe6 33. a4 $11) 32. Qb2 (32.
b5 f6 $14 (32... cxb5 $143 33. Nxd5 Qg5 34. Nb6 $18)) 32... Qg5 33. Qd2 Qd8 (
33... Ra6 34. a4 $11) 34. b5 $14 {Playing against the pawn chain} Qg5 35. b6
Ra6 36. a4 (36. g4 Kh8 $14) 36... Be6 (36... h5 37. Rf1 $11) 37. Qe1 Bd7 38.
Qf2 Rf8 39. g4 f5 40. gxf5 Rxf5 41. g4 {White threatens to win material: g4xf5.
} Rf7 42. Qg3 Ra8 43. Rg1 Raf8 44. Rf2 Kh7 $2 (44... Qe7 $142 $5 {
would keep Black alive} 45. Rfg2 Qg5 $14) 45. h4 $16 Qe7 $2 (45... Qd8 $142 46.
g5 hxg5 47. hxg5 Rh8 $16) 46. g5 $18 hxg5 47. hxg5 (47. Qxg5 $6 Qxg5 48. Rxg5
Rf5 $14) 47... Rh8 48. Kg2 Kg8 49. Ng6 Rxf2+ 50. Qxf2 Qxg5+ 51. Kf1 Qd8 52.
Nxh8 Kxh8 (52... Kxh8 53. Qf7 Bh3+ 54. Ke1 Qd7 55. Rxg7 Qxf7 56. Rxf7 Bc8 57.
Rf8+ Kh7 58. Rxc8 Kg6 59. Rc7 Kh5 60. Rxb7 Kg5 61. Rc7 Kf5 62. b7 Kg4 63. Rf7
Kh5 64. b8=Q Kh4 65. Qg8 Kh5 66. Rh7#) 1-0[/pgn]

Nice game, Wael.

It is time to take up on Stockfish DD.

[Lyudmil Tsvetkov]

Lydumil,
I've always thought grandmasters would be able to draw almost at will against top engines, if not win their share of games in the way you have, by using 'dull' but effective strategies as shown here, but somehow this hasn't materialized in general practice. Your games [of course, not just the games but your thinking that goes with it], along with those of Dr. Deeb's, etc., are a rare but welcome exception.

This seems very puzzling to me -- it is as if some form of group-think hypnosis has taken hold of the top echelon of the chess community (namely, the titled players) so that they cannot play to their true potential against these top engines. In fact, there are hardly any published recent games that I'm aware between top human players and very strong engines. Again, it is as if everyone has collectively thrown in the proverbial towel, and they will not bother to exert themselves against the 'machines', which turn out to be quite beatable after all.

CL

Hi Carl.

The problem is that you play chess to win, and no one is interested in dull draws. Chess is a draw, there is no doubt about that, but you always play for a win. Otherwise just do not start playing. And when you want to win, you have to take chances. When you take chances, you have a reasonable hope to succeed against fallible players, i.e. humans, but much less reasonable hope to do so against computers that rarely make obvious mistakes nowadays. You might play better chess than the engines, but it is enough to make a single mistake, and it is over. As simple as that.

The other important thing is that no grandmaster in the world has ever paid special attention to anticomputer strategies. They just do not cate about that, because they compete against other humans. The sporadical matches and games that have been played between top level GMs and computers have also featured largely unprepared human players for a serious clash against the machines. Kasparov might have been the best prepared of all, but even he did not certainly pay any particular deeper attention to anticomputer strategies. He wanted to simply play straightforward chess, and already in 2003 it was already close to impossible to do that.

I am sure GMs could do much better against engines, but they need a thorough preparation, involving maybe years of time. No one has simply thought of that.

[Daniel Shawul]

I do not try to prove anything, I am just poking some fun at myself and the world.

You took a shot , and now you are joking? In your previous post you insinuated Stockfish testing is wack compared to Houdart's tests with 10 opponents? I guess you are joking everywhere.

Well the point remains that closed positions games don't prove anything about weakness, infact it probably says more about the author's priority with his engine development,i.e. rejecting useless feature requests . How strong are you btw? If you are a 2000 elo, you should be producing these games (wins) 1 in 1000, not at the rate you are producing them

[carldaman]

These games were played as part of a mini-testmatch, using a special test suite of mine. The position used here is one of my favorites, since it works well to highlight engine strengths and weaknesses. It is a subvariation of the B32 Neo-Sveshnikov Sicilian.

Stockfish (one of the recent versions with Syzygy TB support was tested here) has greatly impressed me. SF4 used to play weakly in this test opening; with White it tended to choose 10. Qd1-h5 followed by Qh5-d1 on the next move, which was really dubious, but now opts for the calm and solid 10. 0-0.

However, that is not even the main story. It is SF's strong and brilliant attacking play with Black in a semi-closed position that leaves a lasting impression. At the same time, on the flipside, SF's opponent, Komodo6 is completely crushed in both games with White -- more badly than I have ever seen Komodo lose, with one loss in 27 moves at a time control of 25m+5s! These 2 games were just as revealing about Komodo's play in this type of position involving Kingside attacks. I hope to eventually test the new Komodo with the same openings.



Here we go:

[pgn]

[Event "15m3s-SFsbK6 test"]
[Site "?"]
[Date "2013.11.29"]
[Round "1.1"]
[White "Komodo 6 64-bit"]
[Black "Stockfish 241113 64 SSE4.2sb"]
[Result "0-1"]
[ECO "B32"]
[Annotator "0.06;0.16"]
[PlyCount "76"]
[EventDate "2013.11.29"]

{Intel(R) Core(TM) i3-2120 CPU @ 3.30GHz 3292 MHz W=19.6 plies; 1,143kN/s B=24.
8 plies; 1,580kN/s 1 core, 512 MB hash each, ponder off} 1. e4 c5 2. Nf3 Nc6 3.
d4 cxd4 4. Nxd4 e5 5. Nb5 d6 6. c4 Be7 7. N1c3 a6 8. Na3 f5 9. Bd3 f4 10. Nd5 {
[%eval 6,21] [%emt 0:00:25] first move out of book is somewhat overambitious,
and probably not White's best; recent versions of SF play the sensible 10.0-0
here, whereas SF4 used to play the awful 10.Qh5} Nf6 {[%eval 16,23] [%emt 0:00:
27]} 11. O-O {[%eval 9,22] [%emt 0:00:53] (Bd2)} O-O {[%eval 0,24] [%emt 0:00:
15]} 12. Bd2 {[%eval 1,20] [%emt 0:00:57]} Be6 {[%eval 8,23] [%emt 0:00:21]
(Qe8)} ({I've even seen} 12... Qe8 {!!?} 13. Nc7 Qg6 $40 {with dangerous
attacking prospects for Black; ironically, Komodo considered this in its PV,
yet look how the game turned out}) 13. Bc3 $6 {[%eval 4,21] [%emt 0:01:27]
seems positionally slow and tactically inaccurate} (13. Nc2 $5 {the Knight
needs to be better centralized anyway} Rc8 14. Rc1 f3 $2 15. Nxe7+ $1 $16 Nxe7
16. Qxf3 $16 Nfd5 17. Qg3 {Black would have no compensation for the pawn here})
13... Rc8 {[%eval 10,23] [%emt 0:00:21]} ({Black could already try} 13... f3 $5
14. g3 Qe8 $1 $40 15. Nc7 $4 Qh5 $19) 14. b3 {[%eval 11,21] [%emt 0:01:26]
(Nc2)} f3 $1 {[%eval -34,21] [%emt 0:00:24] (Qe8) the pawn can't be taken} 15.
g3 {[%eval 10,20] [%emt 0:00:19] (Nc2)} (15. Nxe7+ $6 Nxe7 16. Qxf3 Nfd5 $17 {
this now hits both the Queen and misplaced Bishop on c3}) 15... Qe8 {[%eval
-22,23] [%emt 0:00:25]} 16. Nc2 {[%eval 0,20] [%emt 0:00:23]} Ng4 {[%eval -48,
25] [%emt 0:00:50] (Bd8)} 17. h3 {[%eval 2,19] [%emt 0:00:40]} Nf6 $5 {[%eval
-78,25] [%emt 0:01:16] I'd expected Ng4-h6, but the Knight has already
provoked a further weakness} (17... Nh6 $17) 18. h4 {[%eval 0,20] [%emt 0:00:
21]} Nh5 $1 $17 {[%eval -86,24] [%emt 0:00:19] (Bd8)} 19. Nxe7+ {[%eval -3,18]
[%emt 0:00:46]} Nxe7 $1 {[%eval -82,23] [%emt 0:00:20]} 20. Kh2 {[%eval -16,19]
[%emt 0:00:29] (Ne3)} Nf4 $3 {[%eval -78,22] [%emt 0:00:21] (Qg6) White's
position is totally compromised, but Komodo has no awareness of the deeper
dangers} 21. Re1 {[%eval 0,19] [%emt 0:00:37] (Rh1)} (21. gxf4 Rxf4 $19) 21...
Qh5 $19 {[%eval -397,22] [%emt 0:00:19] (Ng2) and now SF sees itself as
clearly winning} 22. Bf1 {[%eval 0,19] [%emt 0:00:12] (Qd2)} Neg6 {[%eval -522,
23] [%emt 0:00:14] (Nc6)} 23. Kg1 {[%eval -230,17] [%emt 0:00:21] Komodo
finally sees it's busted} Nxh4 $1 {[%eval -545,24] [%emt 0:00:13] Diagram [#]
nice} 24. Qxd6 {[%eval -229,20] [%emt 0:00:21]} Nhg2 {[%eval -587,26] [%emt 0:
00:28]} 25. Qxe5 {[%eval -298,21] [%emt 0:00:20]} Nh3+ {[%eval -604,26] [%emt
0:00:13]} 26. Kh2 {[%eval -304,19] [%emt 0:00:03]} Ng5+ {[%eval -622,26] [%emt
0:00:15] (Qxe5)} 27. Kg1 {[%eval -187,17] [%emt 0:00:03]} Rf6 {[%eval -632,28]
[%emt 0:00:18] hehe } 28. Ne3 {[%eval -297,20] [%emt 0:00:21]} Nh3+ {[%eval
-664,26] [%emt 0:00:13]} 29. Kh2 {[%eval -354,20] [%emt 0:00:05]} Qh6 {[%eval
-731,27] [%emt 0:00:17]} 30. Qxf6 {[%eval -385,21] [%emt 0:00:18] position was
hopeless for White} gxf6 {[%eval -759,26] [%emt 0:00:18]} 31. Nxg2 {[%eval
-397,22] [%emt 0:00:09]} fxg2 {[%eval -791,25] [%emt 0:00:17]} 32. Kxg2 {
[%eval -291,18] [%emt 0:00:03]} f5 {[%eval -800,26] [%emt 0:00:21] (Ng5)} 33.
f4 {[%eval -345,18] [%emt 0:00:16] (exf5)} Kf7 {[%eval -953,23] [%emt 0:00:17]
(Nxf4+)} 34. Kf3 {[%eval -454,18] [%emt 0:00:43]} Rd8 {[%eval -1046,25] [%emt
0:00:23]} 35. Bxh3 {[%eval -613,18] [%emt 0:00:19]} fxe4+ {[%eval -1634,25]
[%emt 0:00:16] (Qxh3)} 36. Rxe4 {[%eval -842,19] [%emt 0:00:10]} Qxh3 {[%eval
-2038,26] [%emt 0:00:17]} 37. Rg1 {[%eval -963,21] [%emt 0:00:21]} Qh5+ {
[%eval -9185,29] [%emt 0:00:11] (Rd3+)} 38. Kf2 {[%eval -1077,21] [%emt 0:00:
13]} Rd3 {[%eval -9305,31] [%emt 0:00:16] adj} 0-1

[Event "25m5s-SFsbK6 test"]
[Site "?"]
[Date "2013.11.29"]
[Round "1.1"]
[White "Komodo 6 64-bit MP"]
[Black "Stockfish 241113 64 SSE4.2sb"]
[Result "0-1"]
[ECO "B32"]
[Annotator "0.15;0.00"]
[PlyCount "54"]
[EventDate "2013.11.29"]
[EventType "tourn"]

{Intel(R) Core(TM) i3-2120 CPU @ 3.30GHz 3292 MHz W=20.8 plies; 2,071kN/s
B=27.8 plies; 2,815kN/s; 100 TBAs} {2 cores, 1 GB hash each, ponder off} 1. e4
c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 e5 5. Nb5 d6 6. c4 Be7 7. N1c3 a6 8. Na3 f5 9.
Bd3 f4 10. Nd5 {[%eval 15,23] [%emt 0:01:08]} Bg5 $5 {[%eval 0,26] [%emt 0:00:
30] (Nf6)} 11. Bd2 {[%eval 19,21] [%emt 0:01:03] (Nc2)} Nf6 {[%eval 0,26]
[%emt 0:00:33]} 12. O-O {[%eval 11,22] [%emt 0:01:11]} O-O {[%eval 0,27] [%emt
0:00:41]} 13. Bc3 $2 {[%eval 17,22] [%emt 0:00:57] (Nc2)} ({better is} 13. Nc2
$142 {so that f4-f3 is prevented by Bxg5}) 13... f3 $1 $17 {[%eval -131,25]
[%emt 0:00:26] (Ne7) now White's position is compromised, and Black gets a
crushing attack against the porous Kingside} 14. g3 {[%eval 11,22] [%emt 0:00:
32]} (14. Nxf6+ $143 Rxf6 $19) (14. Nc2 fxg2 15. Kxg2 (15. Re1 Ng4) 15... Ne7
$40 {with a very strong attack to come} 16. f3 (16. Ba5 $5 Qd7 $1) (16. Nxe7+
Qxe7 17. f3 Nh5) 16... Nfxd5 17. exd5 Bf4) 14... Qd7 $1 $19 {[%eval -200,26]
[%emt 0:01:00] Diagram [#] (Bh3) SF takes 1 min to play the best move} 15. Re1
{[%eval 0,19] [%emt 0:00:16] losing, but so was everything else} (15. Nb6 Qh3
16. Qxf3 Nd7 17. Qxf8+ {forced} Nxf8 18. Nxa8 Ng6 19. Nc2 Nf4 $1 20. Ne1 h5 21.
f3 h4 22. gxf4 Bxf4 23. Rf2 Be3 24. Nc2 Bxf2+ 25. Kxf2 Qxh2+ 26. Ke3 h3 $19) (
15. Nxf6+ Rxf6 $19) 15... Qh3 $19 {[%eval -375,25] [%emt 0:00:34] I'm
surprised it'll take so long for Komodo to see that it is being destroyed,
whereas SF already knows it's winning} 16. Bf1 {[%eval -34,20] [%emt 0:00:45]}
Qh5 {[%eval -466,26] [%emt 0:00:44] (Qh6)} 17. c5 {[%eval -33,19] [%emt 0:00:
43]} Ng4 $19 {[%eval -666,26] [%emt 0:00:56] all very straightforward
attacking moves by SF, yet Komodo only now begins to properly evaluate the
unstoppable danger} 18. h3 {[%eval -322,20] [%emt 0:03:54]} Nh6 $1 {[%eval
-674,27] [%emt 0:00:33]} (18... Nxf2 19. Kxf2 {is also strong, but less clear})
19. Re3 {[%eval -321,19] [%emt 0:01:00] desperate attempt to slow down Black's
attack} Bxh3 {[%eval -777,26] [%emt 0:00:33] (dxc5)} 20. Rxf3 {[%eval -311,19]
[%emt 0:00:38] (Bxh3)} Rxf3 {[%eval -830,26] [%emt 0:00:26]} 21. Bxh3 {[%eval
-334,21] [%emt 0:00:32]} Raf8 {[%eval -868,28] [%emt 0:00:33] resignable
already} 22. Be6+ {[%eval -551,20] [%emt 0:02:50] (Ne7+)} Kh8 {[%eval -858,27]
[%emt 0:00:08]} 23. Nc4 {[%eval -614,21] [%emt 0:01:33]} Ng4 {[%eval -913,31]
[%emt 0:00:34]} 24. Bxg4 {[%eval -635,20] [%emt 0:00:13]} Qxg4 {[%eval -965,32]
[%emt 0:00:30]} 25. Qe1 {[%eval -674,22] [%emt 0:01:08]} Rxf2 {[%eval -1062,32]
[%emt 0:00:34]} 26. Qxf2 {[%eval -697,22] [%emt 0:00:11]} Rxf2 {[%eval -1082,
32] [%emt 0:00:33]} 27. Kxf2 {[%eval -667,23] [%emt 0:00:11]} Qxe4 {[%eval
-1098,32] [%emt 0:00:33] Fritz GUI adj} 0-1

[/pgn]

A couple of diagrams, one from the first game

[D]2r2rk1/1p4pp/p2pb3/4p2q/2P1Pn1n/1PB2pP1/P1N2P2/R2QRBK1 w - - 0 24

and one from the second

[D]r1b2rk1/1p1q2pp/p1np1n2/3Np1b1/2P1P3/N1BB1pP1/PP3P1P/R2Q1RK1 w - - 1 15

Cheers,
CL