Stockfish 301213 - Houdini 4 x64A, 1 CPU Core Test

[mwyoung]

I am currently running a superfast 1000 game test (15 sec+.05 sec) Stockfish 8-thread vs Stockfish 4-thread on my hyperthreaded system. Idle Threads Sleep is set to default value True. After 160 games Stockfish 8-thread is up by 17 ELO.

Keep us informed on your results. I also found it helped stockfish in my testing on this. But it was not 17 elo, it was about half that. But it was a much older version of stockfish.

[lkaufman]

Yes upon further review it appears that I am "totally wrong". Apologies to Mark. I even did some testing, Houdini 4 threads is better than Houdini 8 threads on a hyperthreaded system. I am going outside to shovel a whole lot of snow. Lol

I give you credit for testing your theory, that is more then most have done here in this thread. Keep on testing. Thanks.

After 849 games the Stockfish lead over Houdini (4 cores each) is now 27 elo. So I don't think that Mark's running 8 threads vs 4 was the reason for the SF win, it appears to be just that SF is much better at MP. since Houdini is still clearly better at bullet single core. Mark, I would recommend that you rerun your match setting BOTH sides at 4 threads, not because I think your result was wrong, but because it is not at all clear to me that using 8 threads helped SF. MP performance degrades a lot going from 4 to 8 cores, and while SF seems to compensate much better than Houdini for this, it may not be enough to be a net win. I doubt that the SF guys have tested this much if at all. Also there will always be a suspicion that running 8 threads vs 4 somehow biases the test, although I don't see much evidence of this.

I have done a number of test on this, the conclusion from the test is that it does not hurt stockfish at all, and seems to help some. Not a lot, but a few elo. So since the stockfish team said to use Stockfish with 8 threads of a 4 core system for best results. I test what the stockfish team said was best, since I could find no reasons from my pretest to dispute their claim.

So if it does not hurt, and may help why not test this way. If you are right, it does not change any of my results, if the stockfish team is correct. It may give Stockfish a few more deserved elo points in testing.

I am not a promoter of any program, if the way I test stockfish is hurting it by suspicion but is the best and most accurate way to test Stockfish. So be it, I am not here to convince people of anything. Just share my test results, and test the chess programs the most accurate why I know how.

My concern is that running eight threads vs four may bias the test. I am no hardware expert, but it seems reasonable to me that SF might benefit from Houdini taking extra time to "get going" so to speak. We've noticed in testing Komodo that it is very easy to bias a test, very tiny things can do it. In other words, SF may not actuallyi play better on 8 threads than on 4, but it may deprive Houdini of equal access to resources for a very short time each move. Maybe this would not matter at longer time controls.

[mwyoung]

Yes upon further review it appears that I am "totally wrong". Apologies to Mark. I even did some testing, Houdini 4 threads is better than Houdini 8 threads on a hyperthreaded system. I am going outside to shovel a whole lot of snow. Lol

I give you credit for testing your theory, that is more then most have done here in this thread. Keep on testing. Thanks.

After 849 games the Stockfish lead over Houdini (4 cores each) is now 27 elo. So I don't think that Mark's running 8 threads vs 4 was the reason for the SF win, it appears to be just that SF is much better at MP. since Houdini is still clearly better at bullet single core. Mark, I would recommend that you rerun your match setting BOTH sides at 4 threads, not because I think your result was wrong, but because it is not at all clear to me that using 8 threads helped SF. MP performance degrades a lot going from 4 to 8 cores, and while SF seems to compensate much better than Houdini for this, it may not be enough to be a net win. I doubt that the SF guys have tested this much if at all. Also there will always be a suspicion that running 8 threads vs 4 somehow biases the test, although I don't see much evidence of this.

I have done a number of test on this, the conclusion from the test is that it does not hurt stockfish at all, and seems to help some. Not a lot, but a few elo. So since the stockfish team said to use Stockfish with 8 threads of a 4 core system for best results. I test what the stockfish team said was best, since I could find no reasons from my pretest to dispute their claim.

So if it does not hurt, and may help why not test this way. If you are right, it does not change any of my results, if the stockfish team is correct. It may give Stockfish a few more deserved elo points in testing.

I am not a promoter of any program, if the way I test stockfish is hurting it by suspicion but is the best and most accurate way to test Stockfish. So be it, I am not here to convince people of anything. Just share my test results, and test the chess programs the most accurate why I know how.

My concern is that running eight threads vs four may bias the test. I am no hardware expert, but it seems reasonable to me that SF might benefit from Houdini taking extra time to "get going" so to speak. We've noticed in testing Komodo that it is very easy to bias a test, very tiny things can do it. In other words, SF may not actuallyi play better on 8 threads than on 4, but it may deprive Houdini of equal access to resources for a very short time each move. Maybe this would not matter at longer time controls.

I am playing a longer time control games now. I see about the same results so far. +33elo.

I don't see any evidence, or from my knowledge of hardware. How it could be stealing anything because it is not spinning a thread, or because it is using 8 threads. If what you say is true in anyway that would show up in poorer results at longer time controls for stockfish. I am testing longer time controls on my current test.

Even when you test on 4 threads, and you may not know this. But at many points in time every second all 8 pipes are either working or sleeping, the exact same thing that happens when you use 8 threads with sleeping threads enabled. Just because you are only using 4 threads, does not mean you are restricting the cpu to only 4 pipes. If you use 8 thread with sleeping off, then yes all 8 threads spin at 100%. But even this is not stealing anything or time from the next process.

Sleeping threads frees resources, it does not lock them from the next process. That is the whole point of using sleeping threads.

[RJN]

I have run some time-to-depth tests on Houdini in the past, and got the results below. (I had this written down somewhere, can't find it, but the differences were not huge, 10% or so if I recall)

Threads = Real Cores, HT OFF best results. (task manager will show 50% usage) Note that work is only being split into 4

Threads = Total Real + Logical, HT ON obviously. Middle result (task manager will show 100% usage). While all threads are being used 100%, work is being split by 8, which offsets the gain (NPS may be higher, but time to depth is a bit longer)

Threads = Real Cores, HT ON worse result, task manager will show 50% total usage, with the work being split between all of them, a factor of 8. Worse of all worlds, threads at under-utilized, with the maximum splits.

again, I am going by memory, but it's an easy test to run. anyone can feel free to try. But with the above being said, I think the difference was like 10%, which should not be too much of an Elo hit for any of the 3 possible configurations.

Note: edited due to a brain fart!

[arjuntemurnikar]

UPDATE:

I am doing a mini-test on my laptop.

=====
MacBook Air 4,2
Intel i7 2677M @1.8Ghz (Dual Core)
Running in a Windows 8.1 VM (Parallels Desktop 9) with 1 core and 1GB RAM
=====
TC = 2' + 1" (Two minutes + One second)
Book = AH_150_Opening_Suite
GUI = Fritz 13 GUI
All Engines with 1 thread
Hyperthreading = ON (Don't know how to change on Macbook)
SF Idle Threads Sleep = TRUE (Default)
Houdini Contempt = 0
=====

Code: Select all

SF-311213 Gauntlet (After 600/600 games)

Stockfish 311213 64 SSE4.2   - Houdini 4 Pro x64              145.0 - 155.0    +61/=168/-71    48.33%
Stockfish 311213 64 SSE4.2   - Komodo TCEC 64-bit             170.5 - 129.5    +101/=139/-60    56.83%

=====

PGN: https://www.dropbox.com/s/fkysse2p5zi4p ... untlet.pgn
CBV: https://www.dropbox.com/s/wgwqe8bl53940 ... untlet.cbv

I know its blitz, but still a few interesting games I thoroughly enjoyed:

A beautiful win by Stockfish. First a bishop sac, then an exchange sac, then a kinght sac to top it all off. Amazing watch!
[pgn]
[Event "SF-311213"]
[Site "?"]
[Date "2014.01.02"]
[Round "226.2"]
[White "Stockfish 311213 64 SSE4.2"]
[Black "Komodo TCEC 64-bit"]
[Result "1-0"]
[ECO "B81"]
[Annotator "0.46;0.11"]
[PlyCount "151"]
[EventDate "2013.12.31"]
[EventType "simul"]
[Source "Temurnikar"]

{Intel(R) Core(TM) i7-2677M CPU @ 1.80GHz 1802 MHz W=23.1 plies; 1,339kN/s
B=16.4 plies; 1,128kN/s} 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 e6 6.
g4 h6 {[%eval 11,15] [%emt 0:00:03]} 7. Be3 {[%eval 46,17] [%emt 0:00:07]} Nbd7
{[%eval 10,15] [%emt 0:00:03] (Nc6)} 8. f4 {[%eval 48,16] [%emt 0:00:09] (Be2)}
Qb6 {[%eval 14,16] [%emt 0:00:06]} 9. Rb1 {[%eval 50,17] [%emt 0:00:03]} a6 {
[%eval 15,16] [%emt 0:00:03]} 10. Bg2 {[%eval 42,18] [%emt 0:00:06] (g5)} Qc7 {
[%eval -13,15] [%emt 0:00:02]} 11. h3 {[%eval 12,17] [%emt 0:00:04] (0-0)} Nb6
{[%eval -1,16] [%emt 0:00:03] (b5)} 12. Qd3 {[%eval 28,18] [%emt 0:00:03] (0-0)
} Nc4 {[%eval -10,16] [%emt 0:00:03] (e5)} 13. Bc1 {[%eval 22,20] [%emt 0:00:
04]} b5 {[%eval 5,16] [%emt 0:00:03] (Be7)} 14. b3 {[%eval 10,16] [%emt 0:00:
02] (0-0)} Nb6 {[%eval -9,15] [%emt 0:00:02]} 15. O-O {[%eval 26,16] [%emt 0:
00:02]} Nfd7 {[%eval 0,15] [%emt 0:00:03]} 16. a4 {[%eval 28,16] [%emt 0:00:05]
(Ra1)} bxa4 {[%eval -1,14] [%emt 0:00:02] (Nc5)} 17. bxa4 {[%eval 12,17] [%emt
0:00:03]} Nc5 {[%eval -13,15] [%emt 0:00:03]} 18. Qd2 {[%eval -10,17] [%emt 0:
00:12] (Qe3)} Be7 {[%eval -36,16] [%emt 0:00:08] (Bb7)} 19. f5 {[%eval -26,17]
[%emt 0:00:05] (e5)} Nc4 {[%eval -56,15] [%emt 0:00:03]} 20. Qf2 {[%eval -4,17]
[%emt 0:00:03] (Qd1)} Ne5 {[%eval -42,15] [%emt 0:00:03]} 21. Nce2 {[%eval -12,
18] [%emt 0:00:05] (fxe6)} O-O {[%eval -95,14] [%emt 0:00:02] (Bg5)} 22. Bxh6 {
[%eval 0,18] [%emt 0:00:02] (Nf4)} gxh6 {[%eval -33,16] [%emt 0:00:04]} 23. f6
{[%eval -72,19] [%emt 0:00:03]} Bd8 {[%eval -28,19] [%emt 0:00:06]} 24. Qh4 {
[%eval -10,20] [%emt 0:00:02]} Ncd7 {[%eval -34,19] [%emt 0:00:03]} 25. Qxh6 {
[%eval -34,20] [%emt 0:00:02]} Nxf6 {[%eval -35,17] [%emt 0:00:01] (Bxf6)} 26.
Rxf6 {[%eval -10,19] [%emt 0:00:02]} Bxf6 {[%eval -9,18] [%emt 0:00:09]} 27.
Qxf6 {[%eval -72,20] [%emt 0:00:03]} Qc5 {[%eval -10,18] [%emt 0:00:04] (Rb8)}
28. Rf1 {[%eval -6,20] [%emt 0:00:03] (c3)} Rb8 {[%eval -34,16] [%emt 0:00:03]
(Ra7)} 29. Qh6 {[%eval 0,20] [%emt 0:00:02] (Rd1)} Bd7 {[%eval -82,16] [%emt 0:
00:04] (f6)} 30. Kh1 {[%eval 52,18] [%emt 0:00:02] (Rf6)} Bxa4 {[%eval 11,17]
[%emt 0:00:04] (Nxg4)} 31. c3 {[%eval 113,19] [%emt 0:00:03] (Nxe6)} Rfe8 {
[%eval 73,16] [%emt 0:00:07]} 32. Ng3 {[%eval 110,21] [%emt 0:00:02]} Nd7 {
[%eval 56,18] [%emt 0:00:04]} 33. Ngf5 {[%eval 113,22] [%emt 0:00:03]} exf5 {
[%eval 56,16] [%emt 0:00:04]} 34. gxf5 {[%eval 294,22] [%emt 0:00:10]} Qxc3 {
[%eval 106,17] [%emt 0:00:08]} 35. Rf4 {[%eval 303,22] [%emt 0:00:03]} Qc1+ {
[%eval 114,18] [%emt 0:00:05]} 36. Bf1 {[%eval 351,19] [%emt 0:00:03]} Qxf4 {
[%eval 114,18] [%emt 0:00:02]} 37. Qxf4 {[%eval 377,19] [%emt 0:00:02]} Rb1 {
[%eval 116,17] [%emt 0:00:02]} 38. Kg2 {[%eval 397,19] [%emt 0:00:02] (f6)}
Rb2+ {[%eval 114,14] [%emt 0:00:02]} 39. Kg1 {[%eval 414,20] [%emt 0:00:02]}
Rb1 {[%eval 194,16] [%emt 0:00:05]} 40. f6 {[%eval 440,21] [%emt 0:00:04]} Nxf6
{[%eval 204,18] [%emt 0:00:04] (Re5)} 41. Qg5+ {[%eval 496,19] [%emt 0:00:02]}
Kf8 {[%eval 219,18] [%emt 0:00:02]} 42. Qxf6 {[%eval 523,21] [%emt 0:00:02]}
Bb5 {[%eval 223,19] [%emt 0:00:02]} 43. Qxd6+ {[%eval 531,21] [%emt 0:00:02]
(Nxb5)} Kg8 {[%eval 222,18] [%emt 0:00:01]} 44. Nxb5 {[%eval 545,21] [%emt 0:
00:02]} axb5 {[%eval 222,19] [%emt 0:00:01]} 45. Qd3 {[%eval 553,21] [%emt 0:
00:02]} Ra1 {[%eval 230,19] [%emt 0:00:02]} 46. Qxb5 {[%eval 565,21] [%emt 0:
00:05] (Qg3+)} Rxe4 {[%eval 208,15] [%emt 0:00:01]} 47. Qg5+ {[%eval 585,21]
[%emt 0:00:02] (Qb8+)} Kf8 {[%eval 171,16] [%emt 0:00:01]} 48. Qc5+ {[%eval
593,18] [%emt 0:00:01]} Re7 {[%eval 171,17] [%emt 0:00:01]} 49. Kf2 {[%eval
597,20] [%emt 0:00:02] (Kg2)} Rd1 {[%eval 198,16] [%emt 0:00:05] (Rae1)} 50.
Bc4 {[%eval 648,18] [%emt 0:00:02] (Bb5)} Rdd7 {[%eval 212,16] [%emt 0:00:02]
(Rd2+)} 51. h4 {[%eval 688,19] [%emt 0:00:03]} Rc7 {[%eval 256,16] [%emt 0:00:
03]} 52. Qh5 {[%eval 705,21] [%emt 0:00:01]} Red7 {[%eval 286,15] [%emt 0:00:
02]} 53. Qh6+ {[%eval 705,21] [%emt 0:00:01]} Ke7 {[%eval 275,15] [%emt 0:00:
01] (Kg8)} 54. Qf4 {[%eval 712,19] [%emt 0:00:01]} f6 {[%eval 303,16] [%emt 0:
00:02] (Rc5)} 55. h5 {[%eval 785,18] [%emt 0:00:02]} Rc8 {[%eval 329,16] [%emt
0:00:03]} 56. Qe4+ {[%eval 1076,18] [%emt 0:00:03] (Qe3+)} Kd8 {[%eval 335,15]
[%emt 0:00:01]} 57. h6 {[%eval 1127,16] [%emt 0:00:01]} Rc5 {[%eval 334,17]
[%emt 0:00:02] (f5)} 58. h7 {[%eval 1191,16] [%emt 0:00:00]} Rh5 {[%eval 362,
16] [%emt 0:00:01]} 59. Qa8+ {[%eval 8181,23] [%emt 0:00:02] (h8R+)} Kc7 {
[%eval 364,16] [%emt 0:00:01]} 60. h8=R {[%eval 8894,23] [%emt 0:00:02] (h8Q)}
Rxh8 {[%eval 372,17] [%emt 0:00:01]} 61. Qxh8 {[%eval 11067,28] [%emt 0:00:04]}
Rd6 {[%eval 563,18] [%emt 0:00:02]} 62. Qf8 {[%eval 14134,28] [%emt 0:00:06]
(Qg7+)} Rb6 {[%eval 772,18] [%emt 0:00:03]} 63. Qf7+ {[%eval 14836,20] [%emt 0:
00:04]} Kd8 {[%eval 884,19] [%emt 0:00:01] (Kc8)} 64. Qd5+ {[%eval 32734,19]
[%emt 0:00:02] (Be6)} Kc8 {[%eval 32618,17] [%emt 0:00:02] (Ke7)} 65. Qc5+ {
[%eval 32736,18] [%emt 0:00:01]} Kb7 {[%eval 32668,16] [%emt 0:00:00]} 66. Bd5+
{[%eval 32738,18] [%emt 0:00:01]} Ka6 {[%eval 32706,16] [%emt 0:00:00]} 67. Qc7
{[%eval 32740,19] [%emt 0:00:01]} Rb4 {[%eval 32692,16] [%emt 0:00:00]} 68.
Bc4+ {[%eval 32742,18] [%emt 0:00:01]} Rxc4 {[%eval 32688,15] [%emt 0:00:00]}
69. Qxc4+ {[%eval 32746,20] [%emt 0:00:01]} Kb7 {[%eval 32680,14] [%emt 0:00:
00] (Kb6)} 70. Ke3 {[%eval 32754,24] [%emt 0:00:01]} Kb6 {[%eval 32748,14]
[%emt 0:00:00]} 71. Kd4 {[%eval 32756,28] [%emt 0:00:01] (Qe6+)} f5 {[%eval
32758,16] [%emt 0:00:00]} 72. Kd5 {[%eval 32758,43] [%emt 0:00:01]} Kb7 {
[%eval 32760,17] [%emt 0:00:00] (f4)} 73. Qc6+ {[%eval 32760,84] [%emt 0:00:01]
(Kc5)} Kb8 {[%eval 32762,15] [%emt 0:00:00] (Ka7)} 74. Qb6+ {[%eval 32762,100]
[%emt 0:00:00]} Ka8 {[%eval 32764,18] [%emt 0:00:00] (Kc8)} 75. Kc6 {[%eval
32764,100] [%emt 0:00:00]} f4 {[%eval 32766,19] [%emt 0:00:00]} 76. Qb7# {
[%eval 32766,100] [%emt 0:00:00]} 1-0
[/pgn]

Another blazing tactic from SF. Its amazing that stockfish on such miserly hardware (Virtual Machine even!) and at such fast blitz time controls, can come up with such an epic tactic. Komodo was blinded in an instant.
[pgn]
[Event "SF-311213"]
[Site "?"]
[Date "2014.01.02"]
[Round "228.2"]
[White "Stockfish 311213 64 SSE4.2"]
[Black "Komodo TCEC 64-bit"]
[Result "1-0"]
[ECO "B76"]
[Annotator "0.38;0.26"]
[PlyCount "77"]
[EventDate "2013.12.31"]
[EventType "simul"]
[Source "Temurnikar"]
[FEN "1r2r1k1/3q1bb1/3pp1p1/3Pn1P1/p1PB2BQ/1N6/P6R/5RK1 w - - 0 36"]

{Intel(R) Core(TM) i7-2677M CPU @ 1.80GHz 1799 MHz W=19.0 plies; 1,021kN/s
B=16.0 plies; 962kN/s} 36. Nc5 {[%eval 1593,19] [%emt 0:00:02] (Qh7+)} dxc5 {[%eval 970,14]
[%emt 0:00:03] (Qc7)} 37. Qh7+ {[%eval 2585,24] [%emt 0:00:02]} Kf8 {[%eval
1022,17] [%emt 0:00:00]} 38. Qxg7+ {[%eval 3391,27] [%emt 0:00:02]} Kxg7 {
[%eval 1242,20] [%emt 0:00:02]} 39. Bxe5+ {[%eval 9666,30] [%emt 0:00:02]} 1-0
[/pgn]

Finally an attacker's dream position --- The Ultimate Pawn Storm!
[d]r3qrk1/pnp1n1p1/bp1p1p1p/8/2PPPPPP/P1BBN1Q1/8/1R3RK1 b - - 0 26

The game:
[pgn]
[Event "SF-311213"]
[Site "?"]
[Date "2014.01.03"]
[Round "33.1"]
[White "Stockfish 311213 64 SSE4.2"]
[Black "Houdini 4 Pro x64"]
[Result "1-0"]
[ECO "E32"]
[Annotator "0.26;0.12"]
[PlyCount "83"]
[EventDate "2013.12.31"]
[EventType "simul"]
[Source "Temurnikar"]

{Intel(R) Core(TM) i7-2677M CPU @ 1.80GHz 1799 MHz W=17.9 plies; 1,054kN/s
B=15.8 plies; 1,479kN/s} 1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. Qc2 O-O 5. e4 d6 6.
a3 {[%eval 26,18] [%emt 0:00:05]} Bxc3+ {[%eval 12,18] [%emt 0:00:03]} 7. bxc3
{[%eval 22,20] [%emt 0:00:03]} e5 {[%eval 11,19] [%emt 0:00:06] (c5)} 8. Bd3 {
[%eval 38,18] [%emt 0:00:03]} Nc6 {[%eval 11,18] [%emt 0:00:03]} 9. Ne2 {
[%eval 30,19] [%emt 0:00:03]} b6 {[%eval 10,18] [%emt 0:00:02] (Na5)} 10. O-O {
[%eval 42,17] [%emt 0:00:02]} h6 {[%eval 9,17] [%emt 0:00:02] (Ba6)} 11. f3 {
[%eval 44,18] [%emt 0:00:05] (Be3)} Ba6 {[%eval 9,17] [%emt 0:00:04]} 12. Qa2 {
[%eval 36,18] [%emt 0:00:04]} Qe8 {[%eval 6,17] [%emt 0:00:03] (Na5)} 13. Rb1 {
[%eval 40,18] [%emt 0:00:03] (a4)} Na5 {[%eval -7,16] [%emt 0:00:02] (Rc8)} 14.
Ng3 {[%eval 68,18] [%emt 0:00:04]} Qe6 {[%eval 22,16] [%emt 0:00:09] (Rd8)} 15.
Nf5 {[%eval 76,18] [%emt 0:00:05] (d5)} Nh5 {[%eval 25,16] [%emt 0:00:03] (Kh8)
} 16. Qf2 {[%eval 66,19] [%emt 0:00:05] (d5)} Nf4 {[%eval 17,15] [%emt 0:00:02]
(Qf6)} 17. Qg3 {[%eval 90,18] [%emt 0:00:02]} Ng6 {[%eval 59,16] [%emt 0:00:16]
(Qf6)} 18. h4 {[%eval 92,18] [%emt 0:00:03]} Qf6 {[%eval 62,15] [%emt 0:00:05]}
19. f4 {[%eval 179,19] [%emt 0:00:03]} exd4 {[%eval 135,15] [%emt 0:00:06]} 20.
cxd4 {[%eval 212,18] [%emt 0:00:02]} Ne7 {[%eval 132,16] [%emt 0:00:02]} 21.
Ne3 {[%eval 222,20] [%emt 0:00:03]} Qg6 {[%eval 164,17] [%emt 0:00:06]} 22. Qe1
{[%eval 210,17] [%emt 0:00:03]} f6 {[%eval 155,16] [%emt 0:00:03] (f5)} 23. Bd2
{[%eval 246,16] [%emt 0:00:03]} Nb7 {[%eval 162,15] [%emt 0:00:03]} 24. Bc3 {
[%eval 230,16] [%emt 0:00:06] (Qd1)} Qh5 {[%eval 149,16] [%emt 0:00:02] (Rad8)}
25. g4 {[%eval 240,16] [%emt 0:00:05] (Be2)} Qe8 {[%eval 129,15] [%emt 0:00:04]
(Qf7)} 26. Qg3 {[%eval 270,15] [%emt 0:00:02] (Qe2)} Kh8 {[%eval 160,14] [%emt
0:00:06] (Rd8)} 27. g5 {[%eval 284,15] [%emt 0:00:03]} Rg8 {[%eval 158,15]
[%emt 0:00:04]} 28. e5 {[%eval 261,16] [%emt 0:00:19] (Kh2)} f5 {[%eval 168,14]
[%emt 0:00:03]} 29. Rf2 {[%eval 278,15] [%emt 0:00:07] (Kh2)} Rd8 {[%eval 182,
14] [%emt 0:00:05]} 30. d5 {[%eval 266,16] [%emt 0:00:03] (Bc2)} Nc5 {[%eval
148,14] [%emt 0:00:01]} 31. Bc2 {[%eval 264,18] [%emt 0:00:03]} h5 {[%eval 210,
15] [%emt 0:00:03]} 32. Nxf5 {[%eval 327,18] [%emt 0:00:03]} Nxf5 {[%eval 231,
16] [%emt 0:00:02]} 33. Bxf5 {[%eval 365,19] [%emt 0:00:03]} Rf8 {[%eval 270,
15] [%emt 0:00:01]} 34. Bc2 {[%eval 391,20] [%emt 0:00:03]} dxe5 {[%eval 230,
15] [%emt 0:00:01]} 35. Bxe5 {[%eval 412,20] [%emt 0:00:03]} Kg8 {[%eval 289,
16] [%emt 0:00:03]} 36. Re1 {[%eval 474,20] [%emt 0:00:05]} Qd7 {[%eval 290,14]
[%emt 0:00:02]} 37. f5 {[%eval 521,20] [%emt 0:00:02] (Qc3)} Bxc4 {[%eval 452,
16] [%emt 0:00:09] (Rde8)} 38. Qf3 {[%eval 541,19] [%emt 0:00:02] (Bxg7)} g6 {
[%eval 496,15] [%emt 0:00:07] (Bd3)} 39. Bb2 {[%eval 725,20] [%emt 0:00:02]
(Ba1)} Nd3 {[%eval 775,16] [%emt 0:00:07]} 40. Re6 {[%eval 1129,19] [%emt 0:00:
02] (Bxd3)} Qf7 {[%eval 1065,15] [%emt 0:00:06]} 41. Bxd3 {[%eval 2016,15]
[%emt 0:00:02]} Bxd5 {[%eval 1101,15] [%emt 0:00:01]} 42. Qe2 {[%eval 32748,17]
[%emt 0:00:06] (Be4)} 1-0
[/pgn]