SmarThink 1.70 released

[Sergei S. Markoff]

http://genes1s.net/smarthink.php

SmarThink v1.70 — July 2014 release of the engine. Major changes:

— output TT move before search in UCI mode (better output for analysis);
— code for pawn shelter/storm was completely rewritten;
— massive re-tuning of evaluation terms (material, passers/candidates, king attack etc.).

[Sergei S. Markoff]

Thank you, Sergei. Not sure when I'll have time to test it. One of the reasons is that tests of 1.60 are still running...
Cheers,

Thank you for your testing!)
1.70 looks to be at least 60 ELO points stronger and may be even more)

[Graham Banks]

http://genes1s.net/smarthink.php

SmarThink v1.70 — July 2014 release of the engine. Major changes:

— output TT move before search in UCI mode (better output for analysis);
— code for pawn shelter/storm was completely rewritten;
— massive re-tuning of evaluation terms (material, passers/candidates, king attack etc.).

Thanks Sergei. Looking forward to giving it a run.

[Jamal Bubker]

Thank you Sergei !

[Wolfgang]

Thank you, Sergei. Not sure when I'll have time to test it. One of the reasons is that tests of 1.60 are still running...
Cheers,

Thank you for your testing!)
1.70 looks to be at least 60 ELO points stronger and may be even more)

Hi Sergei,

thanks!!

Is this new version compatible to AMD-processors now? Would be nice, but I still see "Intel" in the names of the .exe only...

Thanks and best wishes
Wolfgang
CEGT-Team

PS: to be clear, I'm talking about x64-version. w32 is ok, but not of real interest (for me)

[Sergei S. Markoff]

Is this new version compatible to AMD-processors now? Would be nice, but I still see "Intel" in the names of the .exe only...

32-bit compile at the time only(

SmarThink\SmarThink_v170_x32.exe

64-bit AMD is planned for 1.80 (will be released after few weeks).

[carldaman]

http://genes1s.net/smarthink.php

SmarThink v1.70 — July 2014 release of the engine. Major changes:

— output TT move before search in UCI mode (better output for analysis);
— code for pawn shelter/storm was completely rewritten;
— massive re-tuning of evaluation terms (material, passers/candidates, king attack etc.).

Thanks for your continued (free) development of ST, Sergei.

Regards,
CL

[Frank Quisinsky]

Hi Sergei,

I like your engine very much ... since many years.

Only king safty is a little problem for SmarThink versions. The attacking style in combination with good strategic moves after the openings are great. The style your engine is playing, no other engine have. Perhaps a bit to compare with Junior. Junior's style is also completly an other and very interesting.

It's clear that if you find interest again that your engine jumping to the best in a short time. Hope that you not lost your interest again.

Scala ... 1-6 (1 is the best) reagarding playing style ...
Your engine is one of my favorits and I think 1,5 you will get from myself.

Your new version is scheduled for FCT1 around August 20th, 2014. Perhaps Version 1.8 will be ready to this time? I am sure that the database with 45-minutes games on Haswel i7 4.3 GHz can help you to find mistakes. Can be download on my site.

Frank's Chess Page
http://www.amateurschach.de

Best
Frank

[Werner]

Hi,
just watched a game where SmarThink failed in a 5men endgame although all 5 men tbs are loaded: move 107 ...Rxg3 ??
Werner

[d]R7/7k/8/8/6K1/1r4P1/8/8 b - -
Engine: SmarThink 1.70 x64 (256 MB)
von Sergei S. Markoff
100/100 0:00 0.00 107...Tb4+ (2.736) TB:16
Bester Zug: Tb3-b4 Zeit: 0:00.000 minKnoten: 2.736 TB: 16
-----
1 0:00 0.00 107...Tb4+ (0)
1 0:00 0.00 107...Txg3+ (0)
Bester Zug: Tb3-b4 Zeit: 0:00.000 min

and the game:

[pgn]
[Event "40 Züge in 8 min"]
[Site "smarthink2, AGANDO_HP_PC"]
[Date "2014.07.23"]
[Round "49.1"]
[White "Booot 5.2.0 x64"]
[Black "SmarThink 1.70 x64"]
[Result "1-0"]

1. e4 e5 2. Bc4 Nf6 3. d3 Nc6 4. Nf3 Bc5 5. O-O d6 6. c3 a6
7. Bb3 Ba7 8. Nbd2 h6 9. Re1 O-O 10. h3 Be6 11. Nf1
{+0.11/20 17s} Re8 {+0.16/15 13s} 12. Bxe6 {+0.11/2123s}
Rxe6 {+0.18/18 13s} 13. Be3 {+0.11/21 17s} Qd7 {+0.19/18
12s} 14. Qb3 {+0.11/21 21s (Lxa7)} Ree8 {+0.20/17 11s}
15. Rad1 {+0.12/20 12s (Lxa7)} b5 {+0.15/16 11s} 16. Qc2
{+0.22/21 21s (Lxa7)} a5 {+0.13/16 11s} 17. Bxa7 {+0.22/21
17s (Sg3)} Rxa7 {+0.14/17 12s} 18. Ng3 {+0.29/22 43s (d4)}
Qe6 {+0.11/1712s} 19. d4 {+0.24/22 12s} Ne7 {+0.36/16 30s}
20. a4 {+0.29/21 15s (dxe5)} Rb7 {+0.41/17 11s} 21. axb5
{+0.29/21 9s} Rxb5 {+0.31/18 4s} 22. Ra1 {+0.42/2213s} Reb8
{+0.51/19 17s} 23. Reb1 {+0.35/23 11s} Qc4 {+0.51/18 12s}
24. Ra3 {+0.35/22 14s} Qc6 {+0.33/18 12s} 25. b3 {+0.56/22
9s} Qb6 {+0.42/18 12s} 26. dxe5 {+0.56/20 14s (Se2)} dxe5
{+0.40/17 4s} 27. Nd2 {+0.56/20 14s} Qc5 {+0.42/1813s}
28. Ra4 {+0.49/21 11s} Rf8 {+0.56/18 23s} 29. Nc4 {+0.59/22
14s (Tba1)} Nc6 {+0.60/19 14s} 30. Nf5 {+0.60/22 14s} Rfb8
{+0.61/19 14s} 31. Nce3 {+0.61/21 37s (Ta3)} Qf8 {+0.46/18
14s} 32. Rc4 {+0.56/21 12s} R5b6 {+0.61/18 15s} 33. Nd5
{+0.57/21 20s} Nxd5 {+0.34/19 15s} 34. exd5 {+0.57/22 11s}
Ne7 {+0.48/2116s} 35. Ne3 {+0.51/22 11s} Qd8 {+0.44/21 19s}
36. Rc5 {+0.61/22 11s (Ta4)} Ra8 {+0.63/18 29s} 37. Nc4
{+0.64/19 8s} Rba6 {+0.60/19 17s} 38. Qe4 {+0.64/189s} f6
{+0.79/19 31s} 39. Ra1 {+0.84/17 5s (Dg4)} a4 {+0.73/19
17s} 40. b4 {+0.94/21 7s} Kf8 {+0.68/19 14s} 41. Qg4
{+0.97/20 24s (Sa5)} Rb8 {+0.78/16 11s} 42. Qf3 {+1.06/23
19s} Kg8 {+0.78/18 11s} 43. Qe2 {+1.06/22 22s (Dd1)} Rba8
{+0.90/17 11s} 44. Qc2 {+0.93/21 21s} Kf8 {+0.91/18 10s}
45. Qh7 {+0.93/20 48s (Td1)} Kf7 {+0.67/17 10s} 46. Qd3
{+0.93/21 17s} Kf8 {+0.96/19 26s } 47. Na5 {+1.09/20 14s
(Df3)} Qd7 {+0.82/17 9s} 48. Qc4 {+0.81/21 10s (De4)} R6a7
{+0.97/17 9s} 49. Qa2 {+0.83/20 15s (Dg4)} Rd8 {+0.79/18
10s} 50. Qc2 {+0.91/2115s} a3 {+0.86/19 12s} 51. Qb3
{+0.96/21 8s} Qf5 {+0.86/20 8s} 52. c4 {+0.88/228s} Qe4
{+0.91/19 14s} 53. Qxa3 {+0.99/23 11s} Nf5 {+0.91/19 8s}
54. Qb2 {+0.91/22 13s (Dc1)} Rda8 {+0.86/19 8s} 55. Qb1
{+0.93/22 7s} Qf4 {+0.83/19 9s} 56. Qc1 {+0.89/21 11s
(De1)} Qe4 {+0.82/18 9s} 57. Qe1 {+0.94/21 8s (Db1)} Qf4
{+0.86/19 9s} 58. Rb1 {+0.86/21 10s (Dc3)} Nd4 {+0.83/18
9s} 59. Rd1 {+0.87/227s} Nc2 {+0.85/19 14s} 60. Qc3
{+0.90/23 9s} Qe4 {+0.85/19 9s} 61. Qd2 {+0.86/22 9s} Nd4
{+0.85/17 9s} 62. Re1 {+0.82/22 6s (Kh1)} Qg6 {+0.81/17 9s}
63. Qd1 {+0.82/22 9s} Qf5 {+0.83/20 10s} 64. Rf1 {+0.82/21
9s} Qe4 {+0.83/20 10s} 65. Qd2 {+0.78/21 9s} Qf5 {+1.03/19
17s} 66. f4 {+1.22/21 5s} Qe4 {+1.18/1810s} 67. Kh2
{+1.22/20 9s} Nf5 {+1.26/17 10s} 68. Rf2 {+1.26/20 6s} Re8
{+1.44/18 17s} 69. Nc6 {+1.85/19 6s (fxe5)} Ra1 {+1.64/18
10s} 70. fxe5 {+1.95/206s} Qb1 {+2.18/18 26s} 71. Rxf5
{+1.91/21 6s} Qxf5 {+2.13/19 3s} 72. d6 {+1.91/22 11s} cxd6
{+2.13/20 10s} 73. exd6 {+2.03/22 7s} Qd7 {+2.13/20 11s}
74. Ne7 {+2.03/24 11s} Rxe7 {+2.32/21 15s} 75. dxe7+
{+2.03/23 11s} Kxe7 {+2.32/2111s} 76. Qe3+ {+1.99/24 8s
(Df4)} Kf7 {+2.00/18 12s} 77. Qg3 {+2.08/25 12s} Qd4
{+2.01/21 13s} 78. Rc7+ {+2.08/24 9s} Ke6 {+2.01/21 14s}
79. Qg4+ {+2.08/245s} Qxg4 {+2.08/22 16s} 80. hxg4
{+2.01/27 8s} Rb1 {+2.09/23 14s} 81. b5 {+2.18/29 17s} Rb4
{+2.09/24 11s} 82. Kh3 {+2.18/31 23s (Kg3)} Kd6 {+2.09/22
13s} 83. Rxg7 {+2.28/33 15s} Rxc4 {+2.09/22 11s} 84. Rg6
{+2.40/32 16s (b6)} Rc3+ {+2.29/19 10s} 85. g3 {+2.40/32
17s} Ke7 {+2.75/22 26s} 86. Rxh6 {+2.40/35 16s} Kf7
{+3.22/20 24s} 87. Rh8 {+2.70/33 14s} Rb3 {+3.65/23 23s}
88. Rb8 {+3.08/3812s} Rb2 {+3.65/24 8s} 89. b6 {+2.95/40
16s} Kg7 {+3.65/26 10s} 90. b7 {+2.95/38 14s} Kh7 {+3.65/27
9s} 91. g5 {+2.95/40 16s} fxg5 {+3.65/29 9s} 92. Kg4
{+2.95/40 15s} Rb5 {+3.65/29 13s} 93. Kf3 {+2.01/39 14s}
Rb4 {+3.65/31 12s} 94. Ke2 {+2.95/36 14s (g4)} g4 {+3.65/30
8s} 95. Kd2 {+2.95/39 11s} Rb3 {+3.65/308s} 96. Kc2
{+2.40/43 10s} Rb6 {+3.65/31 8s} 97. Kc3 {+2.40/42 12s
(Kd3)} Kg7 {+3.65/29 8s} 98. Kc4 {+2.01/44 9s} Rb2
{+3.65/30 8s} 99. Kc5 {+2.01/45 6s (Kd5)} Rb3 {+3.65/30 8s}
100. Kd5 {+2.01/43 10s (Kc6)} Rb6 {+3.65/318s} 101. Kd4
{+0.84/41 8s (Ke5)} Rb2 {+3.65/31 8s} 102. Kc4 {+0.84/42 7s
(Ke5)} Rb1 {+3.65/27 8s} 103. Kd3 {+0.69/44 29s (Kd5)} Rb4
{+1.23/21 9s} 104. Ke3 {+0.69/43 6s (Kc3)} Kh7 {+0.01/19
9s} 105. Ra8 {+0.69/42 6s (Kd3)} Rxb7 {0.00/18 9s} 106. Kf4
{+0.69/42 7s} Rb3 {0.00/21 9s} 107. Kxg4 {+0.84/38 6s}
Rxg3+ {0.00/199 0s} 108. Kxg3 {+M11/30 6s} Kg7 {+M10/23 0s}
109. Re8 {+M10/ 38 6s} Kf7 {+M9/21 0s} 110. Re3 {+M9/46 7s}
Kg7 {+M8/19 0s} 111. Kg4 {+M8/53 7s } Kf6 {+M7/17 0s}
112. Re4 {+M7/64 5s} Kg7 {+M6/15 0s} 113. Kg5 {+M6/64 1s}
Kf7 {+M5/13 0s} 114. Re5 {+M5/64 0s} Kg7 {+M4/11 0s}
115. Rf5 {+M4/64 0s} Kh8 {+M3/ 9 0s} 116. Kh6 {+M3/64 0s}
Kg8 {+M3/11 0s} 117. Kg6 {+M2/64 0s} Kh8 {+M3/11 0s}
118. Rf8# {+M1/64 0s} 1-0 [/pgn]

[carldaman]

Hi,
just watched a game where SmarThink failed in a 5men endgame although all 5 men tbs are loaded: move 107 ...Rxg3 ??
Werner

[d]R7/7k/8/8/6K1/1r4P1/8/8 b - -
Engine: SmarThink 1.70 x64 (256 MB)
von Sergei S. Markoff
100/100 0:00 0.00 107...Tb4+ (2.736) TB:16
Bester Zug: Tb3-b4 Zeit: 0:00.000 minKnoten: 2.736 TB: 16
-----
1 0:00 0.00 107...Tb4+ (0)
1 0:00 0.00 107...Txg3+ (0)
Bester Zug: Tb3-b4 Zeit: 0:00.000 min

Interesting that you posted this, since I noticed a significant drop in nps when probing Nalimov TBs. It only uses 10% of regular nps in an endgame with few pieces . My recommendation is to ditch Nalimov use for now, and let the engine play on its own, until Sergei can fix the problems.

Edit: You can create a smarthink.ini file like the one below to switch off Nalimov use.

Code: Select all

;  Setup file for the SmarThink

;
;  TranspositionSize - the size of transposition table
;                                                  
;  Value                   Size (bytes)
;
;  15                      655360
;  16                      1310720
;  17                      2621440
;  18                      5242880
;  19 *                    10485760
;  20                      20971520
;  21                      41943040
;  22                      83886080
;  23                      167772160
;  24                      335544320
;
;  * - default
;

TranspositionSize=21

;  
;  PawnHashSize - the size of pawn hash
;
;  Value                   Size (bytes)
;
;  14                      360448
;  15                      720896
;  16                      1441792
;  17                      2883534
;  18 *                    5767168
;  19                      11534336
;  20                      23068672
;  21                      46137344
;  22                      92274688
;  23                      184549376
;
;  * - default
;

PawnHashSize=18

;
; Aggressivity (%) - causes SmarThink to attack enemy king stronger
; (100 by default)
;
; 0 <= Aggressivity <= 300
;

Aggressivity=100

;
; Cowardice (%) - causes SmarThink to make care of own king
; (100 by default)
;
; 0 <= Cowardice <= 300
;

Cowardice=100

; In learning mode SmarThink saves a costly miscalculations in special
; file for using this info in next games

Learning=off

; When UseLearningData on, SmarThink uses data from learning file

UseLearningData=off

; If LogLearningIntoFile is on than SmarThink shows all learning positions
; in log-file (EPD-notation and several additional debug info)

LogLearningIntoFile=off

; If evaluation is less than *ResignValue* during *ResignTimes* moves
; than SmarThink resigns

ResignValue=800
ResignTimes=5

; In book learning mode SmarThink updates weights in openings book

BookLearning=off

; DISCOVER = play unchecked variations first to collect a weights,
; KILL = try to find best lines to overthrow opponent

BookLearningMode=KILL

; VarietyOfPlay (%) - play book move only if its weight is not less
; than weight of best move by *VarietyOfPlay* %.

VarietyOfPlay=15

; Update weight if abs(evaluation)>*BookLearningBound*

BookLearningBound=260

; No comments

UseBookWeights=on

; Support pondering or no?

Pondering=on

; Use endgame tablebases or no

UseEGTB=off

; Path to tablebase files

EGTBpath=c:\Tablebases

Regards,
CL