Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Test Positions / Tough Tactical Test 1
- - By dorszcz (**) Date 2020-05-06 15:44
TTT1 is a collection of 100 various hard positions for testing engines. Thanks to Dann Corbit who significantly contributed to the creation of this test.

Conditions:
Time limit: 15 seconds per position
CPU: i5-3570 (MP engines: 4x3600 MHz, SP engines: 1x3800MHz)
Tablebases: 6-pieces Syzygy, 5-pieces Gaviota, 5-pieces (and ~80 GB of most popular 6-pieces) Nalimov, all on SSD
Strongest version of engine is used: 64 bit, POPCNT, SSE4.2 etc.
256 MB hash, no books, no learning, 50 move rule is enabled

Results:
76 out of 100 = 76%  -  Bluefish XI-LP FD (Tactical=2; defensive=off)
54 out of 100 = 54%  -  Bluefish XI-LP FD
35 out of 100 = 35%  -  Black Diamond XR7
30 out of 100 = 30%  -  Crystal 260819 64
29 out of 100 = 29%  -  Eman 4.00
28 out of 100 = 28%  -  CorChess 3.1 260819
28 out of 100 = 28%  -  Bluefish FD 100919
25 out of 100 = 25%  -  MateFinder 260819
22 out of 100 = 22%  -  SugaR-NN 130819
21 out of 100 = 21%  -  AsmFishWCP_2019-07-23
20 out of 100 = 20%  -  Cfish 240719
19 out of 100 = 19%  -  Honey X5i
18 out of 100 = 18%  -  Crystal-Honey X5i
15 out of 100 = 15%  -  Stockfish 10
9 out of 100 = 9%  -  Spark-1.0
7 out of 100 = 7%  -  Xiphos 0.5.6
5 out of 100 = 5%  -  Andscacs 0.95
5 out of 100 = 5%  -  Critter 1.6a
5 out of 100 = 5%  -  Xiphos 0.6
5 out of 100 = 5%  -  Xiphos 0.5.3
4 out of 100 = 4%  -  Sting 18
4 out of 100 = 4%  -  Sting 14
4 out of 100 = 4%  -  Naum 4.6
4 out of 100 = 4%  -  Komodo 10
3 out of 100 = 3%  -  Nirvanachess 2.4
3 out of 100 = 3%  -  Wasp 3.60
2 out of 100 = 2%  -  Rybka 3 Dynamic
2 out of 100 = 2%  -  Houdini 1.5a
2 out of 100 = 2%  -  Booot 6.3.1
2 out of 100 = 2%  -  Stockfish 5
2 out of 100 = 2%  -  Arasan 21.3
2 out of 100 = 2%  -  Deep Rybka 4
2 out of 100 = 2%  -  IvanHoe 9.46b
1 out of 100 = 1%  -  SmarThink 1.98 (1CPU)
1 out of 100 = 1%  -  Alfil 13.1
1 out of 100 = 1%  -  Equinox 3.30
1 out of 100 = 1%  -  Gull 3
1 out of 100 = 1%  -  Senpai 2.0
1 out of 100 = 1%  -  rofChade 2.2
1 out of 100 = 1%  -  Rybka WinFinder 2.2
1 out of 100 = 1%  -  Nemorino_5.00
1 out of 100 = 1%  -  Wasp 3.75
1 out of 100 = 1%  -  Sting SF 3 VE
1 out of 100 = 1%  -  Fire 2.2+ xTreme GH
0 out of 100 = 0%  -  Spike 1.4
0 out of 100 = 0%  -  RubiChess 1.5
0 out of 100 = 0%  -  Fire 7.1
0 out of 100 = 0%  -  Texel 1.08a11
0 out of 100 = 0%  -  Topple 0.7.3
0 out of 100 = 0%  -  Pedone 1.9
0 out of 100 = 0%  -  Rodent III 0.273
0 out of 100 = 0%  -  Protector 1.9.0
0 out of 100 = 0%  -  Fritz 11 SE (1CPU)
0 out of 100 = 0%  -  Laser 1.7
0 out of 100 = 0%  -  Marvin 3.4.0
0 out of 100 = 0%  -  ICE 3.0 (1CPU)
0 out of 100 = 0%  -  Amoeba 3.0
0 out of 100 = 0%  -  Bobcat 3.25
0 out of 100 = 0%  -  RofChade 2.1
0 out of 100 = 0%  -  Deep iCE 4.0
0 out of 100 = 0%  -  RubiChess 1.6

Download here: http://www.mediafire.com/file/gwkqps6ihwro14g/Tough_Tactical_Test_1_06.05.2020.zip/file
Parent - - By Clementin (*) Date 2020-05-07 07:49
Thank you for building the test, it is really appreciated.

I ran Allie 0.6 + LS 14.2 at 15sec per move on my gtx 1660 ti and it got 58/100. Fat Fritz got 48/100. Lc0 25.1 was inbetween with LS 14.0. Maybe larger nets on better card will do better.
Parent - - By user923005 (****) Date 2020-05-07 08:23
I am curious if the NN programs solved the same problems or different problems
Parent - By MrKris (***) Date 2020-05-08 04:40
From my post below.

15" , R7 2700X 16ths | RTX 2060
Syzygy 6 50mrOn, 2 GBh
49) 1..  10) 1. secs. if found
f3+ 3 ;  Rd7     Black-Diamond-043020  68  (DeepDef,noCnoPro,Tac2)
f3+ 0 ;  Rd7     Lc0.24.1-SVt40-1541   61
f3+ 0 ;  Rd7     Lc0.24.1-Ls14         61
f3+ 0 ;  Rd7  ** Lc0.24.1-SVlgt60-3010 57
f3+ 0 ;  Rd7     Lc0-.23.2K-lg2036     55
f3+ 1 ;  Bxe4 11 Crystal-200420        50
Qxd5  ;  Bxe4  2 Houdini-6.03-Std-tac1 49  (8ths, 8thsTac @ 30")
Qxd5~ ;  Bxe4 11 Honey-043020          47
Qxd5  ;  Rd7   * Stockfish_20050217    43
Qxd5  ;  Bxe4  9 Houdini-6-Std-tac2    39  (8ths, 4thsTac @ 30")
Qxd5  ;  Bxe4  0 Komodo-13.3           38
Qxd5  ;  Bxe4  4 Houdini-6.03-Std      21  (8ths @ 30")
      ;  solved after: * 16", ** 34"
solved after ~ 02:10


All of my NN's and Stockfish (though about by 1 sec.) missed this.
One of the nets solved it in a maybe respectable 34 secs. afterwards.
[Event "#10"]
[Site "Tough Tactical Test 1"]
[Date ""]
[Round ""]
[White "often solved by CPU's"]
[Black ""]
[Result "1-0"]
[BlackElo "2400"]
[WhiteElo "2400"]
[TimeControl "0+15"]
[SetUp "1"]
[FEN "1r1q1rk1/1bR2pp1/1p5p/p2pNP2/Pb1Pn3/1P1B1R1P/1B2Q1P1/6K1 w - - 0 1"]
[Termination "unterminated"]
[PlyCount "33"]
[WhiteType "human"]
[BlackType "human"]

1. Bxe4 (1. Rd7 {??} Qg5 {= or +=} 2. f6 {quickly, Honey 0.00} Nxf6 3. Nxf7
Rxf7 4. Qe6 Nxd7 5. Qxf7+ Kh8 6. Qxd7) 1. .. dxe4 (1. .. Qxc7 2. f6 dxe4 3.
Rg3 {transposes}) 2. Rg3 Qxc7 3. f6 g6 4. Qe3 Kh7 5. Bc1 Qxc1+ 6. Qxc1 Rbc8
7. Qf4 Rc2 8. Ng4 Bd2 9. Re3 Bxe3+ 10. Nxe3 Rc1+ 11. Kh2 Bc8 12. Ng4 Bxg4
13. Qxc1 Be6 14. Qc6 e3 15. Qe4 Bxb3 16. Qxe3 Be6 (16. .. Bxa4 {??} 17.
Qa3) 17. Qa3 {Black's rook can not leave the 8th rank.} 1-0

Only my NN's, Crystal and Black Diamond found the win, the others did not have time to realize the loss.
The NN's all took under 1 sec., about how long a GM (maybe much less?) would take I think
[Event "#49"]
[Site "Tough Tactical Test 1"]
[Date "2020.05.07"]
[Round ""]
[White ""]
[Black "often solved by NN's"]
[Result "0-1"]
[BlackElo "2400"]
[Time "20:12:47"]
[WhiteElo "2400"]
[TimeControl "0+15"]
[SetUp "1"]
[FEN "k4r1r/Pp6/bP1bq1p1/3N2p1/R1pPppB1/4P1PP/5PK1/3QBR2 b - - 0 1"]
[Termination "normal"]
[PlyCount "21"]
[WhiteType "human"]
[BlackType "human"]
[Comment "bm f3+; id \"TTT1.049\";"]

1. ... f3+

( 1. ... Qxd5 {??} 2. Rxa6 {Lc0 t60-3010 1 min. +1.42, Honey 1 min. +2.42}
bxa6 3. Qa4 {etc.} )

2. Kg1 Qxg4 {Lc0 t60-3010 1 min. -5.05 does not say mate but its PV is:} 3. hxg4 Rh3 4. Rxc4 Rfh8 5. Rc8+ Rxc8 6. Nc7+ Bxc7 7. bxc7 Rch8 8.
c8=Q+ Rxc8 9. Qxf3 exf3 10. Bc3 Rch8 11. d5 Rh1# 0-1
Parent - - By Clementin (*) Date 2020-05-09 06:56
It is the same problems for sure. I reran the test after recreating the database in Fritz GUI, but the results are quite high (say for Allie, 58/100) and I find them surprising. I also reran the test in Arena and with ArasanEpdTEster, and Allie got only 46/100 (with the same settings, network etc.). I absolutely don't get it. I checked the test going in Fritz GUI... Alllie just solves the problems in due time. Say position 100. In Fritz GUI, Allie solves the problem:

63: TTT1.100 Line, TTT1
8/5pkp/1p6/p4p2/3P4/B5rP/1R6/7K w - - 0 1


Analysis by Allie v0.6 (6cd06b9):

1.Bd6
  White is clearly better: +- (2.21)   00:00:04  22N
1.Rg2 Rxg2
  White is better: +/- (1.02)  Depth: 1/2   00:00:04  30N
1.Rg2 Rxg2 2.Kxg2
  White is clearly better: +/- (1.38)  Depth: 1/3   00:00:04  48N
1.Rg2 Rxg2 2.Kxg2 Kf6
  White is better: +/- (1.07)  Depth: 2/4   00:00:04  79N
1.Be7 Rxh3+ 2.Kg2 Re3 3.Bd8
  White is better: +/- (1.18)  Depth: 2/5   00:00:04  120N
1.Be7 Rxh3+ 2.Kg2 Re3 3.Bd8 a4
  White is better: +/- (1.14)  Depth: 2/6   00:00:04  131N
1.Be7 Rxh3+ 2.Kg2 Re3 3.Bd8 a4
  White is clearly better: +/- (1.31)  Depth: 3/7   00:00:04  194N
1.Be7 Rxh3+ 2.Kg2 Re3 3.Bd8 a4 4.Rxb6
  White is better: +/- (1.29)  Depth: 3/8   00:00:04  224N
1.Be7 Rxh3+ 2.Kg2 Re3 3.Bd8 a4 4.Rxb6
  White is better: +/- (1.28)  Depth: 4/8   00:00:04  305N
1.Be7 Rxh3+ 2.Kg2 Re3 3.Bd8 a4 4.Rxb6 Re6
  White is clearly better: +/- (1.32)  Depth: 4/9   00:00:04  351N
1.Be7 Rxh3+ 2.Kg1 Re3 3.Rg2+ Kh6 4.Bf8+
  White is clearly better: +/- (1.39)  Depth: 5/10   00:00:04  542N
1.Be7 Rxh3+ 2.Kg1 f4
  White is clearly better: +/- (1.32)  Depth: 5/11   00:00:04  814N
1.Be7 Rxh3+ 2.Kg1 f4 3.Rxb6
  White is clearly better: +/- (1.31)  Depth: 6/11   00:00:04  1024N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Ra6 h5
  White is better: +/- (1.29)  Depth: 6/12   00:00:04  1126N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Ra6 h5 5.Kf2 h4
  White is better: +/- (1.30)  Depth: 6/13   00:00:04  1809N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Ra6 f6 5.Bxf6+
  White is better: +/- (1.30)  Depth: 7/13   00:00:04  1974N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 a4
  White is better: +/- (1.17)  Depth: 7/14   00:00:04  3359N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Ra6 f6 5.Bxf6+ Kg6 6.Be5+
  White is better: +/- (1.14)  Depth: 7/15   00:00:04  3946N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 h5 5.Ra6
  White is better: +/- (1.11)  Depth: 8/15   00:00:04  4410N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 h5 5.Ra6 Kh7
  White is better: +/- (1.07)  Depth: 8/16   00:00:04  4983N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 Rf3+ 5.Kg2
  White is better: +/- (1.05)  Depth: 8/17   00:00:04  9025N
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 h5 5.Ra6 a4 6.Bf6+
  White is better: +/- (1.03)  Depth: 9/17   00:00:05  10kN
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 h5 5.Ra6 a4 6.Bf6+
  White is better: +/- (1.02)  Depth: 9/18   00:00:05  11kN
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Bd8
  White is better: +/- (1.00)  Depth: 9/19   00:00:05  12kN
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Bd8
  White is better: +/- (1.00)  Depth: 9/20   00:00:05  13kN
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Bd8
  White is better: +/- (1.00)  Depth: 9/21   00:00:05  14kN
1.Be7 Rxh3+ 2.Kg1 Rd3 3.Rxb6 f4 4.Kf1 h5 5.Ra6 a4 6.Bf6+
  White is better: +/- (0.96)  Depth: 10/21   00:00:05  21kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 Kg6 5.Rxf6+ Kg5 6.Kf2 f4
  White is better: +/- (0.95)  Depth: 10/21   00:00:05  22kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 Kh6 4.Kf1 Kg5 5.Bc5 f4 6.Ke2
  White is better: +/- (1.01)  Depth: 10/22   00:00:05  28kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 Kh6 4.Bc5+ Kg5 5.Kf2 Kf4 6.Rb7
  White is better: +/- (1.03)  Depth: 10/23   00:00:06  29kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 Kg6 4.Bc5+ f6 5.Kf1 f4 6.Ke2 Re3+ 7.Kd2 h5 8.Rb8
  White is better: +/- (1.05)  Depth: 10/24   00:00:06  32kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 Kg6 5.Rxf6+ Kg5 6.Kf2 f4 7.Rf7+ Kg4 8.Bc5 h5 9.Ke2 Re3+ 10.Kd2 h4
  White is better: +/- (1.03)  Depth: 11/24   00:00:06  33kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 Kg6 5.Rxf6+ Kg5 6.Kf2 f4 7.d5 Kg4 8.d6 a4 9.Ke2 Rd5 10.Rf8 a3
  White is better: +/- (1.03)  Depth: 11/25   00:00:06  37kN
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 Kg6 4.Bc5+ f6 5.Kf1 f4 6.Ke2 Re3+ 7.Kd2 h5 8.Rb8
  White is better: +/- (1.05)  Depth: 11/26   00:00:06  39kN, tb=1
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 Kg6 4.Kf2 f4 5.Bxf4+ Kf5 6.Be3 Ke4 7.Rf6 Rb3
  White is better: +/- (1.08)  Depth: 12/27   00:00:07  51kN, tb=2
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 a4 5.Ra6 f4 6.Bxf6+ Kf7 7.Be5 a3 8.Kf2 f3 9.Rf6+ Ke7 10.Rh6
  White is better: +/- (1.10)  Depth: 13/27   00:00:08  76kN, tb=3
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 a4 5.Bxf6+ Kg6 6.Ra6 f4 7.Be5+ Kf5 8.Rxa4 Ke4 9.Ra1 h5 10.Re1+ Re3 11.Rf1 Re2+ 12.Rf2 Rxf2+
  White is better: +/- (1.06)  Depth: 13/28   00:00:09  102kN, tb=6
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 a4 5.Ra6 f4 6.Bxf6+ Kg6 7.Be5+ Kf5 8.Rxa4 Ke4 9.Ra1 h5 10.Re1+ Re3 11.Rf1 Re2+ 12.Rf2 Rxf2+ 13.Kxf2 h4
  White is better: +/- (1.03)  Depth: 14/28   00:00:10  134kN, tb=9
1.Rg2 Rxg2 2.Kxg2 Kf6 3.Bd6 Ke6 4.Bc7 Kd5 5.Bxb6 a4 6.Bc5 Ke4 7.h4 f4 8.Kf2 h5
  White is better: +/- (1.01)  Depth: 14/28   00:00:11  163kN, tb=9
1.Bd6 Rxh3+ 2.Kg2 Rd3 3.Rxb6 f6 4.Be7 a4 5.Ra6 f4 6.Bxf6+ Kg6 7.Be5+ Kf5 8.Rxa4 Ke4 9.Ra1 h5 10.Re1+ Re3 11.Rg1 Rg3+ 12.Kh2
  White is better: +/- (1.00)  Depth: 14/28   00:00:11  163kN, tb=9
1.Rg2 f4 2.Be7 f5 3.Bh4 Rxg2 4.Kxg2 b5 5.d5 Kf7 6.Bd8 a4 7.Ba5 a3 8.Kf3 a2 9.Bc3 Ke7
  White is better: +/- (1.03)  Depth: 14/28   00:00:11  172kN, tb=9
1.Rg2 f4 2.Be7 Kg6 3.Rd2 Rxh3+ 4.Kg1 f6 5.d5 Kf7 6.d6 Rg3+ 7.Kf2 Rg8 8.Re2 Ke8 9.Bxf6+ Kd7 10.Be5 a4 11.Kf3 a3 12.Rh2 h5 13.Rxh5 a2 14.Rh7+ Ke6
  White is better: +/- (1.22)  Depth: 14/29   00:00:14  215kN, tb=9
1.Rg2 f4 2.Rxg3+ fxg3 3.d5 Kf6 4.Kg2 b5 5.Bc5 Ke5 6.d6 Ke6 7.Kxg3 f5 8.Bb6 b4 9.Bxa5 b3 10.Bc3 Kxd6 11.Kf4
  White is clearly better: +- (1.70)  Depth: 14/29   00:00:16  302kN, tb=9


While in ArasanEpdTester I get from the logfile:
Pos 100
8/5pkp/1p6/p4p2/3P4/B5rP/1R6/7K w - - bm Rg2; id "TTT1.100";


2020-05-09T01:29:17.102000 >> isready
2020-05-09T01:29:17.114000 << readyok
2020-05-09T01:29:17.114000 >> ucinewgame
2020-05-09T01:29:17.114000 >> position fen
8/5pkp/1p6/p4p2/3P4/B5rP/1R6/7K w - - 0 1

2020-05-09T01:29:17.114000 >> go movetime 15000
2020-05-09T01:29:18.100000 << info depth 0 seldepth 1 nodes 22 nps 880 score cp 554 time 25 hashfull 0 tbhits 0 pv a3d6
2020-05-09T01:29:18.108000 << info depth 1 seldepth 2 nodes 31 nps 939 score cp 432 time 33 hashfull 0 tbhits 0 pv b2g2
2020-05-09T01:29:18.122000 << info depth 1 seldepth 3 nodes 51 nps 1085 score cp 155 time 47 hashfull 0 tbhits 0 pv b2g2 g3g2 h1g2
2020-05-09T01:29:18.130000 << info depth 1 seldepth 4 nodes 70 nps 1273 score cp 127 time 55 hashfull 0 tbhits 0 pv b2g2 g3g2 h1g2 g7f6
2020-05-09T01:29:18.138000 << info depth 2 seldepth 5 nodes 98 nps 1556 score cp 125 time 63 hashfull 0 tbhits 0 pv b2g2 g3g2 h1g2 g7f6 d4d5
2020-05-09T01:29:18.147000 << info depth 2 seldepth 6 nodes 132 nps 1833 score cp 101 time 72 hashfull 0 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6
2020-05-09T01:29:18.155000 << info depth 3 seldepth 7 nodes 161 nps 2013 score cp 95 time 80 hashfull 0 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 h7h5
2020-05-09T01:29:18.167000 << info depth 3 seldepth 8 nodes 213 nps 2315 score cp 123 time 92 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g2 h3e3 e7d8
2020-05-09T01:29:18.201000 << info depth 4 seldepth 8 nodes 323 nps 2563 score cp 151 time 126 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g2 h3e3 e7d6 e3d3 b2b6 f7f6
2020-05-09T01:29:18.211000 << info depth 4 seldepth 9 nodes 366 nps 2691 score cp 146 time 136 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g2 h3e3 e7d8 a5a4 b2b6 e3e6
2020-05-09T01:29:18.222000 << info depth 4 seldepth 10 nodes 441 nps 3000 score cp 140 time 147 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3e3 b2g2 g7h6
2020-05-09T01:29:18.249000 << info depth 5 seldepth 11 nodes 654 nps 3759 score cp 134 time 174 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 a5a4
2020-05-09T01:29:18.280000 << info depth 5 seldepth 12 nodes 892 nps 4351 score cp 129 time 205 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 f5f4 b2b6
2020-05-09T01:29:18.304000 << info depth 6 seldepth 12 nodes 1092 nps 4769 score cp 136 time 229 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3d3 b2b6 f5f4 b6a6 a5a4
2020-05-09T01:29:18.384000 << info depth 6 seldepth 13 nodes 1775 nps 5744 score cp 138 time 309 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3d3 b2b6 f5f4 b6a6 h7h5 g1f2 g7h7
2020-05-09T01:29:18.447000 << info depth 7 seldepth 13 nodes 2405 nps 6465 score cp 128 time 372 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3d3 b2b6 f5f4 b6a6 h7h5 g1f2 h5h4
2020-05-09T01:29:18.496000 << info depth 7 seldepth 14 nodes 3066 nps 7283 score cp 117 time 421 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3d3 b2b6 f5f4 b6a6 f7f6 e7f6 g7g6
2020-05-09T01:29:18.564000 << info depth 8 seldepth 15 nodes 3933 nps 8043 score cp 109 time 489 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3d3 b2b6 f5f4 g1f1 a5a4
2020-05-09T01:29:18.618000 << info depth 8 seldepth 16 nodes 4738 nps 8726 score cp 108 time 543 hashfull 0 tbhits 0 pv a3e7 g3h3 h1g1 h3d3 b2b6 f5f4 g1f1 h7h5 e7f6
2020-05-09T01:29:19.039000 << info depth 8 seldepth 16 nodes 11260 nps 11680 score cp 99 time 964 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7g6 d6f4 f7f6
2020-05-09T01:29:19.225000 << info depth 9 seldepth 16 nodes 12913 nps 11229 score cp 105 time 1150 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7h6 g2f2 h6g5 d6c5
2020-05-09T01:29:19.255000 << info depth 9 seldepth 17 nodes 13434 nps 11385 score cp 101 time 1180 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7h6 g2f1
2020-05-09T01:29:19.296000 << info depth 9 seldepth 18 nodes 14032 nps 11492 score cp 99 time 1221 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7h6 d6e5 h6g5 g2f2 a5a4
2020-05-09T01:29:19.418000 << info depth 9 seldepth 19 nodes 15783 nps 11752 score cp 101 time 1343 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7h6 g2f1 h6g5 d6c5 g5f4
2020-05-09T01:29:19.501000 << info depth 9 seldepth 20 nodes 17291 nps 12126 score cp 103 time 1426 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7g6 d6e5 g6g5 b6f6 a5a4 f6f7 f5f4 g2f2
2020-05-09T01:29:19.596000 << info depth 9 seldepth 21 nodes 18790 nps 12354 score cp 103 time 1521 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7g6 g2f1 g6g5 d6c5 g5f4
2020-05-09T01:29:19.631000 << info depth 9 seldepth 22 nodes 19528 nps 12550 score cp 102 time 1556 hashfull 1 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7g6 g2f1 g6g5 d6c5 g5f4
2020-05-09T01:29:19.809000 << info depth 10 seldepth 22 nodes 22745 nps 13117 score cp 95 time 1734 hashfull 2 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 g7g6 e7f6 a5a4 b6a6 a4a3 f6e7 g6h5 e7c5 h5g4 a6a3
2020-05-09T01:29:19.954000 << info depth 10 seldepth 23 nodes 25482 nps 13561 score cp 94 time 1879 hashfull 2 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 g7g6 e7f6 a5a4 b6a6 a4a3 f6e7 g6h5 e7c5 h5g4 a6a3 d3d2 g2f1
2020-05-09T01:29:20.006000 << info depth 10 seldepth 24 nodes 26440 nps 13685 score cp 95 time 1932 hashfull 2 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 g7g6 e7f6 a5a4 b6a6 a4a3 f6e7 g6h5 e7c5 a3a2 a6a2
2020-05-09T01:29:20.074000 << info depth 10 seldepth 25 nodes 27469 nps 13741 score cp 97 time 1999 hashfull 2 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 g7g6 e7f6 a5a4 b6a6 a4a3 f6e7 g6h5 e7c5 h5g4 a6a3 d3d2 g2f1 f5f4 f1e1 d2c2
2020-05-09T01:29:20.135000 << info depth 10 seldepth 26 nodes 28686 nps 13925 score cp 99 time 2060 hashfull 2 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 g7g6 e7f6 a5a4 b6a6 f5f4 f6e5 g6f5 a6a4 d3d2 g2h3
2020-05-09T01:29:20.332000 << info depth 11 seldepth 26 nodes 33090 nps 14655 score cp 102 time 2258 hashfull 2 tbhits 0 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 g7g6 e7f6 a5a4 b6a6 f5f4 f6e5 g6f5 a6a4 d3d2 g2h3
2020-05-09T01:29:20.744000 << info depth 11 seldepth 27 nodes 43487 nps 16293 score cp 105 time 2669 hashfull 3 tbhits 1 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 f5f4 e7f6 g7g6 f6e5 g6g5 b6f6 a5a4 f6f4 a4a3 f4f8 h7h5 g2f2 h5h4 f2e2 d3b3
2020-05-09T01:29:20.883000 << info depth 12 seldepth 28 nodes 46837 nps 16680 score cp 106 time 2808 hashfull 3 tbhits 2 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 f5f4 e7f6 g7g6 f6e5 g6g5 b6a6 h7h5 a6a5 g5g4 a5a7 d3d2 g2g1 h5h4
2020-05-09T01:29:22.505000 << info depth 13 seldepth 28 nodes 76033 nps 17159 score cp 111 time 4431 hashfull 5 tbhits 3 pv a3d6 g3h3 h1g2 h3d3 b2b6 g7g6 g2f2 g6g5 d6c5 g5f4 b6b7 f4e4 b7e7 e4d5
2020-05-09T01:29:24.104000 << info depth 14 seldepth 28 nodes 124693 nps 20682 score cp 108 time 6029 hashfull 8 tbhits 5 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 g7g6 e7f6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1f1 f4f3 g2g3 e3e2 f1f3
2020-05-09T01:29:24.761000 << info depth 14 seldepth 29 nodes 146886 nps 21969 score cp 104 time 6686 hashfull 9 tbhits 6 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 g7g6 e7f6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1d1 e3c3
2020-05-09T01:29:24.826000 << info depth 14 seldepth 30 nodes 148639 nps 22014 score cp 104 time 6752 hashfull 9 tbhits 6 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 g7g6 e7f6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1d1 e3c3
2020-05-09T01:29:24.927000 << info depth 14 seldepth 31 nodes 151060 nps 22046 score cp 105 time 6852 hashfull 9 tbhits 7 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 g7g6 e7f6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1d1 e3c3
2020-05-09T01:29:25.218000 << info depth 14 seldepth 32 nodes 158976 nps 22253 score cp 106 time 7144 hashfull 9 tbhits 10 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 e7f6 g7g6 b6a6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1h1 f4f3 g2g3 e3e2 h1h4 e4e3 e5f4 e3d4
2020-05-09T01:29:25.336000 << info depth 14 seldepth 33 nodes 162108 nps 22326 score cp 107 time 7261 hashfull 10 tbhits 16 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 e7f6 g7g6 b6a6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1h1 f4f3 g2g3 e3e2 h1h4 e4e3 e5f4 e3d4 g3f3 e2e6
2020-05-09T01:29:26.217000 << info depth 15 seldepth 33 nodes 191632 nps 23536 score cp 101 time 8142 hashfull 11 tbhits 21 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 g7g6 e7f6 f5f4 f6e5 g6f5 a6a4 f5e4 a4a1 h7h5 a1e1 d3e3 e1h1 e3g3 g2h2 g3d3 h1f1 d3d2
2020-05-09T01:29:28.745000 << info depth 15 seldepth 33 nodes 285464 nps 26754 score cp 97 time 10670 hashfull 16 tbhits 30 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 f5f4 e7f6 g7f7 f6e5 a4a3 g2f2 f4f3 a6h6 f7e7 f2g3 e7d7 g3f2 d7e7 h6h7
2020-05-09T01:29:31.268000 << info depth 15 seldepth 33 nodes 368362 nps 27921 score cp 93 time 13193 hashfull 21 tbhits 48 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 f5f4 e7f6 g7f7 g2f2 a4a3 f6e5 f4f3 a6h6 f7e7 f2g3 e7d7 g3f2 d7e7 h6f6
2020-05-09T01:29:31.597000 << info depth 15 seldepth 34 nodes 379966 nps 28100 score cp 92 time 13522 hashfull 22 tbhits 50 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 f5f4 e7f6 g7f7 f6e5 a4a3 a6h6 f7e7 g2f2 f4f3 f2g3 e7d7 e5h8
2020-05-09T01:29:32.343000 << info depth 15 seldepth 35 nodes 408802 nps 28652 score cp 92 time 14268 hashfull 23 tbhits 57 pv a3d6 g3h3 h1g2 h3d3 b2b6 f7f6 d6e7 a5a4 b6a6 f5f4 e7f6 g7f7 f6e5 a4a3 a6f6 f7e8 f6h6 d3d1 h6a6 d1d3 a6h6 e8d7 g2f2 f4f3 f2g3 a3a2 h6a6
2020-05-09T01:29:32.775000 << bestmove a3d6 ponder g3h3
epd bm: Rg2
Engine bestmove in san: Bd6
Engine bestmove does not match the epd bm??
Total position to be evaluated: 100
Correct: 46, Evaluated: 100, CorrectRate: 46.0%

Done!! TotalPos: 100, Correct: 46 (46.0%)


Maybe i'm doing something stupid, or there might be some kind of communication problems between Allie and some utilities or GUI. In any case, for once, the real problem seems to be the low results in Arena and ArasanEpdTester rather than the high one in Fritz GUI (i'm used to see the opposite).
Parent - By dorszcz (**) Date 2020-05-09 10:09
There is some random factor in searching. You can test several times the same engine with the same settings in the same GUI and get different results.
It might be a coincidence. When big difference in results repeats in several runs it might be a problem somewhere.
Parent - By Clementin (*) Date 2020-06-12 19:37
Allie 0.7dev + LS 15.0 is a beast! On my GTX 1660 Ti it got 67/100 at 17sec per position!

Allie 0.6 + LS 15.0 got 61/100.
Lc0 24.1 + LS 15.0, with MultiPv=2, DirichletNoise=true, Cpuct=5 got 59/100.

Allie 0.7dev + LS 15.0
Solving: C:\Users\redoi\OneDrive\Documents\ChessBase\EPD TEST\realTTT1.cbh
Maximum time = 17s.

1. TTT1.001,  TTT1   > 17s.
2. TTT1.002,  TTT1   > 17s.
3. TTT1.003,  TTT1   Solved in 1.89s/6; Solved: 1
4. TTT1.004,  TTT1   > 17s.
5. TTT1.005,  TTT1   Solved in 1.17s/1; Solved: 2
6. TTT1.006,  TTT1   Solved in 2.74s/8; Solved: 3
7. TTT1.007,  TTT1   > 17s.
8. TTT1.008,  TTT1   Solved in 1.27s/4; Solved: 4
9. TTT1.009,  TTT1   > 17s.
10. TTT1.010,  TTT1   > 17s.
11. TTT1.011,  TTT1   Solved in 12.74s/10; Solved: 5
12. TTT1.012,  TTT1   Solved in 1.09s/1; Solved: 6
13. TTT1.013,  TTT1   > 17s.
14. TTT1.014,  TTT1   Solved in 1.94s/10; Solved: 7
15. TTT1.015,  TTT1   > 17s.
16. TTT1.016,  TTT1   > 17s.
17. TTT1.017,  TTT1   Solved in 1.14s/1; Solved: 8
18. TTT1.018,  TTT1   > 17s.
19. TTT1.019,  TTT1   Solved in 1.16s/2; Solved: 9
20. TTT1.020,  TTT1   Solved in 1.28s/5; Solved: 10
21. TTT1.021,  TTT1   Solved in 1.14s/1; Solved: 11
22. TTT1.022,  TTT1   > 17s.
23. TTT1.023,  TTT1   > 17s.
24. TTT1.024,  TTT1   Solved in 5.67s/12; Solved: 12
25. TTT1.025,  TTT1   Solved in 1.61s/6; Solved: 13
26. TTT1.026,  TTT1   Solved in 15.44s/18; Solved: 14
27. TTT1.027,  TTT1   Solved in 1.17s/2; Solved: 15
28. TTT1.028,  TTT1   Solved in 1.11s/1; Solved: 16
29. TTT1.029,  TTT1   Solved in 4.94s/12; Solved: 17
30. TTT1.030,  TTT1   > 17s.
31. TTT1.031,  TTT1   > 17s.
32. TTT1.032,  TTT1   Solved in 1.41s/5; Solved: 18
33. TTT1.033,  TTT1   Solved in 16.98s/18; Solved: 19
34. TTT1.034,  TTT1   > 17s.
35. TTT1.035,  TTT1   Solved in 1.88s/8; Solved: 20
36. TTT1.036,  TTT1   Solved in 5.33s/15; Solved: 21
37. TTT1.037,  TTT1   Solved in 1.25s/4; Solved: 22
38. TTT1.038,  TTT1   Solved in 2.56s/10; Solved: 23
39. TTT1.039,  TTT1   Solved in 7.11s/11; Solved: 24
40. TTT1.040,  TTT1   Solved in 1.13s/1; Solved: 25
41. TTT1.041,  TTT1   > 17s.
42. TTT1.042,  TTT1   > 17s.
43. TTT1.043,  TTT1   > 17s.
44. TTT1.044,  TTT1   Solved in 2.58s/8; Solved: 26
45. TTT1.045,  TTT1   Solved in 1.20s/1; Solved: 27
46. TTT1.046,  TTT1   Solved in 5.28s/10; Solved: 28
47. TTT1.047,  TTT1   > 17s.
48. TTT1.048,  TTT1   Solved in 1.20s/4; Solved: 29
49. TTT1.049,  TTT1   Solved in 1.11s/1; Solved: 30
50. TTT1.050,  TTT1   Solved in 2.22s/7; Solved: 31
51. TTT1.051,  TTT1   > 17s.
52. TTT1.052,  TTT1   > 17s.
53. TTT1.053,  TTT1   > 17s.
54. TTT1.054,  TTT1   > 17s.
55. TTT1.055,  TTT1   Solved in 2.33s/8; Solved: 32
56. TTT1.056,  TTT1   Solved in 16.17s/19; Solved: 33
57. TTT1.057,  TTT1   Solved in 1.16s/3; Solved: 34
58. TTT1.058,  TTT1   Solved in 1.22s/4; Solved: 35
59. TTT1.059,  TTT1   Solved in 1.11s/1; Solved: 36
60. TTT1.060,  TTT1   Solved in 2.66s/10; Solved: 37
61. TTT1.061,  TTT1   Solved in 1.19s/6; Solved: 38
62. TTT1.062,  TTT1   > 17s.
63. TTT1.063,  TTT1   Solved in 2.59s/10; Solved: 39
64. TTT1.064,  TTT1   > 17s.
65. TTT1.065,  TTT1   Solved in 4.28s/11; Solved: 40
66. TTT1.066,  TTT1   > 17s.
67. TTT1.067,  TTT1   Solved in 6.47s/10; Solved: 41
68. TTT1.068,  TTT1   Solved in 5.78s/13; Solved: 42
69. TTT1.069,  TTT1   Solved in 3.70s/13; Solved: 43
70. TTT1.070,  TTT1   Solved in 1.11s/1; Solved: 44
71. TTT1.071,  TTT1   Solved in 1.16s/3; Solved: 45
72. TTT1.072,  TTT1   Solved in 1.09s/1; Solved: 46
73. TTT1.073,  TTT1   > 17s.
74. TTT1.074,  TTT1   Solved in 3.47s/11; Solved: 47
75. TTT1.075,  TTT1   Solved in 1.14s/2; Solved: 48
76. TTT1.076,  TTT1   Solved in 1.27s/4; Solved: 49
77. TTT1.077,  TTT1   > 17s.
78. TTT1.078,  TTT1   > 17s.
79. TTT1.079,  TTT1   > 17s.
80. TTT1.080,  TTT1   Solved in 3.42s/6; Solved: 50
81. TTT1.081,  TTT1   > 17s.
82. TTT1.082,  TTT1   Solved in 5.86s/12; Solved: 51
83. TTT1.083,  TTT1   Solved in 12.45s/13; Solved: 52
84. TTT1.084,  TTT1   Solved in 2.80s/16; Solved: 53
85. TTT1.085,  TTT1   Solved in 2.39s/8; Solved: 54
86. TTT1.086,  TTT1   Solved in 2.58s/10; Solved: 55
87. TTT1.087,  TTT1   Solved in 1.19s/4; Solved: 56
88. TTT1.088,  TTT1   Solved in 1.66s/4; Solved: 57
89. TTT1.089,  TTT1   > 17s.
90. TTT1.090,  TTT1   Solved in 1.50s/7; Solved: 58
91. TTT1.091,  TTT1   Solved in 3.83s/7; Solved: 59
92. TTT1.092,  TTT1   Solved in 5.08s/14; Solved: 60
93. TTT1.093,  TTT1   Solved in 2.23s/9; Solved: 61
94. TTT1.094,  TTT1   Solved in 8.78s/16; Solved: 62
95. TTT1.095,  TTT1   Solved in 5.36s/11; Solved: 63
96. TTT1.096,  TTT1   Solved in 6.25s/15; Solved: 64
97. TTT1.097,  TTT1   Solved in 3.11s/12; Solved: 65
98. TTT1.098,  TTT1   Solved in 1.84s/9; Solved: 66
99. TTT1.099,  TTT1   > 17s.
100. TTT1.100,  TTT1   Solved in 7.47s/13; Solved: 67

Result: 67 out of 100 = 67.0%. Average time = 3.64s / 7.74
- - By MrKris (***) Date 2020-05-07 17:35
Thanks again ... great !!
Also there: http://talkchess.com/forum3/viewtopic.php?f=2&t=73852

15" , R7 2700X 16ths | RTX 2060
Syzygy 6 50mrOn, 2 GBh

Black-Diamond-043020  68  (DeepDef,noCnoPro,Tac2)
Lc0.24.1-SVt40-1541   61
Lc0.24.1-Ls14         61
Lc0.24.1-SVlgt60-3010 57
Lc0-.23.2K-lg2036     55
Crystal-200420        50
Houdini-6.03-Std-tac1 49  (8ths, 8thsTac @ 30")
Honey-043020          47
Stockfish_20050217    43
Houdini-6-Std-tac2    39  (8ths, 4thsTac @ 30")
Komodo-13.3           38
Houdini-6.03-Std      21  (8ths @ 30")


By Clementin:

> I ran Allie 0.6 + LS 14.2 at 15sec per move on my gtx 1660 ti and it got 58/100. Fat Fritz got 48/100. Lc0 25.1 was inbetween with LS 14.0. Maybe larger nets on better card will do better.


Excellent GTX 1660 Ti results! The only excuse for mine is the many 0's or few seconds, below - here it looks like the nets generally solved quite quickly or not at all.

By user923005:

> I am curious if the NN programs solved the same problems or different problems


Below 27 of Ls14 61 solved where not solved by Honey,
and 34 of Honey 47 where solved by Ls14,
Sorry, I don't have time to collate more.
Honey-043020 = Y  
  Analyzing engine: Lc0.24.1-Ls14
  1) .. Qe7-e4          Kd7-c8   
  2) .. Ra8-g8          Ra8-g8  * 7 Seconds
  3) .. Rb2-b3          Rb2-b3  * 0 Seconds
  4) Bd3xh7             Qd1-h5   
Y  5) .. Nf6-h5          Nf6-h5  * 0 Seconds
Y  6) Qd2-f4             Qd2-e3   
Y  7) .. Nd6xe4          Ra8-g8   
Y  8) .. h5xg4           h5xg4   * 0 Seconds
  9) .. a4-a3           Bd8-a5   
Y 10) Bd3xe4             Rc7-d7   
 11) .. Bf8-e7          g7xf6    
Y 12) f2-f4              f2-f4   * 0 Seconds
 13) .. Ng3-f1          Nf4xe2   
 14) Ne1-d3, Ne1-f3     Ne1-d3  * 3 Seconds
Y 15) .. Bc8xh3          Ne7-g6   
 16) Bb2-c1             Bb2-c1  * 0 Seconds
Y 17) Nf3-h2             Nf3-h2  * 0 Seconds
Y 18) .. h7-h5           Rc8-f8   
 19) Qa7xd7             Qa7xd7  * 0 Seconds
Y 20) .. Qc7-c8          Qc7-c8  * 0 Seconds
Y 21) .. Rd8-d4          Rd8-d4  * 0 Seconds
 22) .. Bc8xh3          Bc8xa6   
 23) Ne5-g6             Qd4-b2   
 24) Rf1-f3             f5xg6    
Y 25) Rh3-h5             Rh3-h5  * 0 Seconds
 26) Rd1-b1             Rd1-d3   
Y 27) .. Ra4xe4          Ra4xe4  * 0 Seconds
Y 28) Bf3xh5             Bf3xh5  * 0 Seconds
 29) .. Qf6-g5          Qf6-g5  * 0 Seconds
 30) Rb1-e1             Ne6xf8   
 31) .. Be5xd4          Ke7-d6   
 32) Qb2xb5             Qb2xb5  * 0 Seconds
Y 33) Rd2-d3             Ne5xf7   
Y 34) Nb3-c5             Nb3-c5  * 12 Seconds
 35) .. Ng6-h4          Rh3xe3   
 36) Nd4-f5             Qe2-h5   
Y 37) Rf1-f5             Rf1-f5  * 0 Seconds
Y 38) .. Ra2xe2          Ra2xe2  * 7 Seconds
Y 39) .. Ng3-e2          Ng3-e2  * 5 Seconds
Y 40) Qe2-h5             Qe2-h5  * 0 Seconds
 41) Rf5xf7             h2xg3    
 42) .. Be7-d6          g5-g4    
Y 43) .. Rc3xc7          Rc3xc7  * 13 Seconds
 44) Bg2-h3             Bg2-h3  * 0 Seconds
Y 45) Ra8-a7             Ra8-a7  * 10 Seconds
 46) Nf3xd4             Bh6-e3   
Y 47) .. Qc5-d4          Qc5-d4  * 14 Seconds
 48) Bc2xh7             Bc2xh7  * 0 Seconds
 49) .. f4-f3           f4-f3   * 0 Seconds
 50) Re1-e4             Re1-e4  * 0 Seconds
 51) Bh6-g7             Ra3xa4   
Y 52) Qf1-g1             Rg7-g1   
 53) .. Ba8xe4          Ng7xh5   
 54) .. Ng5xf3, Bg4xf3  Qh6xh2   
 55) Bf4xh6             Bf4xh6  * 0 Seconds
 56) Bf7xg6             Bf7xg6  * 4 Seconds
 57) .. Bg6-e4          Bg6-e4  * 0 Seconds
Y 58) .. Bd6xh2          Bd6xh2  * 0 Seconds
Y 59) Rc6-d6             Rc6-d6  * 0 Seconds
 60) Rg1-g4             Rg1-g4  * 9 Seconds
Y 61) e5xf6              e5xf6   * 0 Seconds
 62) Qf3xf8             Qf3xc6   
Y 63) .. c7-c6           c7-c6   * 0 Seconds
 64) Qe4-e5, Qe4-c6, Qe4-b7, Kg2-g1  f6-f7    
 65) .. Nc6xd4          Nc6xd4  * 0 Seconds
 66) .. Ra7-g7          Ra7xa1   
 67) Bc3xe5             Bc3xe5  * 0 Seconds
Y 68) .. Qh6-h3          Qh6-h3  * 1 Second
Y 69) Bg6-h7             Bg6-h7  * 0 Seconds
Y 70) Rh1xh7             Rh1xh7  * 0 Seconds
Y 71) .. Rg7-h7          Bb6xf2   
Y 72) Bd3xh7             Bd3xh7  * 0 Seconds
Y 73) .. Bb5-c6          Bb5xa4   
 74) Rb6xc6             Rb6xc6  * 4 Seconds
Y 75) d5-d6              d5-d6   * 0 Seconds
Y 76) .. Rh8xh2          Rh8xh2  * 0 Seconds
Y 77) f4-f5              Be7-b4   
Y 78) c4xd5              c4-c5    
 79) .. e4-e3           Re7xe6   
 80) Rf1-f3             Rf1-f3  * 2 Seconds
 81) Kc4xd4             Nc8-b6   
 82) .. Rd7xd6          Rd7xd6  * 8 Seconds
Y 83) Nh4-f3             Nh4-f3  * 14 Seconds
Y 84) .. Kf6xg6          Kf6xg6  * 1 Second
Y 85) Bc2xg6             Bc2xg6  * 2 Seconds
 86) Kg1-h2             Kg1-h2  * 1 Second
 87) f5-f6              f5-f6   * 0 Seconds
 88) Bg7-e5             Rg1xh1   
 89) .. Nc1xe2          Nc1xb3   
 90) Qf2xd2             Qf2xd2  * 3 Seconds
Y 91) Rc4xc5             Rc4xc5  * 5 Seconds
Y 92) Be2-c4             Be2-c4  * 0 Seconds
Y 93) Ng5-e6             Rg1-f1   
Y 94) .. g4-g3           g4-g3   * 0 Seconds
Y 95) .. g5-g4           g5-g4   * 7 Seconds
 96) e5-e6              e5-e6   * 9 Seconds
Y 97) g5-g6              g5-g6   * 3 Seconds
 98) .. Nh4xg2          Nh4xg2  * 0 Seconds
 99) Ng3-f5             Nd5xb4   
100) Rb2-g2             Rb2-g2  * 8 Seconds
   Lc0: 61 of 100 matching moves
   7-5-20 03:21:35, Total time: 01:45:49 Rated time: 12:17 = 737 Seconds
Honey: 47 of 100 matching moves
7-5-20 05:33:53, Total time: 03:58:07 Rated time: 17:54 = 1074 Seconds

== 34 by both 
Parent - - By user923005 (****) Date 2020-05-07 18:50
You have really good tactical results with NN.
The LC0 run I did had the biggest net of all, which is probably not good for one minute tactical analysis.
Parent - - By user923005 (****) Date 2020-05-07 18:52
Specifically, I used:
384x30-t60-3350-lr0.002.pb.gz
Parent - By MrKris (***) Date 2020-05-07 20:58
https://www.comp.nus.edu.sg/~sergio-v/t60/384x30/
3350 has LR0.002, evidentally the others are "Ongoing, LR: 0.001", the smaller the better I think -but not sure.

The much older (2 months!, also must be from old official t60 nets) 384x30-t60-3010 is supposed to be the strongest of this type:
http://lczero.org/play/bestnets/
(It says the older t40 1705 is better than the 2036 I used, also.)

And it was used in the just completed TCEC 17 Superfinal and CCC 13 Final (instead of newer ones at the SV link at top).

And, I have this so far in my net elimination matches: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=586344
Lc0 v0.24.1:
#1  SV384x30-t60-3200  126.5-98.0   128x10-591219
#2  SV384x30-t60-3200  150.5-127.5  256x20-42850
#3  SV384x30-t60-3200  154.0-126.0  320x24J13B.2-178
#4  256x20-t40-1541SV  178.0-170.0  SV384x30-t60-3200  {8 Elo}
#5  256x20-t40-1541SV   36.5-5.5    test-512x40-320SV
#6  Leelenstein 13.1   175.0-173.0  256x20-t40-1541SV  {2 Elo}
#7  Leelenstein 13.1   148.5-113.5  384x30-t40-2036SV
#8  Leelenstein 13.2   142.0-120.0  320x24-net-63222
#9  Leelenstein 13.2    39.5-20.5   128x10-net-701473
#10 Leelenstein 13.2   118.5-87.5   SV384x30-t60-3350 - stopped early at Ls13.2 +39 -8 =159
#11 SV384x30-t60-3010  175.0/173.0  Leelenstein 13.2  {2 Elo}

3010 was slightly ahead of Ls13.2, but Ls13.2 had crushed 3350.

Strange situation.
Parent - - By dorszcz (**) Date 2020-05-08 11:49 Edited 2020-05-08 19:28
Good results. Unfortunately I don't have a decent graphics card to test GPU engines.
Parent - - By MrKris (***) Date 2020-05-14 04:50
Next time I compile a Lc0 for my Ubuntu I'll try to make a Lc0-CPU also just to see if it can solve any.

Thanks for your easy to use web site and you analysis and pgn files!
Parent - By dorszcz (**) Date 2020-05-16 13:56
When you have please share the results.
- - By MrKris (***) Date 2020-06-03 06:12
Add to my 2020-05-07 post above from:
https://github.com/joergoster/Moonfish

TTT1 @ 15" , R7 2700X 16ths | RTX 2060
Syzygy 6 50mrOn, 2 GBh

Black-Diamond-043020  68  (DeepDef,noCnoPro,Tac2)
Lc0.24.1-SVt40-1541   61
Lc0.24.1-Ls14         61
Lc0.24.1-SVlgt60-3010 57
Lc0-.23.2K-lg2036     55
Crystal-200420        50
Houdini-6.03-Std-tac1 49  (8ths, 8thsTac @ 30")
Moonfish-multiPV-03feb2020 49
Moonfish-13apr2020     47

Honey-043020          47
Stockfish_20050217    43
Houdini-6-Std-tac2    39  (8ths, 4thsTac @ 30")
Komodo-13.3           38
Houdini-6.03-Std      21  (8ths @ 30")
Parent - By dorszcz (**) Date 2020-07-29 20:07
64 out of 100 = 64% - SF NNUE halfkp-256 090720

http://dorszcz.blogspot.com/p/ttt1.html
- - By Peter Grayson (****) Date 2020-07-30 17:19
I used the Stockfish abrok BMI2 version on my slower i7-8700 CPU giving it 30s/position. There was no doubt the nnue engines were the  best solvers in this test set.
Clean load for each run ...

TTT.100 test. 30s/pos. i7-8700 12 threads 1 Gb LP hash allocated.

Stockfish 290720 64 BMI2 nnue 20200728-2138 - 77
Stockfish 290720 64 BMI2 nnue 20200730-1934 - 77
Stockfish 290720 64 BMI2 - 58
Stockfish 170720 64 BMI2 - 57
Stockfish 11 64 BMI2 - 51

Peter
Parent - - By dorszcz (**) Date 2020-08-06 16:01
2 different engines with the same net.

72 out of 100 = 72% - SF NNUE halfkp-256 090720 (1403)
59 out of 100 = 59% - Stockfish 050820 (1403)

http://dorszcz.blogspot.com/p/ttt1.html
Parent - - By Peter Grayson (****) Date 2020-08-08 13:34 Edited 2020-08-08 13:40
Yes, I have found the same thing introduced with the abrok 060820 updates, probably the move to hybrid use of the evaluation.

i7-8700 engines 12 threads + 1Gb LP hash  30s per position.
Running the pre hybrid evauation engine with the 1802 net gave a score of

Stockfish 040820 nnue 64 BMI2 20200806-1802 scored 80/100
but the hybrid eval engine ...

Stockfish 060820 nnue 64 BMI2 20200806-1802 scored 58/100

Incidentally the final 310720 alpha-beta engine did better suggesting the hybrid makes it worse than single evaluation method.
Stockfish 310720 64 BMI2 scored 64

the different scores are way outside the typical variation seen over say 5 to 10 runs alone.

The move to hybrid changed the way the engine nnue switch operates
Previously when
switch was checked neural network evaluations were used
switch unchecked classical alpha-beta were used

now
switch is checked hybrid nnue and alpha-beta evaluations automatically determined
switch is unchecked alpha-beta evaluation is used
thus losing the ability to run with nnue evaluation alone resulting in a significantly worse tactical solving capability in the TTT1 100 test set.

Ideally instead of a binary switch it would be useful to have a choice of three options,
0= default alpha-beta + nnue evaluation as current configuration
1= nnue evaluation used
2= alpha-beta evaluation used

In addition I started a match Game in 10m + 3s/move between the 040820 engine that uses nnue and the 060820 engine that changed to nnue + alpha-beta evaluation. Only thirty games completed so far but no indication of any benefit using the hybrid evaluation with the score 1/28/1. However it would have been useful to be able to carry a direct relational match between the two seperate evalution schemes using the same engine to avoid engine distortion. There is every chance the hybrid benefit tested at bullet speeds dissipates very quickly with longer time controls.

The nnue evaluation may be slower than the classical evaluation but there is evidence it is more accurate finding tactical opportunities. However it is not absolute when for example in the early Richter 14 position test set the 010720 engine was solving all positions very quickly but the 040820 nnue only engine failed to solve 3. Perhaps more work should be done on this assessment.

The Richter.epd positions ...
http://www.mediafire.com/file/782sc7lql51qhrb/Richter.epd/file

Peter
Parent - - By user923005 (****) Date 2020-08-08 16:31 Upvotes 1
I made a version where you can turn off SF eval and have it use nnue 100%
https://drive.google.com/file/d/1TzPjGwuultaEn3fZt8MPQzdx2rBvrCNo/view?usp=sharing

The new UCI parameter is called AlwaysNNUE

You will probably want to turn off logging if you use my build.
I only have the one net that they use in the project, so you will want to browse to a good one and use that instead.
Parent - - By Peter Grayson (****) Date 2020-08-08 18:58
Thanks very much Dann, I'll try it out.
Peter
Parent - - By user923005 (****) Date 2020-08-08 19:07 Upvotes 1
It might be interesting to try a series of contests:
1) tournament with the three programs at short, medium and long time controls [*]
2) problem solving at ten seconds per problem with 500 difficult problems for all three engines

[*] by "the three programs" I mean SF alpha-beta, SF hybrid, SF nnue
Parent - - By Peter Grayson (****) Date 2020-08-09 13:06
I have previously run a series of games at G/60s + 1s/move between the pure nnue and alpha-beta engines using my dual Xeon machine and they confirmed the superiority of the nnue engine. However, looking through the games I was not very impressed with the quality, particularly with the scoring when in one case the nnue engine could not win with a 7+ score advantage. Therefore I have settled with running fewer matches but at the G/10m +3s/move time control that I found well suited to opening book testing and the quality of the games looks significantly better with less frequent erroneous scoring of the positions during the games.

The original concern with the hybrid approach was that position solving indicated a worse performance than the pure alpha-beta and significantly worse than the pure nnue engines. As the match between the 040820 pure nnue engine and the 060820 hybrid engine draws to a close it does look as though there will be a small gain for the hybrid engine under match conditions and I suppose the Stockfish engine has always been about producing the best game playing engine on which there is little doubt the programmes have achieved that. In just a month or so, already the noticeable improvement is the reduction of the slope in strength deterioration with increasing time controls graph where the improvements are holding up quite well at my time control.

With the engine updates from abrok coming thick and fast, the engine characteristics may look very different in a month or two's time when perhaps the hybrid evaluation approach may turn out to be best for problem solving too. I note the hybrid criteria has already been refined and the only problem is finding the time to run so many tests with each new release. Perhaps we wil have to be reliant on the derivative engines for best position solving that I believe has been the case with the alpha-beta engine?

Peter
Parent - - By user923005 (****) Date 2020-08-09 13:10 Upvotes 1
I am hoping that enough interested people will run enough matches at different settings and time controls that we will have enough data to make intelligent decisions.
Right now, we are all shooting from the hip. (Country expression for making a crude approximation)
Parent - - By Peter Grayson (****) Date 2020-08-10 08:31
The completed match saw the hybrid engine performing better. Generally there were no outrageous scoring errors. Game 97 promised something for the nnue engine but it appears to have been overly optimistic although the hybrid engine also scored it was adverse.

The 040820 engine was the last pure nn engine release by abrok used for the position solving of TTT1 100.
The 060820 engine was the first of the abrok hybrid releases that scored significantly worse in the TTT1 100 test but match result confirms the engine is game result oriented.

PC: Intel i7-8700 + 64Gb RAM H-T on,
O/S: Windows 10 64 Pro v2004
Engines: abrok Stockfish 040820 +NNUE  64 BMI2 + Sergio 20200806-1802.bin network,
Stockfish 060820 64 BMI2 + Sergio 20200806-1802.bin network
GUI: Fritz 14 64 bit update 46. No adjudication, games played to a finish.
Engine resources: 12 Threads + 4Gb hash per engine. Ponder off.
Time control: G/10 minutes + 3s/m
Openings: Fixed PG 260816 x50 lines openings set. Each engine plays both colours.
Fritz 14 Tournament mode. Engines are unloaded after each game ensuring clean hash for next game. No adjudication.

nn or nn+a-b  2020
                                        
1   Stockfish 060820 64 BMI2 1802   +10  +5/=93/-2 51.50%   51.5/100
2   Stockfish 040820 64 BMI2 1802   -10  +2/=93/-5 48.50%   48.5/100

Games available in pgn and cbv format:
http://www.mediafire.com/file/74wz52ocie2iafk/nn_or_nn%252Ba-b.zip/file

Peter
Parent - By user923005 (****) Date 2020-08-10 12:26
That is still a pretty close match.  I wonder if the hybridization will stunt the growth of the net
Parent - By Peter Grayson (****) Date 2020-08-10 17:12
I ran the position that was posted in the abrok comments section. I forgot on the old Q9550 quad the base version 040820 engine was relatively slow in nnue mode but the result was quite revealing especially given the hybrid bersion failed to solve after 10 minutes on the faster i7-8700 machine. Needless to say the pure nnue 040820 engine solved it instantly on the i7 8700 machine.

167: New game, Rd8! 2020
r1b1qr1k/2p3pp/4p3/1pb1PpN1/pn3N1P/8/PPP1QPP1/2KR3R w - - 0 1


Q9550 + 1Gb hash or engine.
Analysis by Stockfish 040820 64 20200806-1802:

1.Qf3 Ra7 2.Kb1 Nd5 3.h5 Nxf4 4.h6 g6 5.Qxf4 Be7 6.Qe3 c5
  -+ (-2.88)  Depth: 11/20   00:00:01
1.Qf3 Ra7 2.Kb1 Nd5 3.Rxd5 exd5 4.h5 c6 5.h6 g6 6.e6 b4 7.Qg3
  -+ (-3.07)  Depth: 12/22   00:00:01  249kN
1.Qf3 Ra7 2.Kb1 Kg8 3.g4 Qc6 4.Qe2 Nd5 5.Nxd5 exd5 6.e6 fxg4
  -+ (-2.60)  Depth: 13/23   00:00:01  278kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qd7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1
  =/+ (-0.66)  Depth: 14/23   00:00:02  706kN
1.Rd8 Be3+ 2.Kb1 Qxd8 3.Qh5 h6 4.fxe3 Kg8 5.Qg6 hxg5 6.hxg5 Nd5 7.Qh7+ Kf7 8.Qg6+ Ke7 9.Qxg7+ Rf7
  -/+ (-1.04)  Depth: 15/32   00:00:03  752kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Rxf6 7.exf6 Nxa2+ 8.Kb1 Qf8 9.Rh7
  = (-0.06)  Depth: 16/32   00:00:03  771kN
1.Rd8 Be3+ 2.fxe3 Nxa2+ 3.Kb1 Qxd8 4.Qh5 Nc3+ 5.bxc3 h6 6.Qg6 hxg5 7.hxg5+ Kg8 8.Nh5 Qd7 9.Nf6+ Rxf6 10.exf6
  +/= (0.49)  Depth: 17/32   00:00:03  821kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 gxf6 9.exf6+ Kxd8 10.Qxf7 a3 11.Kb1
  +- (1.64)  Depth: 18/32   00:00:03  904kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 gxf6 9.exf6+ Kxd8 10.Qxf7 a3 11.Kb1
  +- (1.73 ++)  Depth: 19/32   00:00:03  991kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 gxf6 9.exf6+ Kxd8 10.Qxf7 a3 11.Kb1
  +- (1.83 ++)  Depth: 19/32   00:00:03  998kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 gxf6 9.exf6+ Kxd8 10.Qxf7 a3 11.Kb1
  +- (1.97 ++)  Depth: 19/32   00:00:03  1000kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 gxf6 9.exf6+ Kxd8 10.Qxf7 a3 11.Kb1
  +- (2.16 ++)  Depth: 19/32   00:00:03  1006kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 gxf6 9.exf6+ Kxd8 10.Qxf7 a3 11.Kb1
  +- (2.43 ++)  Depth: 19/32   00:00:03  1011kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (2.27)  Depth: 19/32   00:00:03  1012kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (2.36 ++)  Depth: 20/32   00:00:03  1097kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (2.46 ++)  Depth: 20/32   00:00:03  1139kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (2.59 ++)  Depth: 20/32   00:00:03  1166kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (2.79 ++)  Depth: 20/32   00:00:03  1182kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (3.05 ++)  Depth: 20/32   00:00:03  1196kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Qe7 6.Nf6+ Rxf6 7.gxf6 Nxa2+ 8.Kb1 Nc3+ 9.bxc3 Qf7 10.Rh8+ Kxh8 11.Qxf7
  +- (3.41 ++)  Depth: 20/33   00:00:04  1255kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (3.50)  Depth: 20/33   00:00:04  1265kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (3.59 ++)  Depth: 21/28   00:00:04  1614kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (3.68 ++)  Depth: 21/30   00:00:04  1629kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (3.82 ++)  Depth: 21/34   00:00:04  1667kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (4.01 ++)  Depth: 21/34   00:00:05  1708kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (4.28 ++)  Depth: 21/34   00:00:05  1790kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (4.63 ++)  Depth: 21/34   00:00:05  1872kN
1.Rd8 Qxd8 2.Qh5 h6 3.Qg6 hxg5 4.hxg5+ Kg8 5.Nh5 Rf7 6.Nf6+ Kf8 7.Rh8+ Ke7 8.Rxd8 Kxd8 9.Qxf7 Nxa2+ 10.Kb1 gxf6 11.g6 Nc3+ 12.bxc3 fxe5 13.g7
  +- (4.89)  Depth: 21/34   00:00:05  1920kN

(Grayson, Newport, South Wales 10.08.2020)
Parent - By user923005 (****) Date 2020-08-08 19:10 Upvotes 1
I should also mention that I turn off the default contempt, because I mostly use SF for analysis and not game play.
Parent - - By rov (*) Date 2020-08-08 19:57 Edited 2020-08-08 20:09
I get error during unzip archive, and after make:

*** No rule to make target 'src/stockfish.o', needed by 'Stockfish-master2'.  Stop.".

This error can say what I did wrong ?
Parent - By user923005 (****) Date 2020-08-08 22:46
There is a binary already present.
If you want make it yourself ( a good idea ) you need msys2 and g++ (a very recent version because it uses C++ 17 features ).
You can type make from the bash prompt and you should see this:


dcorbit@DCORBIT MINGW64 /g/project/dcorbit/stockfish-master2/src
# make

To compile stockfish, type:

make target ARCH=arch [COMP=compiler] [COMPCXX=cxx]

Supported targets:

build                   > Standard build
profile-build           > Standard build with PGO
strip                   > Strip executable
install                 > Install executable
clean                   > Clean up
net                     > Download the default nnue net

Supported archs:

x86-64-avx512           > x86 64-bit with avx512 support
x86-64-bmi2             > x86 64-bit with bmi2 support
x86-64-avx2             > x86 64-bit with avx2 support
x86-64-sse42            > x86 64-bit with sse42 support
x86-64-modern           > x86 64-bit with sse41 support (x86-64-sse41)
x86-64-sse41            > x86 64-bit with sse41 support
x86-64-ssse3            > x86 64-bit with ssse3 support
x86-64-sse3-popcnt      > x86 64-bit with sse3 and popcnt support
x86-64-sse3             > x86 64-bit with sse3 support
x86-64                  > x86 64-bit generic
x86-32                  > x86 32-bit (also enables SSE)
x86-32-old              > x86 32-bit fall back for old hardware
ppc-64                  > PPC 64-bit
ppc-32                  > PPC 32-bit
armv7                   > ARMv7 32-bit
armv8                   > ARMv8 64-bit
apple-silicon           > Apple silicon ARM64
general-64              > unspecified 64-bit
general-32              > unspecified 32-bit

Supported compilers:

gcc                     > Gnu compiler (default)
mingw                   > Gnu compiler with MinGW under Windows
clang                   > LLVM Clang compiler
icc                     > Intel compiler

Simple examples. If you don't know what to do, you likely want to run:

make -j build ARCH=x86-64    (This is for 64-bit systems)
make -j build ARCH=x86-32    (This is for 32-bit systems)

Advanced examples, for experienced users:

make -j build ARCH=x86-64-modern COMP=clang
make -j profile-build ARCH=x86-64-bmi2 COMP=gcc COMPCXX=g++-4.8

The selected architecture x86-64-modern enables the following configuration:

make[1]: Entering directory '/g/project/dcorbit/stockfish-master2/src'

Config:
debug: 'no'
sanitize: 'no'
optimize: 'yes'
arch: 'x86_64'
bits: '64'
kernel: 'MINGW64_NT-6.3-9600'
os: 'Windows_NT'
prefetch: 'yes'
popcnt: 'yes'
sse: 'yes'
sse3: 'yes'
ssse3: 'yes'
sse41: 'yes'
sse42: 'no'
avx2: 'no'
pext: 'no'
avx512: 'no'
neon: 'no'

Flags:
CXX: g++
CXXFLAGS: -Wall -Wcast-qual -fno-exceptions -std=c++17  -pedantic -Wextra -Wshadow -m64 -DNDEBUG -O3 -mtune=native -DIS_64BIT -msse -msse3 -mpopcnt -DUSE_POPCNT -DUSE_SSE41 -msse4.1 -DUSE_SSSE3 -mssse3 -DUSE_SSE3 -msse3 -DUSE_SSE2 -flto
LDFLAGS:  -m64 -Wl,--no-as-needed -lpthread -Wall -Wcast-qual -fno-exceptions -std=c++17  -pedantic -Wextra -Wshadow -m64 -DNDEBUG -O3 -mtune=native -DIS_64BIT -msse -msse3 -mpopcnt -DUSE_POPCNT -DUSE_SSE41 -msse4.1 -DUSE_SSSE3 -mssse3 -DUSE_SSE3 -msse3 -DUSE_SSE2 -flto

Testing config sanity. If this fails, try 'make help' ...

make[1]: Leaving directory '/g/project/dcorbit/stockfish-master2/src'

dcorbit@DCORBIT MINGW64 /g/project/dcorbit/stockfish-master2/src
#

For the binary in the archive, I typed:
make profile-build ARCH=x86-64-mingw COMP=mingw
Parent - - By user923005 (****) Date 2020-08-08 22:48
I guess also that the error than happened when you unzipped the archive was important.
What exactly did the error say?
Parent - - By rov (*) Date 2020-08-09 07:42
Finally problem was in ubuntu, I get decopresed internal error,
In windows 7zip can do it without problem.
Compilation in ubuntu works ok,
Parent - By user923005 (****) Date 2020-08-09 12:29
I am glad you got it working.
And sorry for the terrible insult of assuming you were using Windows.
:razz:
Parent - - By dorszcz (**) Date 2020-08-09 19:41
Engine updates is one thing, but a lot depends on net.
Shortly after net 1403 I tested, 1802 was published. Both in the same day. I tested it with the same engine before hybrid use of the evaluation was introduced. The result was only 58%. It is 24% less than 1403 on the same engine. Stockfish with 1403 and with 1802 played very differently. It is almost impossible to say which net is better because they are published often and I am not going to test them all.
To make it even more complex I think that different version of Stockfish with different net will play even more differently so with everyday Stockfish updates and daily net updates it is impossible to verify what mixture is best.
Parent - By user923005 (****) Date 2020-08-09 20:03
Unfortunately,  the guy who wrote SFNNUE is not going to maintain it now that the code has been incorporated into SF.
Parent - By Peter Grayson (****) Date 2020-08-09 22:24
Yes it is impossible to keep up with all of the different changes. In the space of my 100 game match over 2 days the update and network releases make the match obsolete or at least unrepresentative of what is available today. I still have Dann's offering to check out too! No chance of absolute confirmation which is best for position solving and it may come down to one of the specialist engines to provide that consistency, perhaps Crystal, ShashChess, CorChess or maybe even Dann's engine!

Peter
Up Topic The Rybka Lounge / Test Positions / Tough Tactical Test 1

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill