New net: The White Rose

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

dkappe
Posts: 1631
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

New net: The White Rose

Post by dkappe »

I’ve approached the training with Dark Horse data a little differently. It’s a little under inflated, but not as much as DH-0.2.

https://www.patreon.com/posts/47771982

Note that this net is distributed under the Creative Commons attribution, no derivatives license. As such it cannot be embedded and distributed in a GPLv3 work or others with an incompatible license.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
peter
Posts: 3186
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: New net: The White Rose

Post by peter »

dkappe wrote: Sat Feb 20, 2021 1:16 am I’ve approached the training with Dark Horse data a little differently. It’s a little under inflated, but not as much as DH-0.2.

https://www.patreon.com/posts/47771982

Note that this net is distributed under the Creative Commons attribution, no derivatives license. As such it cannot be embedded and distributed in a GPLv3 work or others with an incompatible license.
Thanks a lot, Dietrich, at least tactiacally again a very strong one!

SugaR AI ICCF 1.40 on an old Xeon X5670 CPU (SSE4.1 popcnt) 24 threads of 3x12Ghz, 4G hash, MulitPV=3:

Code: Select all

Löse: C:\Users\Dr. Martan\Documents\ChessBase\Testsets\ERET - CBH\Eigenmann Rapid Engine Test.cbh
Maximale Lösungszeit = 30s.

1. ERET 001 - Entlastung,  Akopian-Khenkin, Tilburg 1994   Gelöst in 0.50s/9; Gelöst: 1
2. ERET 002 - Zugzwang,  Akerblom 1961   Gelöst in 4.20s/10; Gelöst: 2
3. ERET 003 - Linienoeffnen,  Alvarez-Kokkila, CorrGame 1997   Gelöst in 5.10s/15; Gelöst: 3
4. ERET 004 - Endspiel L&S vs L&S,  Anand-Shirov, Advanced Chess 2000   Gelöst in 3.46s/20; Gelöst: 4
5. ERET 005 - Laeuferopfer f7,  Andriulaitis-Ruefenacht, CorrGame 1999   Gelöst in 2.37s/12; Gelöst: 5
6. ERET 006 - Springeropfer,  Annageldyev-Akopian, Moskau 1994   Gelöst in 2.82s/13; Gelöst: 6
7. ERET 007 - Laeuferpaar,  Aronian-Anand, WiikAanZee 2013   Gelöst in 4.26s/14; Gelöst: 7
8. ERET 008 - Zentrum,  Aronian-Kramnik, Monte Carlo 2011   Gelöst in 0.52s/9; Gelöst: 8
9. ERET 009 - Springeropfer,  Aseev-Popv, Kazan 1995   Gelöst in 3.07s/13; Gelöst: 9
10. ERET 010 - Freibauer,  Atalik-Bregadze, St Louis 2013   Gelöst in 0.45s/8; Gelöst: 10
11. ERET 011 - Rochadeangriff,  Atalik-Sax, Szeged 1997   Gelöst in 3.87s/16; Gelöst: 11
12. ERET 012 - Entlastung,  Babula-Hoffmann, Dt-BL 2011   Gelöst in 0.73s/10; Gelöst: 12
13. ERET 013 - Zentrum,  Bang-Umansky, CorrGame 1996   Gelöst in 0.39s/7; Gelöst: 13
14. ERET 014 - Endspiel T&L&S vs T&L&S,  Barbre-D'Arruda, CorrGame 1994   Gelöst in 3.46s/16; Gelöst: 14
15. ERET 015 - Endspiel D&B vs T&L&B,  Becker 2015   Gelöst in 3.37s/15; Gelöst: 15
16. ERET 016 - Pos. Opfer,  Bischoff-Fressinet, Dt. Bundesliga 2008   Gelöst in 0.55s/9; Gelöst: 16
17. ERET 017 - Koenigsangriff,  Boudy-Szabo, Varna 1979   Gelöst in 18.27s/19; Gelöst: 17
18. ERET 018 - Entwicklung,  Bromberger-Areshchenko, Dt-BL 2010   Gelöst in 16.35s/16; Gelöst: 18
19. ERET 019 - Endspiel L&L&S&S vs L&S&B,  Broni 1970   > 30s.
20. ERET 020 - Laeuferopfer h7,  Burmakin-Ziatdinov, Moskau 2009   Gelöst in 6.19s/16; Gelöst: 19
21. ERET 021 - Prophylaxe,  Carlsen-Nakamura, London 2010   Gelöst in 4.21s/14; Gelöst: 20
22. ERET 022 - Freibauer,  Carlsen-Nakamura, Medias 2011   Gelöst in 0.34s/7; Gelöst: 21
23. ERET 023 - Rochadeangriff,  Copie-Patrici, CorrGame 1986   Gelöst in 1.08s/11; Gelöst: 22
24. ERET 024 - Entwicklung,  Critter-Komodo, CompGame 2016   Gelöst in 0.39s/7; Gelöst: 23
25. ERET 025 - Endspiel T&L vs T&L,  Critter-Naum, CompGame 2015   Gelöst in 0.39s/9; Gelöst: 24
26. ERET 026 - Springeropfer,  Degraeve-Aseev, St.Petersburg 1996   Gelöst in 6.99s/17; Gelöst: 25
27. ERET 027 - Zugzwang,  DeLimburg 1900   Gelöst in 3.53s/21; Gelöst: 26
28. ERET 028 - Vergifteter Bauer,  Diep-DeepJunior, WCCC 2004   Gelöst in 0.39s/7; Gelöst: 27
29. ERET 029 - Qualitaetsopfer,  Dolinskij-Tosic, Moskau 1999   Gelöst in 1.30s/13; Gelöst: 28
30. ERET 030 - Initiative,  Dutra-Alfredo, CorrGame 2008   Gelöst in 3.50s/13; Gelöst: 29
31. ERET 031 - Endspiel L vs L,  Estrin-Ivashin, CorrGame 1947   Gelöst in 0.61s/11; Gelöst: 30
32. ERET 032 - Zugzwang,  Fahrni 1922   Gelöst in 0.50s/12; Gelöst: 31
33. ERET 033 - Initiative,  Fedorchuk-Sokolov, Khanty-Mansiysk 2009   Gelöst in 4.31s/16; Gelöst: 32
34. ERET 034 - Laeuferpaar,  Fischer-Incutto, Mar Del Plata 1960   Gelöst in 1.90s/13; Gelöst: 33
35. ERET 035 - Qualitaetsopfer,  FischerW-Siigur, CorrGame 2009   Gelöst in 1.05s/12; Gelöst: 34
36. ERET 036 - Endspiel T&L&S vs T&L&S,  Fritz 1965   Gelöst in 10.30s/18; Gelöst: 35
37. ERET 037 - Zugzwang,  Garcia 2000   Gelöst in 10.27s/21; Gelöst: 36
38. ERET 038 - Endspiel S&S vs L&B,  Garcia 2000   Gelöst in 4.23s/23; Gelöst: 37
39. ERET 039 - Entwicklung,  Gelfand-Ivanchuk, Dagomys 2009   Gelöst in 0.47s/9; Gelöst: 38
40. ERET 040 - Koenigssicherheit,  Gerasin-Dronov, CorrGame 1991   Gelöst in 0.42s/7; Gelöst: 39
41. ERET 041 - Springeropfer,  Gerber-Georgiev, Luzern 2003   Gelöst in 0.41s/7; Gelöst: 40
42. ERET 042 - Endspiel S vs L,  Glaurung-Rybka (Var), CompGame 2009   Gelöst in 0.36s/8; Gelöst: 41
43. ERET 043 - Springeropfer,  Glek-Ulibin, Wien 1998   Gelöst in 0.48s/9; Gelöst: 42
44. ERET 044 - Linienoeffnen,  Gligoric-Ghitescu, Zagreb 1970   Gelöst in 0.55s/10; Gelöst: 43
45. ERET 045 - Initiative,  Gouw-BlancoG, CorrGame 2001   Gelöst in 2s/13; Gelöst: 44
46. ERET 046 - Endspiel T&S vs T&S,  Grandelius-Raznikov, Albena 2011   Gelöst in 20.76s/23; Gelöst: 45
47. ERET 047 - Entlastung,  Grischuk-Aronian, Ohrid 2009   Gelöst in 0.92s/11; Gelöst: 46
48. ERET 048 - Starke Felder,  Grivas-Svetushkin, Glifada 2001   Gelöst in 0.44s/8; Gelöst: 47
49. ERET 049 - Entwicklung,  Gufeld-Ivanovic, Socchi 1979   Gelöst in 0.47s/8; Gelöst: 48
50. ERET 050 - Koenigsangriff,  Hansen-Barua, Biel 1993   Gelöst in 3.40s/14; Gelöst: 49
51. ERET 051 - Verteidigung,  Harding-Zavanelli, CorrGame 1985   Gelöst in 0.38s/7; Gelöst: 50
52. ERET 052 - Starke Felder,  HarEven-Kofidis, CorrGame 1996   Gelöst in 0.91s/11; Gelöst: 51
53. ERET 053 - Pos. Opfer,  Ivanisevic-Nisipeanu, Istanbul 2003   Gelöst in 1.01s/11; Gelöst: 52
54. ERET 054 - Endspiel D&T&L vs D&T&L,  Jussupow-Kasparov, Linares 1990   Gelöst in 1.05s/12; Gelöst: 53
55. ERET 055 - Laeuferopfer h6,  Kamsky-Lautier, Dortmund 1993   Gelöst in 0.55s/9; Gelöst: 54
56. ERET 056 - Zwischenzug,  Karjakin-Kramnik, Moskau 2010   Gelöst in 1.84s/12; Gelöst: 55
57. ERET 057 - Abtausch,  Kashlyak-Vandermeulen, CorrGame 2008   Gelöst in 6.19s/16; Gelöst: 56
58. ERET 058 - Endspiel T&L vs T&L,  Kasparian 1954   Gelöst in 1.06s/12; Gelöst: 57
59. ERET 059 - Endspiel D&L vs D&T,  Komodo-Stockfish, CompGame 2014   Gelöst in 1.15s/13; Gelöst: 58
60. ERET 060 - Koenigsangriff,  Kortschnoj-Andersson, Ubeda 1997   Gelöst in 1.95s/13; Gelöst: 59
61. ERET 061 - Koenigsangriff,  Kortschnoj-Shirov, Madrid 1996   Gelöst in 1.03s/11; Gelöst: 60
62. ERET 062 - Starke Felder,  Kotronias-Davies, Gausdal 1094   Gelöst in 0.41s/7; Gelöst: 61
63. ERET 063 - Verteidigung,  Kovacs-Sperkin, CorrGame 2012   Gelöst in 22s/19; Gelöst: 62
64. ERET 064 - Endspiel T vs T,  Ktulu-Hiarcs (Var), CompGame 2005   Gelöst in 0.39s/7; Gelöst: 63
65. ERET 065 - Zwischenzug,  Kuligowski-Knaak, Slupsk 1982   Gelöst in 5.55s/15; Gelöst: 64
66. ERET 066 - Endspiel T&S vs L&S,  Marwitz 1965   Gelöst in 2.96s/19; Gelöst: 65
67. ERET 067 - Raeumung,  Moisieev-Simagin, Moskau 1951   Gelöst in 1.17s/11; Gelöst: 66
68. ERET 068 - Offene Linie,  Morozevich-Kogan, London 1994   Gelöst in 6.27s/18; Gelöst: 67
69. ERET 069 - Koenigsangriff,  Morozevich-Petursson, London 1994   Gelöst in 0.39s/7; Gelöst: 68
70. ERET 070 - Starke Felder,  Motylev -Li, China 2010   Gelöst in 3.42s/14; Gelöst: 69
71. ERET 071 - Ablenkung,  Myers-Baufays, CorrGame 2010   Gelöst in 0.36s/7; Gelöst: 70
72. ERET 072 - Zentralisierung,  N.N.-N.N. (Analyse)   Gelöst in 2.09s/12; Gelöst: 71
73. ERET 073 - Mobilitaet,  N.N.-N.N. (Analyse)   Gelöst in 0.39s/7; Gelöst: 72
74. ERET 074 - Endspiel T&S vs T&S,  N.N.-N.N. (Analyse)   Gelöst in 0.38s/7; Gelöst: 73
75. ERET 075 - Festung,  N.N.-N.N. (Analyse)   Gelöst in 18.25s/18; Gelöst: 74
76. ERET 076 - Entwicklung,  Nakamura-VanWely, Amsterdam 2010   Gelöst in 0.44s/8; Gelöst: 75
77. ERET 077 - Rochadeangriff,  Negi-Volokitin, ESP 2008   Gelöst in 0.39s/7; Gelöst: 76
78. ERET 078 - Freibauer,  Niedermayr-Bachschmidt, CorrGame 1968   Gelöst in 0.44s/8; Gelöst: 77
79. ERET 079 - Damenopfer,  Nijboer-Winants, WijkAanZee 1988   Gelöst in 5.35s/16; Gelöst: 78
80. ERET 080 - Raeumung,  Nirvana-Hannibal, CompGame 2015   Gelöst in 12.17s/17; Gelöst: 79
81. ERET 081 - Koenigsangriff,  Nunn-Ward, GB 1998   Gelöst in 0.55s/9; Gelöst: 80
82. ERET 082 - Endspiel S vs L,  Nurmamed.-Volovici (Var), CorrGame 1960   Gelöst in 0.61s/11; Gelöst: 81
83. ERET 083 - Abtausch,  Orlov-Georgiev, Belgrad 1988   Gelöst in 0.42s/7; Gelöst: 82
84. ERET 084 - Koenigsangriff,  Persson-Teichmeister, CorrGame 2009   Gelöst in 8.22s/17; Gelöst: 83
85. ERET 085 - Abtausch,  Petrosian-Bronstein, UdSSR-Ch 1967   Gelöst in 4.01s/15; Gelöst: 84
86. ERET 086 - Qualitaetsopfer,  Petrosian-Fischer, BuenosAires 1971   Gelöst in 5.66s/15; Gelöst: 85
87. ERET 087 - Raeumung,  Petrosian-Larsen, PalmaDeMallorca 1968   Gelöst in 3.79s/13; Gelöst: 86
88. ERET 088 - Endspiel S vs S,  Razuvaev-Ostojic,  1988   Gelöst in 0.39s/8; Gelöst: 87
89. ERET 089 - Unterverwandlung,  Rumänien 1977   Gelöst in 0.42s/7; Gelöst: 88
90. ERET 090 - Endspiel L vs L,  Salai 2011   > 30s.
91. ERET 091 - Endspiel B vs B,  Shredder-Fritz (Var), CompGame 2007   Gelöst in 6.07s/25; Gelöst: 89
92. ERET 092 - Vergifteter Bauer,  Shredder-Ruffian, CompGame 2005   Gelöst in 1.40s/11; Gelöst: 90
93. ERET 093 - Endspiel T&S&B vs T&T&L,  Simkhovich 1923   > 30s.
94. ERET 094 - Damenopfer,  Sirobaba-Chaika, CorrGame 2011   > 30s.
95. ERET 095 - Damenopfer,  Smagin-Sahovic, Biel 1990   Gelöst in 0.58s/9; Gelöst: 91
96. ERET 096 - Endspiel D vs T&L&S,  Sochnev 2009   Gelöst in 3.60s/14; Gelöst: 92
97. ERET 097 - Endspiel L vs L,  Spassky-Barne, SanJuan 1974   Gelöst in 25.30s/26; Gelöst: 93
98. ERET 098 - Endspiel T vs T,  Stockfish-Komodo, CompGame 2015   Gelöst in 2.18s/15; Gelöst: 94
99. ERET 099 - Endspiel D&T vs D&T&S,  Stuovsky-Movsesian, EU-Ch 2001   Gelöst in 0.44s/8; Gelöst: 95
100. ERET 100 - Initiative,  Sutovsky-Agrest, Harplinge 1998   Gelöst in 0.41s/8; Gelöst: 96
101. ERET 101 - Entwicklung,  Tronhjem-Christensen, CorrGame 1984   Gelöst in 0.41s/7; Gelöst: 97
102. ERET 102 - Koenigsangriff,  Tumurhuyag-Uhlmann, Novi Sad 1990   Gelöst in 3.54s/14; Gelöst: 98
103. ERET 103 - Zugzwang,  VanEssen 2004   Gelöst in 5.93s/16; Gelöst: 99
104. ERET 104 - Entwicklung,  VanOosterom-Timmerman, CorrGame 1982   Gelöst in 0.76s/10; Gelöst: 100
105. ERET 105 - Starke Felder,  Vlasveld-Neagu, CorrGame 2014   Gelöst in 0.37s/8; Gelöst: 101
106. ERET 106 - Koenigssicherheit,  Wahls-Chandler (Var), NoviSad 1990   Gelöst in 0.50s/10; Gelöst: 102
107. ERET 107 - Verteidigung,  Wiacek-Kolarik, CorrGame 1999   Gelöst in 3.18s/13; Gelöst: 103
108. ERET 108 - Endspiel T&L vs T&L,  Winckelmann-Bastos, CorrGame 2003   Gelöst in 0.45s/9; Gelöst: 104
109. ERET 109 - Entlastung,  Xie-Bischoff, Baden-Baden 1992   Gelöst in 0.77s/11; Gelöst: 105
110. ERET 110 - Freibauer,  Ye-Lutz, Batumi 2001   Gelöst in 11.19s/16; Gelöst: 106
111. ERET 111 - Festung,  Zueger-Kindermann(Var), Muenchen 1989   Gelöst in 19.06s/10; Gelöst: 107

Ergebnis: 107 aus 111 = 96.3%. Durchschnittszeit = 3.51s / 12.26
Peter.
User avatar
Ozymandias
Posts: 1535
Joined: Sun Oct 25, 2009 2:30 am

Re: New net: The White Rose

Post by Ozymandias »

White Rose? Mr. Robot?
dkappe
Posts: 1631
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

Re: New net: The White Rose

Post by dkappe »

I had tested against DH-0.2 to make sure I was making progress. Tested against the newly released SF13. Shocker. Perhaps within 20 elo. Those final 30 elo will be hard.

[pgn]

[Event "?"]
[Site "?"]
[Date "2021.02.20"]
[Round "53"]
[White "sf13"]
[Black "thewhiterose"]
[Result "0-1"]
[ECO "C78"]
[GameDuration "00:02:08"]
[GameEndTime "2021-02-20T11:18:12.615 CST"]
[GameStartTime "2021-02-20T11:16:04.138 CST"]
[Opening "Ruy Lopez"]
[PlyCount "100"]
[TimeControl "60+0.6"]
[Variation "5.O-O"]

1. e4 {book} e5 {book} 2. Nf3 {book} Nc6 {book} 3. Bb5 {book} a6 {book}
4. Ba4 {book} Nf6 {book} 5. O-O {book} b5 {book} 6. Bb3 {book} Be7 {book}
7. Re1 {book} O-O {book} 8. c3 {book} d5 {book} 9. exd5 {book} Nxd5 {book}
10. Nxe5 {book} Nxe5 {book} 11. Rxe5 {book} c6 {book} 12. d4 {book} Bd6 {book}
13. Re1 {book} Qh4 {book} 14. g3 {book} Qh3 {book} 15. Re4 {0.00/23 3.1s}
g5 {0.00/24 4.9s} 16. Qe2 {+0.63/24 2.4s} f5 {+0.27/21 0.88s}
17. Bxd5+ {-0.25/30 14s} cxd5 {+0.32/22 1.5s} 18. Re6 {-0.24/26 0.77s}
Bxe6 {+0.40/23 1.7s} 19. Qxe6+ {-0.43/25 1.4s} Kg7 {+0.31/23 1.3s}
20. Bxg5 {-0.31/24 1.6s} Qh5 {+0.32/25 1.8s} 21. Bh4 {0.00/25 5.7s}
Qh6 {+0.37/20 1.2s} 22. Qxd5 {0.00/23 1.9s} f4 {+0.39/23 2.8s}
23. Na3 {0.00/23 1.0s} Rae8 {+0.39/21 1.5s} 24. Qb7+ {-0.31/25 2.1s}
Rf7 {+0.94/26 2.3s} 25. Qxa6 {-2.07/26 4.1s} Re2 {+1.92/22 1.4s}
26. Nxb5 {-2.90/24 3.1s} fxg3 {+2.29/24 1.7s} 27. Qxd6 {-3.97/29 15s}
gxf2+ {+2.52/21 1.3s} 28. Bxf2 {-4.13/22 0.79s} Rfxf2 {+2.75/23 1.6s}
29. Qxh6+ {-3.81/24 2.0s} Kxh6 {+3.11/24 1.3s} 30. b3 {-5.03/24 8.1s}
Rg2+ {+4.05/22 1.2s} 31. Kh1 {-5.03/17 0.18s} Rxh2+ {+4.48/21 1.3s}
32. Kg1 {-5.16/18 0.38s} Reg2+ {+4.63/20 1.3s} 33. Kf1 {-5.03/19 0.27s}
Rg5 {+4.97/22 3.4s} 34. Re1 {-5.60/21 2.9s} Rxb5 {+5.48/19 1.3s}
35. c4 {-5.55/18 0.47s} Rf5+ {+5.57/20 1.3s} 36. Kg1 {-5.23/18 0.17s}
Rxa2 {+6.12/21 1.3s} 37. Re6+ {-7.15/19 1.5s} Kg5 {+8.55/21 1.4s}
38. Re7 {-127.61/20 0.78s} Rff2 {+152.59/40 2.5s} 39. Rg7+ {-152.60/28 0.43s}
Kh6 {+152.60/39 0.77s} 40. Re7 {-152.61/24 0.074s} Rfd2 {+152.62/42 2.7s}
41. Re1 {-152.61/32 0.20s} Kg5 {+M19/45 0.83s} 42. Kf1 {-152.62/32 0.18s}
Kf4 {+M17/50 1.2s} 43. Re5 {-152.62/31 0.18s} Rd1+ {+M15/56 1.1s}
44. Re1 {-152.62/31 0.22s} Rxe1+ {+M13/62 1.1s} 45. Kxe1 {-M12/41 0.54s}
h5 {+M11/68 1.1s} 46. b4 {-M10/50 0.26s} h4 {+M9/127 1.1s}
47. Kf1 {-M8/207 0.27s} h3 {+M7/245 0.27s} 48. Kg1 {-M6/245 0.029s}
h2+ {+M5/245 0.009s} 49. Kh1 {-M4/245 0.007s} Kg3 {+M3/245 0.006s}
50. b5 {-M2/245 0.005s} Ra1# {+M1/245 0.004s, Black mates} 0-1

[/pgn]
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
matejst
Posts: 364
Joined: Mon May 14, 2007 8:20 pm
Full name: Boban Stanojević

Re: New net: The White Rose

Post by matejst »

Dietrich,

Could you be a bit more specific about this net (and the DH and Harmon nets). I guess you probably posted more detailed explanations somewhere, but I cannot find them.

I tried the Harmon net and I was very impressed by the way it plays -- if the aim was to have a "human" kind of chess, I felt it was it.
dkappe
Posts: 1631
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

Re: New net: The White Rose

Post by dkappe »

matejst wrote: Sat Feb 20, 2021 11:57 pm Dietrich,

Could you be a bit more specific about this net (and the DH and Harmon nets). I guess you probably posted more detailed explanations somewhere, but I cannot find them.

I tried the Harmon net and I was very impressed by the way it plays -- if the aim was to have a "human" kind of chess, I felt it was it.
The dark horse is essentially like the night nurse net — mcts/nn engine data — but night nurse is at 800 nodes while dark horse is at twice or more.

Harmon was an attempt to train on FIDE 2300+ games (shy of 1M games). My first attempt relied on only game outcomes, and while it definitely played an entertaining game, it did not understand the value of pieces. I augmented the data with qsearch on a naive material eval (I didn’t want to pollute the net with the HCE of a real engine) and trained with 80% outcome, 20% eval. This played a much more reasonable game.

I can’t judge if it is more “human.” It’s definitely weaker. Stefan Pohl had it at 3472 vs the 3733 of the strongest sf nets. https://www.sp-cc.de/files/nnue_quicktests.txt

I do wonder if this sort of net might be more useful for prepping against human opponents.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
dkappe
Posts: 1631
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

Re: New net: The White Rose

Post by dkappe »

Oh, the additional twist is adding dropouts to the neural net training, which really do seem to help the long training runs.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
matejst
Posts: 364
Joined: Mon May 14, 2007 8:20 pm
Full name: Boban Stanojević

Re: New net: The White Rose

Post by matejst »

I used it for analysis of openings and games, and we had a discussion about the net on Ed forum (one can easily find the thread).
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: New net: The White Rose

Post by carldaman »

dkappe wrote: Sat Feb 20, 2021 1:16 am I’ve approached the training with Dark Horse data a little differently. It’s a little under inflated, but not as much as DH-0.2.

https://www.patreon.com/posts/47771982

Note that this net is distributed under the Creative Commons attribution, no derivatives license. As such it cannot be embedded and distributed in a GPLv3 work or others with an incompatible license.
Thanks for your efforts, but The Baron 3.441, clearly rated below 3000 CCRL, can seemingly draw at will with White against recent Cfish + White Rose net, at rapid TCs + no book. A massive fail, as far as I'm concerned, but this deficiency is also shared with some of the latest SF NNUE nets.

I frequently like to test against much weaker engines and without a book, things that relatively very few others do. This can bring to light such aberrations. Of course, these overly drawish tendencies from the NNUE Black side is something that became apparent soon after NNUE originally came out. The absence of a contempt setting combined with risk-averse play from the Black side continues to be a general problem with NNUE.

I will test some of your other recent nets against The Baron, under the same conditions, and report back my impressions.
dkappe
Posts: 1631
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

Re: New net: The White Rose

Post by dkappe »

carldaman wrote: Sun Feb 21, 2021 7:33 am
dkappe wrote: Sat Feb 20, 2021 1:16 am I’ve approached the training with Dark Horse data a little differently. It’s a little under inflated, but not as much as DH-0.2.

https://www.patreon.com/posts/47771982

Note that this net is distributed under the Creative Commons attribution, no derivatives license. As such it cannot be embedded and distributed in a GPLv3 work or others with an incompatible license.
Thanks for your efforts, but The Baron 3.441, clearly rated below 3000 CCRL, can seemingly draw at will with White against recent Cfish + White Rose net, at rapid TCs + no book. A massive fail, as far as I'm concerned, but this deficiency is also shared with some of the latest SF NNUE nets.

I frequently like to test against much weaker engines and without a book, things that relatively very few others do. This can bring to light such aberrations. Of course, these overly drawish tendencies from the NNUE Black side is something that became apparent soon after NNUE originally came out. The absence of a contempt setting combined with risk-averse play from the Black side continues to be a general problem with NNUE.

I will test some of your other recent nets against The Baron, under the same conditions, and report back my impressions.
Interesting. Can we expect The Baron in the SuFi soon? :wink:
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".