Alpha Zero's 21. Bg5! revisited

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
zullil
Posts: 6442
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: Alpha Zero's 21. Bg5! revisited

Post by zullil » Tue Jul 21, 2020 10:00 am

Spliffjiffer wrote:
Mon Jul 20, 2020 8:22 pm
Raphexon wrote:
Mon Jul 20, 2020 2:46 pm
Sergio's net:
https://www.comp.nus.edu.sg/~sergio-v/nnue/

With halfkp 256x binary from 7-13:
Threads 1
Hash 16 mb
No TB

Didn't even need 25 million nodes to find it.
Absolutely incredible.
i think ive seen the future of SF now, thx very much for sharing Raph !
look eg what it does with this position:
[d]4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - 0 1
bm Qxe5
it says completely winning after just a couple of seconds...

or with this position:
[d]rnb2rk1/1pq1bppp/p3p3/2npP1P1/3N1P2/2NBB3/PPP4P/R2QK2R w KQ - 0 1
bm Bxh7
stunning, solved within a minute or so !

and im quite sure that this is just the beginning as well, hf, stay patient ;-)
Yes, amazing. The first is solved in seconds. The latter is solved in less than a second:

info depth 17 seldepth 29 multipv 1 score cp 117 nodes 15154873 nps 22720949 tbhits 0 time 667 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 e1c1 c5e4 h4h7 g8f8 d4b5 a6b5 c3e4 b8c6
info depth 18 seldepth 30 multipv 1 score cp 119 nodes 15617237 nps 22899174 tbhits 0 time 682 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 h4h7 g8f8 e1c1 b7b5 h3h6 b8c6 h2h4 c6e7 h4h5 g6h5 a2a3
info depth 19 seldepth 36 multipv 1 score cp 112 nodes 17103548 nps 23206985 tbhits 0 time 737 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 h4h7 g8f8 e1c1 b7b5 h3h6 b8c6 h2h4 c6d4 e3d4 b5b4 c3e2 b4b3
info depth 20 seldepth 39 multipv 1 score cp 102 nodes 19118166 nps 23544539 tbhits 0 time 812 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 h4h7 g8f8 e1c1 b7b5 a2a3 c8d7 h3h6 d7e8 c1b1 b8d7 h2h4 d7e5 f4e5 c7e5 d1e1 e5g3
info depth 21 seldepth 40 multipv 1 score cp 83 nodes 21212506 nps 23914888 tbhits 0 time 887 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 h4h7 g8f8 e1c1 b7b5 a2a3 c8b7 h3h6 b8c6 h2h4 c6d4 e3d4 c5e4 h4h5 g6h5
info depth 22 seldepth 46 multipv 1 score cp 108 nodes 23420399 nps 24119875 tbhits 0 time 971 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 e1c1 g8f8 h4h7 b7b5 a2a3 c8b7 b2b4 c5d7 d4e6 f7e6 h7g6 d8e8 f4f5 d7e5 g6h7 e6f5 e3c5 e8e7 h7f5 f8g8 f5h7 g8f8 h3e3
info depth 23 seldepth 45 multipv 1 score cp 108 nodes 27836923 nps 24700020 hashfull 4 tbhits 0 time 1127 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 e1c1 f8g7 g3h3 b7b5 a2a3 g8f8 h4h7 c5a4 c3a4 b5a4 h3h6 b8c6 h2h4 c8d7 h4h5 a8c8 d1d2 c6d4 e3d4 g6h5 g5g6 c7c4 h6h5
info depth 24 seldepth 39 multipv 1 score cp 175 nodes 31832281 nps 25283781 hashfull 4 tbhits 0 time 1259 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 h4h7 g8f8 e1c1 c5e4 d4b5 a6b5 c3e4 b8a6 e4f6 a6b4 c2c3 b4a2 c1d2 a2c3 b2c3
info depth 25 seldepth 46 multipv 1 score cp 104 nodes 39826568 nps 25711147 hashfull 4 tbhits 0 time 1549 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 e1c1 g8f8 h4h7 b7b6 h3h6 b8c6 h2h4 c6d4 e3d4 g7h6 g5h6 f8e7 h7g7 e7d7 b2b4 d7c6 h6h7 c5e4 c3e4 d5e4 h7h8q d8h8 g7h8 c6b7
info depth 26 seldepth 50 multipv 1 score cp 130 nodes 69508066 nps 26378772 hashfull 7 tbhits 0 time 2635 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 e1c1 f8g7 g3h3 g8f8 h4h7 b7b6 h3g3 b8c6 h2h4 c6e7 d1g1 e7f5 h4h5 f5g3 g1g3 g6h5 g5g6 c5e4 c3e4 d5e4 g6f7 c7f7 f4f5 e6f5
info depth 27 seldepth 46 multipv 1 score cp 156 nodes 76315460 nps 26507627 hashfull 8 tbhits 0 time 2879 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 h4h7 g8f8 e1c1 b7b5 h3h6 b8c6 c3e2 c5e4 h2h4 c8d7 h4h5 g7h6 g5h6 g6h5 h7g7 f8e7 d4c6 d7c6 f4f5 d8h8 h6h7 h8h7 f5f6 e7d8 g7h7 d8c8 c1b1 c8b7 d1g1
info depth 28 currmove h2h4 currmovenumber 2
info depth 28 currmove c3e2 currmovenumber 3
info depth 28 currmove d1g4 currmovenumber 4
info depth 28 currmove h1g1 currmovenumber 5
info depth 28 currmove a2a3 currmovenumber 6
info depth 28 currmove d3e2 currmovenumber 7
info depth 28 currmove h2h3 currmovenumber 8
info depth 28 currmove e1f2 currmovenumber 9
info depth 28 currmove e1f1 currmovenumber 10
info depth 28 currmove g5g6 currmovenumber 11
info depth 28 currmove d4c6 currmovenumber 12
info depth 28 currmove d4f5 currmovenumber 13
info depth 28 currmove f4f5 currmovenumber 14
info depth 28 currmove d4b5 currmovenumber 15
info depth 28 currmove d1e2 currmovenumber 16
info depth 28 currmove e1e2 currmovenumber 17
info depth 28 currmove c3e4 currmovenumber 18
info depth 28 currmove b2b3 currmovenumber 19
info depth 28 currmove h1f1 currmovenumber 20
info depth 28 currmove e1g1 currmovenumber 21
info depth 28 currmove a1c1 currmovenumber 22
info depth 28 currmove e3d2 currmovenumber 23
info depth 28 currmove d1c1 currmovenumber 24
info depth 28 currmove d1h5 currmovenumber 25
info depth 28 currmove b2b4 currmovenumber 26
info depth 28 currmove e3c1 currmovenumber 27
info depth 28 currmove c3b1 currmovenumber 28
info depth 28 currmove e3g1 currmovenumber 29
info depth 28 currmove c3b5 currmovenumber 30
info depth 28 currmove e3f2 currmovenumber 31
info depth 28 currmove d4b3 currmovenumber 32
info depth 28 currmove d4e2 currmovenumber 33
info depth 28 currmove d1d2 currmovenumber 34
info depth 28 currmove d3b5 currmovenumber 35
info depth 28 currmove d1b1 currmovenumber 36
info depth 28 currmove e1d2 currmovenumber 37
info depth 28 currmove a1b1 currmovenumber 38
info depth 28 currmove d3f1 currmovenumber 39
info depth 28 currmove d3f5 currmovenumber 40
info depth 28 currmove d3g6 currmovenumber 41
info depth 28 currmove d3c4 currmovenumber 42
info depth 28 currmove d3e4 currmovenumber 43
info depth 28 currmove d1f3 currmovenumber 44
info depth 28 currmove a2a4 currmovenumber 45
info depth 28 currmove c3a4 currmovenumber 46
info depth 28 currmove d4f3 currmovenumber 47
info depth 28 currmove c3d5 currmovenumber 48
info depth 28 currmove d4e6 currmovenumber 49
info depth 28 currmove d3a6 currmovenumber 50
info depth 28 seldepth 51 multipv 1 score cp 155 nodes 81204163 nps 26606868 hashfull 8 tbhits 0 time 3052 pv d3h7 g8h7 d1h5 h7g8 h1g1 g7g6 h5h4 f8d8 g1g3 e7f8 g3h3 f8g7 e1c1 g8f8 h4h7 b7b6 h3g3 b8c6 h2h4 c5e4 c3e4 d5e4 h4h5 c6d4 e3d4 g6h5 g5g6 d8d4 d1d4 c8b7 h7h5 b7d5 g6f7 c7f7 h5g5 a8c8 d4d1 e4e3 g3e3 f8e8

User avatar
MikeB
Posts: 4360
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Alpha Zero's 21. Bg5! revisited

Post by MikeB » Tue Jul 21, 2020 12:51 pm

Spliffjiffer wrote:
Mon Jul 20, 2020 8:22 pm
Raphexon wrote:
Mon Jul 20, 2020 2:46 pm
Sergio's net:
https://www.comp.nus.edu.sg/~sergio-v/nnue/

With halfkp 256x binary from 7-13:
Threads 1
Hash 16 mb
No TB

Didn't even need 25 million nodes to find it.
Absolutely incredible.
i think ive seen the future of SF now, thx very much for sharing Raph !
look eg what it does with this position:
[d]4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - 0 1
bm Qxe5
it says completely winning after just a couple of seconds...

or with this position:
[d]rnb2rk1/1pq1bppp/p3p3/2npP1P1/3N1P2/2NBB3/PPP4P/R2QK2R w KQ - 0 1
bm Bxh7
stunning, solved within a minute or so !

and im quite sure that this is just the beginning as well, hf, stay patient ;-)
hmmm I must be getting jaded ...took a little longer than expected...

mover viewpoint fewer / Multi-PV setting = 1 / more
exclude: none best +tail
dep score nodes time (not shown: tbhits knps seldept
48 +2.80 2.31G 0:32.79 Qxe5 fxe5 Rf1 Qe7 Bd1 Rc4 Bc2 Qg7 Bxg7 Kxg7 Rf7+ Kh6 Rxa7 b5 e7 b4 Ba4 bxc3 bxc3 Rxc3 e8=R Rxe8 Bxe8 Rc4 Kf2 Kg5 Rxh7 Rxe4 a4 Rb4 Ke3 Rb2 Ke4 Re2+ Kf3 Re1 h4+ Kf5 Bd7+ Kf6 Bb5 Ra1 Ke4 Ra3 a5 g5 Rh6+ Kg7 hxg5 Rxa5 Bc4 Ra1 Rc6 Rg1
47 23 seconds
47 +2.39 997.8M 0:15.05 Qxe5 fxe5 Rf1 Qe7 Bd1 Rc4 Bc2 Qg7 Bxg7 Kxg7 Rf7+ Kh6 Rxa7 Re8 e7 Kg7 Ba4 Rxa4 Rxa4 Rxe7 Ra6 b5 Rb6 Ra7 a3 Ra4 Rxb5 Rxe4 b4 Re1+ Kf2 Rc1 Rc5 Rc2+ Kg3 Kf6 a4 Ra2 Rc6+ Kf5 a5 Ke4 c4 Ra3+ Kg4 Kd3 c5 e4 a6 e3 Kf3 e2 Rd6+ Kc4+ Kxe2 K
46 13 seconds
46 3 seconds
46 0.00 197.7M 0:03.12 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 b3 Qe8 Rf2 Qa8 h3 Qd8 Kh2 a5 Rf7 Qe8 Rg7+ Kf8 Rb7+ Kg8
45 0.00 188.1M 0:02.97 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 h4 Qe8 Rf2 Qe7 b3 Qd8 c4 Qe7 Rf3 bxc4 bxc4 a5 Rf2 Qa3 Rf3
44 0.00 158.0M 0:02.50 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 Kg1 Qd1+ Kf2 Qc2+ Kf3 Qd1+
43 0.00 132.9M 0:02.09 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 a4 bxa4 Kh2 g5 Kg3 Qd3+ Rf3 Qd8
42 0.00 110.4M 0:01.74 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 a4 bxa4 Ra7 a5 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
41 0.00 91.6M 0:01.43 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 a4 bxa4 Ra7 a5 Rf7 g5 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
40 0.00 86.9M 0:01.35 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 Kf1 Qd1+ Kf2 Qc2+ Kg3 Qd3+ Kh2 Qc2+ Kg1
39 0.00 73.4M 0:01.14 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 c4 bxc4 Rb7 Qe8 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
38 0.00 63.3M 0:00.97 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 c4 bxc4 Rb7 g5 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
37 0.00 51.8M 0:00.80 Qxe5 fxe5 Rf1 Rc8 Ba6 b5 Bb7 Rd8 Bd5 Rxd5 exd5 Qe7 Kg2 Qc5 Rf7 Qxd5+ Kf2 Qd8 Kg2
36 0.00 39.7M 0:00.61 Qxe5 fxe5 Rf1 Rc8 Ba6 b5 Bb7 Rd8 Kg2 Qe7 Bd5 Rxd5 exd5 e4 d6 Qxd6 Rf7 Qd8 Rxa7 e3 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
35 0.00 35.7M 0:00.55 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 c4 bxc4 Kh2 Qe8 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
34 0.00 32.3M 0:00.50 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 a4 bxa4 h4 Qe8 Rg7+ Kf8 Rb7+ Kg8
33 0.00 29.5M 0:00.45 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 h3 b5 b4 Qe8 Rg7+ Kf8 Rb7+ Kg8
32 0.00 26.4M 0:00.41 Qxe5 fxe5 Rf1 Rc8 Ba6 Qd8 Bxc8 Qxc8 Rf7 Qd8 Kg2 a6 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8 Rh7+
31 0.00 21.3M 0:00.33 Qxe5 fxe5 Rf1 Rc8 Ba6 b5 Bb7 Rd8 Kg2 Qe7 Bd5 Rxd5 exd5 e4 d6 Qxd6 Rf7 Qd8 Rg7+ Kf8 Rxh7+ Kg8 Rg7+ Kf8
30 0.00 13.3M 0:00.21 Qa3 Rxe6 Qxa7 Qe7 Qa8+ Qe8




but this one took a little less time than expected ...
mover viewpoint fewer / Multi-PV setting = 1 / more
exclude: none best +tail
dep score nodes time (not shown: tbhits knps seldep)
38 50 seconds
38 +1.22 2.67G 0:48.99 Bxh7+ Kxh7 Qh5+ Kg8 O-O-O Rd8 g6 fxg6 Qxg6 Bf8 Rdg1 Nc6 Rg3 Qf7 Qh6 Nxd4 Bxd4 Ne4 Nxe4 dxe4 Bb6 Rd7 Rhg1 Qf5 Bc5 e3 Bxf8 Qxf8 Rxe3 Rd4 Rxg7+ Qxg7 Rg3 Rd7 Qxe6+ Kf8 Rxg7 Rd1+ Kxd1 Bxe6 Rxb7 Rc8 c3 Rc6 Rb8+ Ke7 Kd2 Bxa2 Ke3 Bb1 h4
37 40 seconds
37 30 seconds
37 +0.55 1.66G 0:30.00 Bxh7+
37 +0.61 1.57G 0:28.34 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b6 Rf3 Nc6 h4 Ne7 Nce2 Ne4 Ng3 Nf5 Ngxf5 exf5 h5 gxh5 Rh3 Re8 Rxh5 Be6 g6 f6 Nxe6+ Rxe6 Bd4 Rc8 c3 Ng3 exf6 Rxf6 Bxf6 Bxf6 Qxc7 Rxc7 Rh7 Rg7 Rxg7 Kxg7
36 +0.59 1.56G 0:28.09 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b6 Rf3 Nc6 h4 Ne7 Nce2 Ne4 Rg1 Nf5 h5 gxh5 Kb1 a5 Nxf5 exf5 Nd4 a4 a3 b5 g6 f6 Nxb5 Qe7 Nd4 Ra6 Qxh5 Re8 c3 fxe5 fxe5 Qxe5 Bf4 Qf6 Ka1 Ke7 Bh6 Kd8
35 20 seconds
35 +1.13 933.4M 0:16.61 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 O-O-O Kf8 Qh7 b6 Rf3 Nc6 h4 Ne7 Nce2 Ne4 Ng3 Nxg3 Rxg3 Bxe5 fxe5 Qxe5 Ne2 Qg7 Qxg7+ Kxg7 Bxb6 Re8 Bd4+ Kf8 Be5 Bd7 Nf4 Nf5 Rh3 Ke7 h5 gxh5 Rxh5 Rg8 g6 Be8
34 +0.93 623.7M 0:11.09 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b6 Rg3 Nc6 h4 Ne7 Nce2 Ra7 Rg2 Ne4 Rdg1 Rb7 Kb1 Bxe5 fxe5 Qxe5 Nf3 Qg7 Qxg7+ Kxg7 Bd4+ Kf8 Nf4 e5 Bxe5 Bf5 Nd4 Re8 Ka1 Bd7 Bf6 Rc8 h5 gxh5 Nxh5 Ng6
33 10 seconds
33 +0.79 563.9M 0:10.00 Bxh7+
33 +0.85 530.6M 0:09.40 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 O-O-O b6 Qh7+ Kf8 Rg3 Nc6 h4 Ne7 h5 gxh5 g6 fxg6 Rdg1 Ra7 Rxg6 Nxg6 Rxg6 Qf7 Qxh5 Rc7 f5 exf5 Bh6 Bxh6+ Qxh6+ Ke8 Qh8+ Ke7 Qh4+ Ke8 Rh6 Re7 Nc6 Be6 Rh8+ Kd7 Nxd8 Qg7 Nxe6 Kxe6
32 +0.89 502.1M 0:08.90 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b6 Rg3 Nc6 h4 Ne7 h5 gxh5 g6 fxg6 Rdg1 Ra7 Rxg6 Nxg6 Rxg6 Qf7 Qxh5 Re8 f5 exf5 Bh6 Qg8 Nxf5 Bxf5 Bxg7+ Rxg7 Qxf5+ Qf7 Rf6 Rg1+ Nd1 Rg7 Rxf7+ Rxf7 Qh5 d4 Qh6+ Kg8
31 +0.97 415.2M 0:07.37 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 O-O-O b5 Qh7+ Kf8 a3 Nbd7 Rf3 Nb6 h4 Nc4 h5 gxh5 Bf2 Ne4 Be1 a5 g6 fxg6 Bh4 Nxc3 Rg1 Na2+ Kd1 Qa7 Bf2 Qf7 Rxg6 Qg8 Qxh5 Qh8 Qg5 Re8 Rg3 Ra7 Nc6 Rd7 Bc5+ Kg8
30 +1.01 373.0M 0:06.69 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 O-O-O Kf8 Qh7 b5 a3 Bb7 Rh6 Nc6 f5 exf5 Nxf5 Nb3+ cxb3 gxf5 Bc5+ Ne7 Bd4 Bxh6 Qh8+ Ng8 gxh6 Ke7 h7 Qc6 Be3 Kd7 Rxd5+ Qxd5 Nxd5 Bxd5 Bg5 Rdc8+ Kd2 Be6 hxg8=Q Rxg8 Qf6 Rh8 h4 Rag
29 +0.91 330.2M 0:05.93 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 O-O-O Kf8 Qh7 b5 a3 Bb7 Rh6 Nc6 f5 exf5 Nxf5 Nb3+ cxb3 gxf5 Bc5+ Ne7 Bd4 Bxh6 Qh8+ Ng8 gxh6 Ke7 Re1 Kd7 Qg7 Ne7 Qxf7 Rf8 e6+ Kc6
28 +0.82 226.2M 0:04.11 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 O-O-O Kf8 Qh7 b5 Nce2 Ne4 Rf3 Nc6 h4 Ne7 Ng3 Nf5 Ngxf5 exf5 Kb1 a5 Rh1 b4 h5 gxh5 Rxh5 Ra6 Rh6 Rxh6 gxh6
27 +0.95 132.5M 0:02.42 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b6 Rg3 Nc6 h4 Ne4 Nxe4 dxe4 h5 Nxd4 Bxd4 gxh5 g6 Rxd4 Rxd4 Bb7 Kb1 e3 Rxe3 fxg6 Qxg6 Rd8 Rdd3 Bd5 Qxh5 Qf7 Qg5 Ke8 a3
26 +0.92 124.2M 0:02.29 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O Nbd7 Kb1 Nb6 Nb3 d4 Bxd4 Rxd4 Nxd4 Bd7 Rg3 Nd5 Nxd5 exd5 h4 Ne4 Rgg1 Rc8 h5 gxh5 Qxh5 Ke7 g6
25 +0.95 94.8M 0:01.75 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b5 Nce2 Nbd7 Rh6 Nb6 h4 Nc4 Bf2 Ne4 Be1 a5 h5 Nxe5 Bh4 Ng4 hxg6
24 +0.69 81.8M 0:01.53 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Rd8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 O-O-O b5 Nce2 Ne4 Rh6 Nc6 h4 Bxh6 gxh6 Ke7 h5 gxh5 Qg7 Bd7 Nxc6+ Bxc6 f5 exf5 h7 Kd7 Nd4 Kc8 h8=Q Rxh8
23 +0.49 68.1M 0:01.28 Bxh7+ Kxh7 Qh5+ Kg8 Rg1 g6 Qh4 Re8 Rg3 Bf8 Rh3 Bg7 Qh7+ Kf8 Rh6 Nbd7 O-O-O Nxe5 fxe5 Qxe5 Re1 Ne4 Nf3 Qf5 Nxe4 dxe4
22 0.00 44.1M 0:00.83 h4 Bd7 h5 Nc6 Be2 Rac8 Qd2 b5 a3 Nb7 Bd3 Nc5
21 0.00 24.8M 0:00.46 h4 Bd7 h5 Nc6 Be2 b5 a3 Rac8 Qd2 Nb7 Bd3 Nc5
Image

Spliffjiffer
Posts: 248
Joined: Thu Aug 02, 2012 5:48 pm
Location: Germany

Re: Alpha Zero's 21. Bg5! revisited

Post by Spliffjiffer » Tue Jul 21, 2020 1:31 pm

hi Louis, hi Mike :-)
very interesting times hehe...this projekt seems also very strong with fortresses at first glance...although i observe that "pure" SF is here and there significantly faster (on my old APU-machine) in various positions, NNue shines bright, for example 2 fortress-testpositions we all know about:
Brzozka-Bronstein, 1963:
[d]1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - 0 1
bm Rxb3
a matter of seconds for NNue !

or this one:
[d]3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P1K2P/PPB5/8 w - - 0 1
bm Ba4
i dont know if it needs milli- or microseconds for this :-)
ok, it doesnt show 0.00 but sth like -7.5 but it solves it !

[tested with Sergio's 20200721-1907.bin]

ps: gl with your new HoneyNN Mike, i didnt miss any Rx release so far :-)
Wahrheiten sind Illusionen von denen wir aber vergessen haben dass sie welche sind.

User avatar
MikeB
Posts: 4360
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Alpha Zero's 21. Bg5! revisited

Post by MikeB » Tue Jul 21, 2020 2:16 pm

Spliffjiffer wrote:
Tue Jul 21, 2020 1:31 pm
hi Louis, hi Mike :-)
very interesting times hehe...this projekt seems also very strong with fortresses at first glance...although i observe that "pure" SF is here and there significantly faster (on my old APU-machine) in various positions, NNue shines bright, for example 2 fortress-testpositions we all know about:
Brzozka-Bronstein, 1963:
[d]1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - 0 1
bm Rxb3
a matter of seconds for NNue !

or this one:
[d]3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P1K2P/PPB5/8 w - - 0 1
bm Ba4
i dont know if it needs milli- or microseconds for this :-)
ok, it doesnt show 0.00 but sth like -7.5 but it solves it !

[tested with Sergio's 20200721-1907.bin]

ps: gl with your new HoneyNN Mike, i didnt miss any Rx release so far :-)
Good stuff, the latest Sergio net you referenced is about 10 Elo stronger than the original Sergio net in self play testng

Score of Stockfish-XI-NN-20200721-1907 vs Stockfish-XI-NN-sergio: 347 - 292 - 1361 [0.514]
Elo difference: 9.6 +/- 8.6, LOS: 98.5 %, DrawRatio: 68.0 %

2000 of 2000 games finished.
Image

User avatar
MikeB
Posts: 4360
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Alpha Zero's 21. Bg5! revisited

Post by MikeB » Tue Jul 21, 2020 2:21 pm

Spliffjiffer wrote:
Tue Jul 21, 2020 1:31 pm
hi Louis, hi Mike :-)
very interesting times hehe...this projekt seems also very strong with fortresses at first glance...although i observe that "pure" SF is here and there significantly faster (on my old APU-machine) in various positions, NNue shines bright, for example 2 fortress-testpositions we all know about:
Brzozka-Bronstein, 1963:
[d]1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - 0 1
bm Rxb3
a matter of seconds for NNue !

or this one:
[d]3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P1K2P/PPB5/8 w - - 0 1
bm Ba4
i dont know if it needs milli- or microseconds for this :-)
ok, it doesnt show 0.00 but sth like -7.5 but it solves it !

[tested with Sergio's 20200721-1907.bin]

ps: gl with your new HoneyNN Mike, i didnt miss any Rx release so far :-)
Black-Diamond-NN gets the zero score ( same net as you referenced)

Code: Select all

                                     
dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 46	20 seconds
 46	  0.00 	1.20G	0:15.91	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 45	  0.00 	1.01G	0:13.16	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 44	  0.00 	846.4M	0:10.89	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 43	10 seconds
 43	  0.00 	654.6M	0:08.28	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 42	  0.00 	471.1M	0:05.93	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 41	  0.00 	372.2M	0:04.66	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 40	  0.00 	329.4M	0:04.12	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 39	  0.00 	263.9M	0:03.29	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 38	  0.00 	187.3M	0:02.27	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 37	  0.00 	151.9M	0:01.83	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 36	  0.00 	126.0M	0:01.52	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 35	  0.00 	105.3M	0:01.26	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 34	  0.00 	84.4M  	0:00.98	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 33	  0.00 	64.0M  	0:00.75	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 32	  0.00 	56.0M  	0:00.66	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 31	  0.00 	42.0M  	0:00.49	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 30	  0.00 	36.6M  	0:00.43	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 29	  0.00 	32.2M  	0:00.38	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 28	  0.00 	27.4M  	0:00.32	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 27	  0.00 	21.6M  	0:00.26	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 26	  0.00 	18.6M  	0:00.22	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 25	  0.00 	16.4M  	0:00.20	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 24	  0.00 	14.1M  	0:00.17	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 23	  0.00 	11.3M  	0:00.14	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 22	  0.00 	9.64M  	0:00.12	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 21	  0.00 	8.67M  	0:00.11	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 20	  0.00 	7.77M  	0:00.10	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Bxd8 Kd3 Ra4 Ke3
 19	  0.00 	7.08M  	0:00.09	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Ke8 Ke4 Kf8 Kd3 Rc7 Ke4
 18	  0.00 	6.55M  	0:00.09	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Ke8 Ke4 Kf8 Kd3 Rc7 Ke4
 17	  0.00 	6.09M  	0:00.08	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Ke8 Ke4 Kf8 Kd3 Rc7 Ke4
 16	  0.00 	5.95M  	0:00.08	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Ke8 Ke4 Kf8 Kd3 Rc7 Ke4
 15	  0.00 	5.90M  	0:00.08	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Ke8 Ke4 Kf8 Kd3 Rc7 Ke4
 14	  0.00 	5.86M  	0:00.08	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Ke8 Ke4 Kf8 Kd3 Rc7 Ke4
 13	 -1.06 	4.34M  	0:00.06	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Ke8 f5 Rc6 Bxa5 Rc8 Bb6 Kf8 Ke2 Rbc7 Bxc7
 12	  0.00 	3.42M  	0:00.05	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Ke8 f5 Rb5 Ke4 Kxd8 Ke3 Ra8 Ke4
 11	 -2.48 	3.21M  	0:00.05	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Ra8 Kd3 Rb5 Ke4
 10	 -4.39 	2.59M  	0:00.04	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kc8 f5 Kxd8 Kd3 Rab6 Ke4
  9	 -4.54 	2.37M  	0:00.03	Ba4+ Kxa4 b3+ Kb5 c4+ Kc6 d5+ Kd7 e6+ Kxd8 f5 Rb5 cxb5 Ra8
  8	 -4.35 	1.82M  	0:00.03	c4+ Kxc4 Bd3+ Kd5 Be4+ Ke6 Bxb7 Ra7 d5+ Kd7 exf6 exf6
  7	 -2.40 	879086	0:00.02	c4+ Kxc4 Bd3+ Kd5 Be4+ Ke6 Bxb7 Ra7 f5+ Kd7
  6	 -2.32 	330182	0:00.01	c4+ Kxc4 Bd3+ Kd5 Be4+ Ke6 f5+ Kd7 Bxb7
  5	 -2.23 	314574	0:00.01	c4+ Kxc4 Bd3+ Kd5 Be4+ Ke6 f5+ Kd7 Bxb7
  4	 -2.14 	294539	0:00.01	c4+ Kxc4 Bd3+ Kd5 Be4+ Ke6
  3	 -2.14 	288948	0:00.01	c4+ Kxc4 Bd3+ Kd5 Be4+ Ke6
  2	 -3.24 	267161	0:00.01	c4+ Kxc4 Bd3+ Kd5 Bxa6
  1	 -4.10 	136758	0:00.01	Bd3+ c4
  0	#
Image

User avatar
mclane
Posts: 17758
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: Alpha Zero's 21. Bg5! revisited

Post by mclane » Thu Jul 23, 2020 10:42 am

chrisw wrote:
Mon Jul 20, 2020 10:48 pm
zullil wrote:
Mon Jul 20, 2020 10:20 pm
chrisw wrote:
Mon Jul 20, 2020 6:42 pm

Ah, ok. Thanks. Bit tricky on a iPhone in that format, can’t actually see half the move list and so on.
Well, apart from beating the mighty god Stockfish, this wasn’t a particularly fascinating game. You may consider it win at move 15, but it still took until plycount 100 to grind it out into an ending, and as for the other games, draws with huge ply counts.

So contrary to Chris’s statistics with Laskos own set of 100 games and MikeB set of 1000 games, indicating very long tedious games with high ply counts and zero fireworks, you have one anecdote (in six tediums). Ding dong, round two. So far, not impressive, apart from the results of course, although there’s not really enough of those either, bean counter versus net trained on bean counter, all very efficient and mechanical, but, at risk of activating TC, I would say this is not really chess in the sense that AZ, also Leela, have shown.
Twenty more games between Stockfish-nnue and Stockfish-dev. Three wins by Stockfish-nnue, none by Stockfish-dev. 20 threads, 4 GB Hash, 6-man tables:
I prefer this sort of thing ...

Nb5 was nice. 10...Nbc6 was of Course „suboptimal“

But better this kind of chess then 1000 boring games between Stockfish and Komodo.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....

zullil
Posts: 6442
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: Alpha Zero's 21. Bg5! revisited

Post by zullil » Thu Jul 23, 2020 1:21 pm

mclane wrote:
Thu Jul 23, 2020 10:42 am
chrisw wrote:
Mon Jul 20, 2020 10:48 pm
zullil wrote:
Mon Jul 20, 2020 10:20 pm
chrisw wrote:
Mon Jul 20, 2020 6:42 pm

Ah, ok. Thanks. Bit tricky on a iPhone in that format, can’t actually see half the move list and so on.
Well, apart from beating the mighty god Stockfish, this wasn’t a particularly fascinating game. You may consider it win at move 15, but it still took until plycount 100 to grind it out into an ending, and as for the other games, draws with huge ply counts.

So contrary to Chris’s statistics with Laskos own set of 100 games and MikeB set of 1000 games, indicating very long tedious games with high ply counts and zero fireworks, you have one anecdote (in six tediums). Ding dong, round two. So far, not impressive, apart from the results of course, although there’s not really enough of those either, bean counter versus net trained on bean counter, all very efficient and mechanical, but, at risk of activating TC, I would say this is not really chess in the sense that AZ, also Leela, have shown.
Twenty more games between Stockfish-nnue and Stockfish-dev. Three wins by Stockfish-nnue, none by Stockfish-dev. 20 threads, 4 GB Hash, 6-man tables:
I prefer this sort of thing ...

Nb5 was nice. 10...Nbc6 was of Course „suboptimal“

But better this kind of chess then 1000 boring games between Stockfish and Komodo.
Well, if you want "exciting" games (i.e., games full of errors), there are many human "masterpieces" to study. :wink:

User avatar
mclane
Posts: 17758
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: Alpha Zero's 21. Bg5! revisited

Post by mclane » Thu Jul 23, 2020 1:34 pm

I prefer forward pruning engines.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....

User avatar
MikeB
Posts: 4360
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Alpha Zero's 21. Bg5! revisited

Post by MikeB » Sat Aug 08, 2020 3:26 pm

zullil wrote:
Mon Jul 20, 2020 11:37 am


Stockfish-nnue-dev shines:

info depth 30 seldepth 55 multipv 1 score cp 152 lowerbound nodes 343277338 nps 19054026 hashfull 39 tbhits 0 time 18016 pv c1g5
info depth 27 currmove c1g5 currmovenumber 1
info depth 30 seldepth 55 multipv 1 score cp 171 lowerbound nodes 344766231 nps 19062602 hashfull 39 tbhits 0 time 18086 pv c1g5
info depth 26 currmove c1g5 currmovenumber 1
info depth 26 currmove b2b4 currmovenumber 2
[snipped]
info depth 26 currmove g4d7 currmovenumber 39
info depth 26 currmove g4g7 currmovenumber 40
info depth 30 seldepth 55 multipv 1 score cp 171 nodes 345359193 nps 19063766 hashfull 39 tbhits 0 time 18116 pv c1g5 f7f5 g4f4 b7c5 f4d6 h8g8 g5e7 c5e4 g2e4 f5e4 e7f8 a6d3 f2f3 b8a6 f3e4 a8f8 d6d3 a6c5 d3c4 c5e6 e1f1 g7g6 h5g6 h7g6 g1g2 g6g4 f1f8 g8f8 c4f1 f8e7 a1e1 h6h5 f1f3 e6g5 f3f5 g5f7 a2a3 g4f5 e4f5 e7f6 e1h1 f6f5 h1h5 f5e6 h5h7

and later

info depth 36 seldepth 59 multipv 1 score cp 288 nodes 791092688 nps 20170644 hashfull 80 tbhits 0 time 39220 pv c1g5 f7f5 g4f4 b7c5 g5e7 c5e6 f4d6 f5e4 e7f8 a6d3 f8e7 h7f5 a1d1 a7a5 d1d2 a5a4 e1e3 h8g8 a2a3 f5h5 g2e4 d3e4 e3e4 h5g6 d2e2 g8h7 e2e1 g6f5 e4e5 f5f7 f2f4 f7g6 g1f2 c6c5 f4f5 g6f7 f5f6 f7g6 f6g7 e6g7 e5d5 g6f7 f2g2 b8c6 e7h4 g7f5 d6d7 f5h4 g3h4 f7d7 d5d7 h7g8


it did shine, I only see Bluefish Xir4 now getting this in less than a minute, your mileage may vay
$ Blu*

Code: Select all

     ######                                       #    # ~~~ ~~~ ~~~ ~~~ ~~ ~~~ ~~~ ~~~ #     # ###
     #     # #      #    # ###### ###### #  ####  #    # ~~~ ~~~ ~~ /''*._ ~~~~ ~~ ~ ~~~ #   #   #
     #     # #      #    # #      #      # #      #    # ~~~ ~~ .-*'`    `*-.._.-'\ ~ ~~~ # #    #
     ######  #      #    # #####  #####  #  ####  ###### ~~~ ~ < * ) ) )   ,     ( ~~~ ~~~ #     #
     #     # #      #    # #      #      #      # #    # ~~~ ~~ `*-._`._(__.--*'`.) ~ ~~~ # #    #
     #     # #      #    # #      #      # #    # #    # ~~~  ~~~ ~~ ~~~ ~~~ ~~ ~~~~ ~~~ #   #   #
     ######  ######  ####  ###### #      #  ####  #    # ~~~ ~~ ~~~ ~~ ~~~ ~~~ ~~ ~ ~~~ #     # ###


    Bluefish XIr4  by Stockfish Developers as noted in the AUTHORS file
p f rn3r1k/pn1p1ppq/bpp4p/7P/4N1Q1/6P1/PP3PB1/R1B1R1K1 w - - 3 21
s t 60
info string Hash Table: Windows Large Pages, 256 Mb
Confirmation: Threads set to 60
s h 40960
info string Hash Table: Windows Large Pages, 40960 Mb
Confirmation: Hash set to 40960 Mb
s nn true
Confirmation: UseNN set to true
info string: NN evaluation using eval.bin enabled.
go mt 60
info string NNUE evaluation using eval.bin enabled.
info depth 1 seldepth 1 multipv 1 score cp -241 nodes 2832 nps 1416000 tbhits 0 time 2 pv g4g7 h8g7
info depth 2 seldepth 2 multipv 1 score cp -319 nodes 15786 nps 5262000 tbhits 0 time 3 pv g4g7 h7g7
info depth 3 seldepth 3 multipv 1 score cp -324 nodes 24301 nps 8100333 tbhits 0 time 3 pv g4g7 h7g7 c1f4
info depth 4 seldepth 4 multipv 1 score cp -305 nodes 27218 nps 9072666 tbhits 0 time 3 pv g4g7 h7g7 c1f4 d7d5
info depth 5 seldepth 5 multipv 1 score cp -251 nodes 32199 nps 10733000 tbhits 0 time 3 pv g4g7 h7g7 c1d2 d7d5 d2c3
info depth 6 seldepth 6 multipv 1 score cp -60 nodes 39468 nps 13156000 tbhits 0 time 3 pv e4g5 h6g5
info depth 7 seldepth 7 multipv 1 score cp -71 nodes 55058 nps 18352666 tbhits 0 time 3 pv e4g5 h6g5 c1g5 f7f5
info depth 8 seldepth 9 multipv 1 score cp 11 nodes 69250 nps 23083333 tbhits 0 time 3 pv e4g5 h6g5 c1g5 f7f5 g4b4 d7d6 g5e7
info depth 9 seldepth 9 multipv 1 score cp 94 nodes 102748 nps 25687000 tbhits 0 time 4 pv g4f4 d7d5 e4d6 b7d6 f4d6 h8g8 e1e7
info depth 10 seldepth 17 multipv 1 score cp 66 nodes 250459 nps 41743166 tbhits 0 time 6 pv g4f4 d7d5 e4d6 b7d6 f4d6 h8g8 b2b3 h7f5 c1a3 b8d7 d6c6 f8e8 c6d5
info depth 11 seldepth 16 multipv 1 score cp 111 nodes 453476 nps 45347600 tbhits 0 time 10 pv b2b4 d7d5 c1b2 d5e4 g2e4 f7f5 g4f4 b8d7 e4c6
info depth 12 seldepth 18 multipv 1 score cp 67 nodes 993461 nps 49673050 tbhits 0 time 20 pv b2b4 d7d5 c1b2 d5e4 g2e4 h7g8 a1d1 a6c4 g4f4 f7f6 e4f5 c4a2 e1e7
info depth 13 seldepth 18 multipv 1 score cp 84 nodes 1248005 nps 49920200 tbhits 0 time 25 pv b2b4 d7d5 e4d2 b7d6 c1b2 d6c4 d2c4 a6c4 a2a3 c6c5 a1d1 b8c6 g2d5 c4d5
info depth 14 seldepth 24 multipv 1 score cp 71 nodes 1880420 nps 50822162 tbhits 0 time 37 pv g4f4 d7d5 e4d6 b8d7 d6f7 h8g8 e1e7 b7c5 b2b4 c5d3 f7h6 h7h6 f4h6 g7h6 e7d7 d3c1 a1c1 a6b5
info depth 15 seldepth 23 multipv 1 score cp 108 nodes 2213389 nps 51474162 tbhits 0 time 43 pv g4f4 d7d5 e4d6 b7d6 f4d6 h8g8 b2b3 h7f5 c1a3 b8d7 d6c6
info depth 16 seldepth 24 multipv 1 score cp 88 nodes 9277608 nps 53939581 tbhits 0 time 172 pv b2b4 d7d5 e4d2 b7d8 d2f3 d8e6 f3e5 h7c2 a2a4 a6c8 b4b5 b8d7 e5c6 d7c5 g4d1 c2d1
info depth 17 seldepth 27 multipv 1 score cp 64 nodes 11177350 nps 53996859 tbhits 0 time 207 pv b2b4 d7d5 e4d2 h7c2 d2f3 c2c4 c1f4 b7d6 a2a3 a6c8 g4h4 d6f5 h4h1 b8d7 a1c1 c4d3 c1c6 d7f6 g2h3 f6h5 f3e5
info depth 18 seldepth 31 multipv 1 score cp 74 nodes 14656843 nps 54084291 tbhits 0 time 271 pv b2b4 d7d5 e4d2 h7c2 a2a4 a6d3 d2f3 b8a6 c1h6 g7h6 g4d4 h8g8 a1c1 c2a4 d4d3 a8e8 d3d2 a4b4 d2h6 b4d6 h6g5 g8h7 g5f5 h7g8 e1e8 f8e8 f3g5
info depth 19 seldepth 33 multipv 1 score cp 83 nodes 20321114 nps 53759560 tbhits 0 time 378 pv b2b4 d7d5 e4d2 b7d6 c1b2 d6c4 d2c4 a6c4 a2a3 a7a5 g2h3 a5b4 g4f4 f7f6 a3b4 a8a1 b2a1 b8a6 h3f5 h7g8 f5e6 f8e8 e6g8 e8e1 g1g2 h8g8
info depth 20 seldepth 38 multipv 1 score cp 72 nodes 27826285 nps 53615192 tbhits 0 time 519 pv b2b4 d7d5 e4d2 b7d6 c1b2 d6c4 d2c4 a6c4 e1e7 b8a6 a2a3 a8e8 e7e8 f8e8 g4d7 e8e2 b2d4 c6c5 g2f1 h7e4 f1e2 c4e2 d4b2 e2f3 d7c8 h8h7 c8a6 e4c2 a6a7 c2b2 a1e1 b2f6 a7e7 f6f5
info depth 21 seldepth 42 multipv 1 score cp 49 nodes 36747006 nps 53645264 tbhits 0 time 685 pv b2b4 d7d5 e4d2 b7d6 c1b2 d6c4 d2c4 a6c4 e1e7 b8a6 a2a3 a8e8 e7a7 c6c5 g2f1 c4f1 a1f1 a6b8 a7c7 b8a6 c7b7 e8e4 g4f3 c5b4 a3b4 f7f6 b7b6 a6c7
info depth 22 seldepth 39 multipv 1 score cp 70 nodes 40222347 nps 53629796 tbhits 0 time 750 pv b2b4 d7d5 e4d2 b7d6 g4f4 d6c4 e1e7 h7c2 d2c4 c2d1 g2f1 a6c4 f4h6 g7h6 c1b2 h8g8 a1d1 c4f1 g1f1 f8c8 e7b7 b8a6 d1d4 c8c7 d4g4 g8f8 b2g7 f8e7 b7c7 a6c7 g7h6 c7e6 g4h4
info depth 23 seldepth 42 multipv 1 score cp 87 nodes 80232625 nps 52958828 hashfull 12 tbhits 0 time 1515 pv c1g5 h6g5 e4g5 h7g8 g4h4 f8e8 h5h6 a6d3 h4d4 b8a6 d4d3 a6c5 d3d4 f7f6 g5f3 a7a5 f3h4 g7h6 g2h3 g8g7 h4f5 g7g5 f5e7 d7d5 e7c6
info depth 24 seldepth 44 multipv 1 score cp 99 nodes 90795665 nps 53252589 hashfull 13 tbhits 0 time 1705 pv c1g5 h6g5 e4g5 h7g8 g4h4 f8e8 h5h6 a6d3 h4d4 b8a6 d4d3 a6c5 d3d4 f7f6 g5f3 a7a5 g2h3 g8d5 h6g7 h8g7 d4g4 g7f8 e1d1 d5f7 f3h4 a8d8 b2b4 e8e4 g4f3 a5b4 h3f5
info depth 25 seldepth 50 multipv 1 score cp 115 nodes 127254487 nps 54058830 hashfull 19 tbhits 0 time 2354 pv c1g5 f7f5 g4f4 b7c5 f4d6 h8g8 g5e7 f5e4 e7f8 a6d3 g2h3 b8a6 f8e7 a6b4 h3d7 c5d7 d6e6 g8h8 e7b4 d7f6 e6c6 h7g8 b4c3 a8f8 e1e3 g8f7 c3f6 f7h5
info depth 26 seldepth 42 multipv 1 score cp 115 nodes 139739958 nps 54204793 hashfull 21 tbhits 0 time 2578 pv c1g5 f7f5 g4f4 b7c5 f4d6 h8g8 g5e7 c5e4 g2e4 f5e4 e7f8 a6d3 f2f3 b8a6 f3e4 d3c4 f8e7 c4e6 d6d3 a6c5 e7c5 b6c5 a1d1 a8f8 d1d2 g8h8 d2f2 f8f2 g1f2 h7g8 e1d1 g8f7 d3f3
info depth 27 seldepth 57 multipv 1 score cp 109 nodes 164209569 nps 54320069 hashfull 24 tbhits 0 time 3023 pv c1g5 h6g5 e4g5 h7g8 g4h4 f8e8 h5h6 a6d3 h4d4 b8a6 d4d3 a6c5 d3d4 f7f6 g5f3 e8e1 a1e1 a8e8 h6g7 h8g7 d4g4 g7h6 g4f4 h6g7 e1d1 e8e4 d1d7 c5d7 f4e4 b7c5 e4d4 g8d5 d4g4 g7f7 b2b4 c5e6 f3h4 d5d2 g4h5 f7e7 h4f5 e7d8 h5h8 d8c7 h8a8
info depth 28 seldepth 44 multipv 1 score cp 128 nodes 186936837 nps 54484650 hashfull 28 tbhits 0 time 3431 pv c1g5 h6g5 e4g5 h7g8 g4h4 a6d3 h5h6 b7c5 b2b4 c5e6 h6g7 h8g7 g5e6 f7e6 h4d4 f8f6 d4d3 d7d5 d3c3 g7f7 a1d1 g8f8 d1d4 f8d6 c3c2 b8d7 c2h7 f7e8 d4g4 d7f8 h7h8 e8e7 g4g7 f6f7 b4b5 c6b5 h8h4 e7e8
info depth 29 seldepth 55 multipv 1 score cp 169 nodes 351616875 nps 54438283 hashfull 51 tbhits 0 time 6459 pv c1g5 h6g5 e4g5 h7g8 g4h4 b7c5 h5h6 a6d3 b2b4 b8a6 b4c5 a6c5 e1e7 a8e8 g5f7 f8f7 h6g7 h8g7 h4g5 g7h7 g5h5 h7g7 h5e5 g7g6 e7e8 g8g7 e5e1 g7f6 a1d1 g6g7 e1e3 d3g6 f2f4 d7d5 e8e5 g6f5 g2f3 c5e4 g3g4 f5g6 f3e4 g6e4 d1f1 g7g8 e5g5 f7g7 g5g7 g8g7
info depth 30 seldepth 52 multipv 1 score cp 184 nodes 461403504 nps 54532975 hashfull 67 tbhits 0 time 8461 pv c1g5 f7f5 g4f4 b7c5 g5e7 c5d3 f4f3 d3e1 a1e1 f5e4 f3f8 h7g8 g2e4 g8f8 e7f8 a6c4 f8d6 b8a6 e4g6 c4e6 f2f4 a6c5 f4f5 e6d5 e1e7 c5e4 d6e5 e4f6 g3g4 d7d6 e5c3 a8f8 c3f6 g7f6 e7h7 h8g8 h7h6 d5f3 g1f2 f3g4 h6h7 d6d5 h5h6 c6c5 h7a7 d5d4 a7b7 f8d8 f2e1
info depth 31 seldepth 62 multipv 1 score cp 153 nodes 529601391 nps 54558709 hashfull 73 tbhits 0 time 9707 pv c1g5 f7f5 g4f4 b7c5 g5e7 c5d3 f4f3 d3e1 a1e1 f5e4 f3f8 h7g8 g2e4 g8f8 e7f8 a6c4 f8d6 b8a6 e4g6 c4e6 f2f4 a6c5 f4f5 e6d5 e1e7 c5e4 d6e5 e4f6 g3g4 d7d6 e5c3 a8f8 e7a7 b6b5 c3f6 g7f6 g1f2

info 10 seconds
info depth 31 seldepth 44 multipv 1 score cp 145 nodes 545544887 nps 54543579 hashfull 75 tbhits 0 time 10002 pv c1g5 h6g5
info depth 32 seldepth 59 multipv 1 score cp 139 nodes 600700189 nps 54594218 hashfull 82 tbhits 0 time 11003 pv c1g5 h6g5 e4g5 h7g8 g4h4 b7c5 h5h6 a6d3 b2b4 b8a6 b4c5 a6c5 e1e7 a8e8 g5f7 f8f7 h6g7 h8g7 h4g5 d3g6 g5e5 g7h7 e7e8 g8g7 a1f1 g7e5 e8e5 a7a5 f2f4 h7g7 g1h2 g6d3 f1d1 d7d5 e5e8 f7f8 e8f8 g7f8 g3g4 d3e2 d1b1 e2g4 b1b6 g4d7 b6b8 f8e7 b8a8
info depth 33 seldepth 60 multipv 1 score cp 171 nodes 647021579 nps 54651708 hashfull 88 tbhits 0 time 11839 pv c1g5 h6g5 e4g5 h7g8 g4h4 a6d3 h5h6 b7c5 b2b4 b8a6 b4c5 a6c5 e1e7 a8e8 g5f7 f8f7 h6g7 h8g7 h4g5 d3g6 g5e5 g7h7 e7e8 g8g7 a1d1 g7e5 e8e5 a7a5 f2f4 h7g7 g2h3 g6c2 d1d2 c2h7 d2d6 f7f6 d6f6 g7f6 g1h2 a5a4 g3g4 c5d3 e5h5 h7g6 g4g5 f6e7 h5h4
info depth 34 seldepth 66 multipv 1 score cp 178 nodes 925851712 nps 54276686 hashfull 133 tbhits 0 time 17058 pv c1g5 f7f5 g4f4 b7c5 g5e7 c5d3 f4f3 d3e1 a1e1 f5e4 f3f8 h7g8 g2e4 g8f8 e7f8 a6c4 f8d6 b8a6 e4g6 c4e6 f2f4 a6c5 f4f5 e6d5 e1e7 c5e4 d6e5 e4f6 g3g4 a8f8 a2a3 h8g8 e5f6 f8f6 g1f2 a7a5 e7d7 g8f8 f2e3 d5b3 e3f4 c6c5 f4e5

info 20 seconds
info depth 34 seldepth 58 multipv 1 score cp 178 nodes 1083456897 nps 54167428 hashfull 160 tbhits 0 time 20002 pv c1g5
info depth 35 seldepth 70 multipv 1 score cp 170 nodes 1232007406 nps 54130378 hashfull 186 tbhits 0 time 22760 pv c1g5 h6g5 e4g5 h7g8 g4h4 b7c5 h5h6 a6d3 b2b4 b8a6 b4c5 a6c5 e1e7 a8e8 g5f7 f8f7 h6g7 h8g7 h4g5 d3g6 g5e5 g7h7 e7e8 g8g7 a1d1 g7e5 e8e5 d7d5 f2f4 h7g7 g2f3 b6b5 f3h5 g6h5 e5h5 c5e4 g1g2 b5b4 g2f3 a7a5 d1c1 e4c3 c1h1 c3a2 g3g4 a5a4 h5h7 g7f8 h7h8 f8e7 g4g5 a2c3 g5g6 f7g7 f4f5 e7d6 h8h7
info depth 36 seldepth 75 multipv 1 score cp 163 nodes 1400612185 nps 54169716 hashfull 215 tbhits 0 time 25856 pv c1g5 h6g5 e4g5 h7g8 g4h4 a6d3 h5h6 b7c5 b2b4 b8a6 b4c5 a6c5 e1e7 a8e8 g5f7 f8f7 h6g7 h8g7 h4g5 d3g6 g5e5 g7h7 e7e8 g8g7 a1d1 g7e5 e8e5 d7d5 f2f4 h7g7 g2f3 b6b5 f3h5 f7f6 g1g2 g6e4 h5f3 b5b4 f3e4 c5e4 g2f3 a7a5 f4f5 a5a4 f3f4 b4b3 a2b3 a4b3 g3g4 c6c5 g4g5 f6f7 g5g6 f7c7 e5d5 e4f6 d5e5 b3b2 d1b1 c7b7 e5c5 b7b4 f4e3

info 30 seconds
info depth 36 seldepth 47 multipv 1 score cp 172 nodes 1625717387 nps 54186967 hashfull 250 tbhits 0 time 30002 pv c1g5
info depth 37 seldepth 74 multipv 1 score cp 186 nodes 1655248113 nps 54169195 hashfull 254 tbhits 0 time 30557 pv c1g5 h6g5 e4g5 h7g8 g4h4 a6d3 h5h6 f8e8 h4d4 b8a6 d4d7 e8f8 d7d3 g7h6 g5f3 b7c5 d3c4 a6c7 a1d1 g8g7 f3d4 a7a5 d4c6 c7e6 c4h4 g7g5 h4h3 a8e8 e1e5 g5g6 h3h4 e6g5 d1d5 c5e6 d5d6 f7f5 c6d4 f5f4 d4e6 g5e6 d6b6 f4g3

info 40 seconds
info depth 37 seldepth 68 multipv 1 score cp 167 nodes 2137393058 nps 53432154 hashfull 324 tbhits 0 time 40002 pv c1g5
info depth 38 seldepth 76 multipv 1 score cp 213 nodes 2560865999 nps 52953123 hashfull 385 tbhits 0 time 48361 pv c1g5 h6g5 e4g5 h7g8 g4h4 f8e8 h5h6 a6d3 h4d4 b8a6 d4d7 e8f8 d7d3 g7h6 g5f3 b7c5 d3a3 a6c7 a1d1 g8g7 b2b4 c5e6 f3e5 a7a5 b4b5 c7b5 a3e3 b5c7 e3b6 a8d8 e5c6 d8d1 e1d1 g7g5 d1d7 c7e8 b6a5 g5a5 c6a5 e8f6 d7a7 f8d8 a7f7 d8d1 g2f1 f6e4 g1g2 e6g5 f7f4 d1d2 f1c4 e4d6

info 50 seconds
info depth 38 seldepth 76 multipv 1 score cp 213 nodes 2641885981 nps 52834549 hashfull 398 tbhits 0 time 50003 pv c1g5 h6g5 e4g5 h7g8 g4h4 f8e8 h5h6 a6d3 h4d4 b8a6 d4d7 e8f8 d7d3 g7h6 g5f3 b7c5 d3a3 a6c7 a1d1 g8g7 b2b4 c5e6 f3e5 a7a5 b4b5 c7b5 a3e3 b5c7 e3b6 a8d8 e5c6 d8d1 e1d1 g7g5 d1d7 c7e8 b6a5 g5a5 c6a5 e8f6 d7a7 f8d8 a7f7 d8d1 g2f1 f6e4 g1g2 e6g5 f7f4 d1d2 f1c4 e4d6
info depth 39 seldepth 64 multipv 1 score cp 204 nodes 3149638349 nps 52492222 hashfull 462 tbhits 0 time 60002 pv c1g5 h6g5
bestmove c1g5 ponder h6g5
Image

peter
Posts: 2132
Joined: Sat Feb 16, 2008 6:38 am
Full name: Peter Martan

Re: Alpha Zero's 21. Bg5! revisited

Post by peter » Sat Aug 08, 2020 4:56 pm

MikeB wrote:
Sat Aug 08, 2020 3:26 pm
it did shine, I only see Bluefish Xir4 now getting this in less than a minute, your mileage may vay
As it may well be with many positions, that could be solved with SF NNUE (almost as well or even better then with LC0) before the mystery- patch called NNUE-treshold-eval or mixed evalutation or hybrid eval, started switching off the use of nets trained with much hardware- time of their own but then just don't get used by the engine even without any chance for the user to see the difference other then in output.
For nothing else then 11 Elo in STC and selfplay.
:(

https://github.com/official-stockfish/S ... -670598949

viewtopic.php?f=2&t=74719
Peter.

Post Reply