King and knight locked together

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

Moderators: Harvey Williamson, Dann Corbit, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Javier Ros
Posts: 184
Joined: Fri Oct 12, 2012 10:48 am
Location: Seville (SPAIN)
Full name: Javier Ros

King and knight locked together

Post by Javier Ros » Sun Feb 03, 2019 10:11 am

Another incredible position was reached in the following game, where Stockfish 10 played White on 6 cores i7 980X and AF125 on 2xGTX1060 with 15 minutes plus 1 second time control.

After 7 moves of Two Knights Defense from pgn, both engines playing by themselves followed the recommended variant of Hiarcs14zc book until move 29!! where ADF125 played 29...f4. White has two passed pawns advantage.

Image

Previously to 36.g3? Stockfish evaluation was 0.00 and AF125's 1.27, after g3 eval of Lc0 jumped to 4.32 because the prison was closed!

35. Rc1 {0.00/37 21} Kf5 {1.27/15 12} 36. g3 {0.00/31 11} Ba6 {4.32/13 34} 37. Rc5+ {-2.73/34 105}

The love relationship between a chess engine tester and his computer can be summarized in one sentence:
Until heat do us part.

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

Re: King and knight locked together

Post by zullil » Sun Feb 03, 2019 10:35 am

Javier Ros wrote:
Sun Feb 03, 2019 10:11 am
Another incredible position was reached in the following game, where Stockfish 10 played White on 6 cores i7 980X and AF125 on 2xGTX1060 with 15 minutes plus 1 second time control.

Previously to 36.g3? Stockfish evaluation was 0.00 and AF125's 1.27, after g3 eval of Lc0 jumped to 4.32 because the prison was closed!


Indeed, even Stockfish-dev (default settings, 1 thread deterministic search, no endgame tables) takes a long time to realize that g3 is wrong.

info depth 38 seldepth 12 multipv 1 score cp 0 nodes 98663635 nps 2313332 hashfull 1000 tbhits 0 time 42650 pv g2g3 e4e3 d2e3 d3e4 h2h3 h6h5 c1c5 f5e6 c5c4 e6e5 c4c5
info depth 39 currmove g2g3 currmovenumber 1
info depth 39 currmove g2f3 currmovenumber 2
info depth 39 currmove a2a4 currmovenumber 3
info depth 39 currmove c1a1 currmovenumber 4
info depth 39 currmove b3b4 currmovenumber 5
info depth 39 currmove c1f1 currmovenumber 6
info depth 39 currmove g1e2 currmovenumber 7
info depth 39 currmove c1c7 currmovenumber 8
info depth 39 currmove a2a3 currmovenumber 9
info depth 39 currmove h2h3 currmovenumber 10
info depth 39 currmove c1c6 currmovenumber 11
info depth 39 currmove c1d1 currmovenumber 12
info depth 39 currmove g1h3 currmovenumber 13
info depth 39 currmove c1c3 currmovenumber 14
info depth 39 currmove c1b1 currmovenumber 15
info depth 39 currmove c1c5 currmovenumber 16
info depth 39 currmove h2h4 currmovenumber 17
info depth 39 currmove c1c4 currmovenumber 18
info depth 39 currmove c1e1 currmovenumber 19
info depth 39 currmove c1c2 currmovenumber 20
info depth 39 currmove c1c8 currmovenumber 21
info depth 39 currmove g1f3 currmovenumber 22
info depth 39 seldepth 45 multipv 1 score cp -10 upperbound nodes 255422516 nps 2346468 hashfull 1000 tbhits 0 time 108854 pv g2g3 d3a6
info depth 39 currmove g2g3 currmovenumber 1
info depth 39 currmove g2f3 currmovenumber 2
info depth 39 currmove a2a4 currmovenumber 3
info depth 39 currmove c1c5 currmovenumber 4
info depth 39 currmove g1h3 currmovenumber 5
info depth 39 currmove b3b4 currmovenumber 6
info depth 39 currmove c1c6 currmovenumber 7
info depth 39 currmove a2a3 currmovenumber 8
info depth 39 currmove c1c4 currmovenumber 9
info depth 39 currmove c1c8 currmovenumber 10
info depth 39 currmove h2h4 currmovenumber 11
info depth 39 currmove h2h3 currmovenumber 12
info depth 39 currmove c1f1 currmovenumber 13
info depth 39 currmove c1e1 currmovenumber 14
info depth 39 currmove c1c3 currmovenumber 15
info depth 39 currmove c1a1 currmovenumber 16
info depth 39 currmove c1c2 currmovenumber 17
info depth 39 currmove c1c7 currmovenumber 18
info depth 39 currmove g1e2 currmovenumber 19
info depth 39 currmove c1b1 currmovenumber 20
info depth 39 currmove c1d1 currmovenumber 21
info depth 39 currmove g1f3 currmovenumber 22
info depth 39 seldepth 45 multipv 1 score cp -19 upperbound nodes 403696554 nps 2352297 hashfull 1000 tbhits 0 time 171618 pv g2g3 d3a6

...

info depth 39 seldepth 60 multipv 1 score cp -13 nodes 557535008 nps 2349722 hashfull 1000 tbhits 0 time 237277 pv a2a4 d8d5 b3b4 d5d4 a4a5 d4b4 a5a6 b4a4 c1c5 f5f6 c5c6 f6e5 c6h6 a4a6 h6a6 d3a6 g2f3 e4f3 h2h3 e5f4 h3g4 f4g4 h1h2 a6b7 g1h3 b7c8 h3g1 c8a6 g1f3 g4f3 h2g1 f3g4 f2f3 g4f3 g1h2 f3f2 d2d3 a6d3 h2h1 f2f3 h1h2

Javier Ros
Posts: 184
Joined: Fri Oct 12, 2012 10:48 am
Location: Seville (SPAIN)
Full name: Javier Ros

Re: King and knight locked together

Post by Javier Ros » Sun Feb 03, 2019 7:53 pm

It is curious that in a test to show the randomness of Lc0 it is Stockfish who caught this randomness playing differently in the second round, winning with 9.h4! in a splendid attack game.

The love relationship between a chess engine tester and his computer can be summarized in one sentence:
Until heat do us part.

Post Reply