Crystal Update

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.
Dann Corbit
Posts: 11221
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Crystal Update

Post by Dann Corbit » Thu Apr 16, 2020 4:49 pm

Jouni wrote:
Thu Apr 09, 2020 6:56 pm
I tested it in my mate suite. Clearly the best mate solver now! Nice progress from previous version. Thanks.
Is your mate suite available?
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Jouni
Posts: 2133
Joined: Wed Mar 08, 2006 7:15 pm

Re: Crystal Update

Post by Jouni » Thu Apr 16, 2020 5:38 pm

No. Simply selection of long mates from Chest database.
Jouni

jhellis3
Posts: 460
Joined: Fri Aug 16, 2013 10:36 pm

Re: Crystal Update

Post by jhellis3 » Thu Apr 16, 2020 6:42 pm

I hope You have looked at Houdini 6 tactical mode source. It solves some tactics still faster than Crystal even being from 2017!
To be honest, I doubt I will. While the code is GPL, it was undoubtedly not intended to be public, and I have no way of knowing if all of the ideas present are genuinely RH's or possibly someone else's.

Also important to remember that Crystal gives up quite a bit of depth by using a more robust search, so it is natural that deep mates may take slightly longer.

However, if there is very large difference in time or there are positions Crystal just doesn't get at all, by all means post the fens and I will add it to my collection of stuff to work on. :)

nnnnnnnn
Posts: 22
Joined: Mon Nov 18, 2019 1:36 pm
Full name: Mark Thellen

Re: Crystal Update

Post by nnnnnnnn » Fri Apr 17, 2020 4:18 am

I ran the latest Crystal on a 6-CPU iMac with 6 threads and I think lots of hash (although not certain) and set to disregard the 50-move rule.

It failed to give a correct evaluation of this position, which is very difficult for the engines I have tried as the key idea is pruned early:

2R2K2/p3p3/3p4/p2P4/P7/4p3/6pp/6nk w - - 0 1

Are there engines than can solve this from the start position - not just the correct move, which is easy (1. Rc1) but the evaluation (white wins)?

kiroje
Posts: 27
Joined: Wed Jul 25, 2012 8:12 am
Location: Copenhagen
Full name: Kim Jensen
Contact:

Re: Crystal Update

Post by kiroje » Fri Apr 17, 2020 8:20 am

nnnnnnnn wrote:
Fri Apr 17, 2020 4:18 am
I ran the latest Crystal on a 6-CPU iMac with 6 threads and I think lots of hash (although not certain) and set to disregard the 50-move rule.

It failed to give a correct evaluation of this position, which is very difficult for the engines I have tried as the key idea is pruned early:

2R2K2/p3p3/3p4/p2P4/P7/4p3/6pp/6nk w - - 0 1

Are there engines than can solve this from the start position - not just the correct move, which is easy (1. Rc1) but the evaluation (white wins)?
New version of Sting solves this in 4 seconds on my old laptop.
0.18 1.Rc1 (depth 24, 1/16), 116,3M nodes (6,2M nps)
0.01: 0.00/18 1.Rc1 a6 2.Kg7 e2 3.Kh6 e6 4.dxe6 d5 5.e7 d4 6.e8=Q d3 7.Qe3 d2 8.Qxd2 e1=Q 9.Rxe1
0.02: 0.00/19 1.Rc1 a6 2.Ra1 e2 3.Rb1 e6 4.dxe6 d5 5.e7 d4 6.e8=B d3 7.Bh5 d2 8.Bxe2 d1=N 9.Bxd1 Nh3 10.Be2+ Ng1 11.Bxa6
0.03: 0.00/20 1.Rc1 a6 2.Ra1 e2 3.Rb1 e6 4.dxe6 d5 5.e7 d4 6.e8=B d3 7.Bh5 d2 8.Bxe2 d1=N 9.Bxd1 Nh3 10.Be2+ Ng1 11.Bxa6
0.04: +11.11/21 1.Rc1 e2 2.Rb1 e6 3.dxe6 d5 4.e7 d4 5.e8=B d3 6.Bh5 d2 7.Bxe2 d1=B 8.Bxd1 Nh3 9.Bf3+ Ng1 10.Be4 a6 11.Rb2 Nf3 12.Bxf3 Kg1 13.Rxg2+ Kf1 14.Rxh2
0.04: +80.77/22 1.Rc1
0.16: +80.77/23 1.Rc1 e2 2.Rb1 e6 3.dxe6 d5 4.e7 d4 5.e8=B d3 6.Bh5 d2 7.Bxe2 d1=B 8.Bxd1 Nh3 9.Bf3+ Ng1 10.Be4 a6 11.Rb2 Nf3 12.Bxf3 Kg1 13.Rxg2+ Kf1 14.Rxh2 Kg1 15.Rh6
0.18: +80.77/24 1.Rc1 e2 2.Rb1 e6 3.dxe6 d5 4.e7 d4 5.e8=B d3 6.Bh5 d2 7.Bxe2 d1=B 8.Bxd1 Nh3 9.Bf3+ Ng1 10.Be4 a6 11.Rb2 Nf3 12.Bxf3 Kg1 13.Rxg2+ Kf1 14.Rxh2 Kg1 15.Rh6
Want to play a game ?
Lichess handle : kiroje
Chess.com handle : kiroje
Chess24 handle : kiroje

peter
Posts: 2004
Joined: Sat Feb 16, 2008 6:38 am
Full name: Peter Martan

Re: Crystal Update

Post by peter » Fri Apr 17, 2020 9:47 am

nnnnnnnn wrote:
Fri Apr 17, 2020 4:18 am
2R2K2/p3p3/3p4/p2P4/P7/4p3/6pp/6nk w - - 0 1
CorChess was at three trials always about as quick as here:



Analysis by CorChess 6.0 010420 64 POPCNT:

Code: Select all

1.Rc1 a6 2.Kxe7 e2 3.Kxd6 e1Q 4.Rxe1 
  =  (0.00)   Depth: 7/7   00:00:00  10kN
...
1.Rc1 a6 2.Rb1 e2 3.Rc1 e6 4.dxe6 d5 5.e7 d4 6.e8Q d3 7.Ra1 d2 8.Qxe2 d1Q 9.Qxd1 
  =  (0.00)   Depth: 22/18   00:00:00  339kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 a6 7.Ra1 e1B 8.Rxe1 d2 9.Ra1 d1R 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Ne2 13.Be4+ Ng1 14.Ra2 
  +-  (2.42)   Depth: 23/32   00:00:00  643kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 e1N 7.Rxe1 d2 8.Rd1 a6 9.Ra1 d1B 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Nf3 13.Be4+ Ng1 14.Ra2 Ne2 15.Rxe2 Kg1 16.Rxg2+ Kf1 
  +-  (7.75)   Depth: 24/40   00:00:00  1067kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 d2 7.Bxe2 d1Q 8.Bxd1 a6 9.Ra1 Nh3 10.Bc2+ Ng1 11.Bb1 Ne2 
  +-  (54.08)   Depth: 25/40   00:00:00  1205kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 d2 7.Bxe2 d1Q 8.Bxd1 a6 9.Ra1 Nh3 10.Bc2+ Ng1 11.Bb1 Ne2 12.Be4+ Ng1 13.Ra2 Nh3 14.Rxg2 Nf4 15.Rg4+ Ng2 16.Rg3 Kg1 
  +-  (54.08)   Depth: 26/40   00:00:00  1267kN
...
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 d2 7.Bxe2 d1N 8.Bxd1 a6 9.Ra1 Nh3 10.Bc2+ Ng1 11.Bb1 Ne2 12.Be4+ Ng1 13.Ra2 Nh3 14.Rxg2 Nf4 15.Rg8+ Ng2 16.Rg6 Kg1 17.Rxg2+ Kf1 18.Rxh2 Kg1 19.Rg2+ Kh1 20.Rg5+ Kh2 21.Bb7 Kh3 22.Rxa5 Kg3 23.Rxa6 Kg4 24.a5 Kf4 25.Rb6 Kf5 26.a6 Kf4 27.Rc6 Kg3 
  +-  (69.27)   Depth: 49/57   00:00:00  10262kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 e1Q 7.Rxe1 d2 8.Rd1 a6 9.Ra1 d1B 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Ne2 13.Be4+ Ng1 14.Ra2 Nh3 15.Rxg2 Nf4 16.Rg8+ Ng2 17.Bb7 Kg1 18.Rxg2+ Kf1 19.Rxh2 Ke1 20.Bf3 Kf1 21.Ra2 Ke1 22.Rg2 Kf1 23.Bb7 Ke1 24.Bxa6 Kd1 25.Ke7 Kc1 26.Kd6 Kd1 27.Rh2 Kc1 28.Kc5 
  +-  (#35)   Depth: 50/57   00:00:01  22090kN
...
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 a6 7.Ra1 e1R 8.Rxe1 d2 9.Ra1 d1B 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Ne2 13.Be4+ Ng1 14.Ra2 Nh3 15.Rxg2 Nf4 16.Rg8+ Ng2 17.Bb7 Kg1 18.Rxg2+ Kf1 19.Rxh2 Kg1 20.Rg2+ Kh1 21.Rb2+ Kg1 22.Ke7 Kf1 23.Rg2 Ke1 24.Bxa6 Kd1 25.Kd6 Kc1 26.Kc5 Kb1 27.Kd4 Kc1 28.Kc3 Kb1 29.Bc4 Kc1 30.Rg1# 
  +-  (#30)   Depth: 68/60   00:02:34  4123MN
24 threads of 3x12GHz CPU, 32G hash.

Distance to mate is too high, yet that's to be seen als solved, I'd say.

As quickly as that (almost at once) Matefinder gets it too.
Peter.

User avatar
MikeB
Posts: 4181
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Crystal Update

Post by MikeB » Fri Apr 17, 2020 12:26 pm

peter wrote:
Fri Apr 17, 2020 9:47 am
nnnnnnnn wrote:
Fri Apr 17, 2020 4:18 am
2R2K2/p3p3/3p4/p2P4/P7/4p3/6pp/6nk w - - 0 1
CorChess was at three trials always about as quick as here:



Analysis by CorChess 6.0 010420 64 POPCNT:

Code: Select all

1.Rc1 a6 2.Kxe7 e2 3.Kxd6 e1Q 4.Rxe1 
  =  (0.00)   Depth: 7/7   00:00:00  10kN
...
1.Rc1 a6 2.Rb1 e2 3.Rc1 e6 4.dxe6 d5 5.e7 d4 6.e8Q d3 7.Ra1 d2 8.Qxe2 d1Q 9.Qxd1 
  =  (0.00)   Depth: 22/18   00:00:00  339kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 a6 7.Ra1 e1B 8.Rxe1 d2 9.Ra1 d1R 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Ne2 13.Be4+ Ng1 14.Ra2 
  +-  (2.42)   Depth: 23/32   00:00:00  643kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 e1N 7.Rxe1 d2 8.Rd1 a6 9.Ra1 d1B 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Nf3 13.Be4+ Ng1 14.Ra2 Ne2 15.Rxe2 Kg1 16.Rxg2+ Kf1 
  +-  (7.75)   Depth: 24/40   00:00:00  1067kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 d2 7.Bxe2 d1Q 8.Bxd1 a6 9.Ra1 Nh3 10.Bc2+ Ng1 11.Bb1 Ne2 
  +-  (54.08)   Depth: 25/40   00:00:00  1205kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 d2 7.Bxe2 d1Q 8.Bxd1 a6 9.Ra1 Nh3 10.Bc2+ Ng1 11.Bb1 Ne2 12.Be4+ Ng1 13.Ra2 Nh3 14.Rxg2 Nf4 15.Rg4+ Ng2 16.Rg3 Kg1 
  +-  (54.08)   Depth: 26/40   00:00:00  1267kN
...
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 d2 7.Bxe2 d1N 8.Bxd1 a6 9.Ra1 Nh3 10.Bc2+ Ng1 11.Bb1 Ne2 12.Be4+ Ng1 13.Ra2 Nh3 14.Rxg2 Nf4 15.Rg8+ Ng2 16.Rg6 Kg1 17.Rxg2+ Kf1 18.Rxh2 Kg1 19.Rg2+ Kh1 20.Rg5+ Kh2 21.Bb7 Kh3 22.Rxa5 Kg3 23.Rxa6 Kg4 24.a5 Kf4 25.Rb6 Kf5 26.a6 Kf4 27.Rc6 Kg3 
  +-  (69.27)   Depth: 49/57   00:00:00  10262kN
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 e1Q 7.Rxe1 d2 8.Rd1 a6 9.Ra1 d1B 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Ne2 13.Be4+ Ng1 14.Ra2 Nh3 15.Rxg2 Nf4 16.Rg8+ Ng2 17.Bb7 Kg1 18.Rxg2+ Kf1 19.Rxh2 Ke1 20.Bf3 Kf1 21.Ra2 Ke1 22.Rg2 Kf1 23.Bb7 Ke1 24.Bxa6 Kd1 25.Ke7 Kc1 26.Kd6 Kd1 27.Rh2 Kc1 28.Kc5 
  +-  (#35)   Depth: 50/57   00:00:01  22090kN
...
1.Rc1 e2 2.Rb1 e5 3.dxe6 d5 4.e7 d4 5.e8B d3 6.Bh5 a6 7.Ra1 e1R 8.Rxe1 d2 9.Ra1 d1B 10.Bxd1 Nh3 11.Bc2+ Ng1 12.Bb1 Ne2 13.Be4+ Ng1 14.Ra2 Nh3 15.Rxg2 Nf4 16.Rg8+ Ng2 17.Bb7 Kg1 18.Rxg2+ Kf1 19.Rxh2 Kg1 20.Rg2+ Kh1 21.Rb2+ Kg1 22.Ke7 Kf1 23.Rg2 Ke1 24.Bxa6 Kd1 25.Kd6 Kc1 26.Kc5 Kb1 27.Kd4 Kc1 28.Kc3 Kb1 29.Bc4 Kc1 30.Rg1# 
  +-  (#30)   Depth: 68/60   00:02:34  4123MN
24 threads of 3x12GHz CPU, 32G hash.

Distance to mate is too high, yet that's to be seen als solved, I'd say.

As quickly as that (almost at once) Matefinder gets it too.
# Black-Diamond-dev-041620
Black Diamond 041620 by T. Romstad, M. Costalba, J. Kiiski, G. Linscott
info string Hash LargePages 256 Mb
s t 10
Confirmation: Threads set to 10
s h 2048
info string Hash LargePages 2048 Mb
Confirmation: Hash set to 2048
s dpa true
Confirmation: Deep Pro Analysis set to true
p f 2R2K2/p3p3/3p4/p2P4/P7/4p3/6pp/6nk w - - 0 1
g i m 30

Pro Analysis value1: 0


Pro Analysis value2: 14400000

info depth 1 seldepth 1 multipv 1 score cp -839 nodes 1134 nps 567000 tbhits 0 time 2 pv c8c2
info depth 2 seldepth 2 multipv 1 score cp -755 nodes 6594 nps 2198000 tbhits 0 time 3 pv c8c2 g1e2
info depth 3 seldepth 3 multipv 1 score cp -742 nodes 11599 nps 3866333 tbhits 0 time 3 pv c8c1 a7a6 f8e7
info depth 4 seldepth 4 multipv 1 score cp -660 nodes 28762 nps 7190500 tbhits 0 time 4 pv c8c1 a7a6 f8e7 e3e2
info depth 5 seldepth 6 multipv 1 score cp -389 nodes 57504 nps 11500800 tbhits 0 time 5 pv c8c1 a7a6 f8e7 e3e2 e7d6
info depth 6 seldepth 7 multipv 1 score cp -167 nodes 95311 nps 13615857 tbhits 0 time 7 pv c8c1 e7e6 d5e6 e3e2 e6e7 e2e1r c1e1
info depth 7 seldepth 12 multipv 1 score cp -140 nodes 200602 nps 15430923 tbhits 0 time 13 pv c8c1 e7e6 d5e6 e3e2 e6e7 e2e1r c1e1
info depth 8 seldepth 8 multipv 1 score cp 0 nodes 310605 nps 16347631 tbhits 0 time 19 pv c8c1 e3e2 f8e7 a7a6 e7d6 e2e1r c1e1
info depth 9 seldepth 8 multipv 1 score cp 0 nodes 557061 nps 17408156 tbhits 0 time 32 pv c8c1 e3e2 f8e7 a7a6 e7d6 e2e1r c1e1
info depth 10 seldepth 8 multipv 1 score cp 0 nodes 1093647 nps 17639467 tbhits 0 time 62 pv c8c1 e3e2 f8e7 a7a6 e7d6 e2e1r c1e1
info depth 11 seldepth 8 multipv 1 score cp 0 nodes 2078086 nps 17610898 tbhits 0 time 118 pv c8c1 e3e2 f8e7 a7a6 e7d6 e2e1r c1e1
info depth 12 seldepth 8 multipv 1 score cp 0 nodes 4046262 nps 17516285 tbhits 0 time 231 pv c8c1 a7a6 f8e7 e3e2 e7d6 e2e1r c1e1
info depth 13 seldepth 8 multipv 1 score cp 0 nodes 6478327 nps 17652117 tbhits 0 time 367 pv c8c1 a7a6 f8e7 e3e2 e7d6 e2e1r c1e1
info depth 14 seldepth 8 multipv 1 score cp 0 nodes 13390288 nps 17925419 tbhits 0 time 747 pv c8c1 a7a6 f8e7 e3e2 e7d6 e2e1r c1e1
info depth 15 seldepth 8 multipv 1 score cp 0 nodes 23662499 nps 18201922 hashfull 63 tbhits 0 time 1300 pv c8c1 a7a6 f8e7 e3e2 e7d6 e2e1r c1e1
info depth 16 seldepth 18 multipv 1 score cp 0 nodes 23665974 nps 18204595 hashfull 63 tbhits 0 time 1300 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 17 seldepth 18 multipv 1 score cp 0 nodes 23668248 nps 18206344 hashfull 63 tbhits 0 time 1300 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 18 seldepth 18 multipv 1 score cp 0 nodes 23670768 nps 18208283 hashfull 63 tbhits 0 time 1300 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 19 seldepth 18 multipv 1 score cp 0 nodes 23673393 nps 18210302 hashfull 63 tbhits 0 time 1300 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 20 seldepth 18 multipv 1 score cp 0 nodes 23676463 nps 18198664 hashfull 63 tbhits 0 time 1301 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 21 seldepth 18 multipv 1 score cp 0 nodes 23681669 nps 18202666 hashfull 63 tbhits 0 time 1301 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 22 seldepth 18 multipv 1 score cp 0 nodes 23687405 nps 18207075 hashfull 63 tbhits 0 time 1301 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 23 seldepth 18 multipv 1 score cp 0 nodes 23694729 nps 18212704 hashfull 63 tbhits 0 time 1301 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 24 seldepth 18 multipv 1 score cp 0 nodes 23707979 nps 18208893 hashfull 63 tbhits 0 time 1302 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 25 seldepth 18 multipv 1 score cp 0 nodes 23720025 nps 18204163 hashfull 63 tbhits 0 time 1303 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 26 seldepth 18 multipv 1 score cp 0 nodes 25315248 nps 18225520 hashfull 65 tbhits 0 time 1389 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 27 seldepth 18 multipv 1 score cp 0 nodes 33303050 nps 18099483 hashfull 73 tbhits 0 time 1840 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 28 seldepth 18 multipv 1 score cp 0 nodes 36876333 nps 18058928 hashfull 76 tbhits 0 time 2042 pv c8c1 a7a6 f8g7 e3e2 c1a1 e7e5 d5e6 d6d5 e6e7 d5d4 e7e8r e2e1q e8e1 d4d3 e1d1 d3d2 d1d2
info depth 29 seldepth 83 multipv 1 score cp 10017 nodes 62267727 nps 17991253 hashfull 113 tbhits 0 time 3461 pv c8c1
info depth 30 seldepth 102 multipv 1 score mate 27 nodes 62267727 nps 17991253 hashfull 113 tbhits 0 time 3461 pv c8c1 a7a6 c1d1 e7e6 d5e6 d6d5 e6e7 e3e2 d1a1 d5d4 e7e8b d4d3 e8h5 e2e1q a1e1 d3d2 e1a1 d2d1q h5d1 g1h3 d1c2 h3g1 c2b1 g1f3 b1e4 f3g1 a1a2 g1h3 a2g2 h3f4 g2g3 f4g2 e4f3 h1g1 g3g2 g1f1 g2h2 f1e1 h2b2 e1f1 b2c2 f1g1 f8f7 g1f1 f7e6 f1e1 e6d5 e1f1 d5d4 f1e1 d4e3 e1f1 c2c1
bestmove c8c1 ponder a7a6
Image

jhellis3
Posts: 460
Joined: Fri Aug 16, 2013 10:36 pm

Re: Crystal Update

Post by jhellis3 » Fri Apr 17, 2020 2:07 pm

Thanks for the position and for the hint that matefinder sees it. There is really only a couple of things it could be, and I have a pretty good idea which one it is. Will see if I can do something about it.....

nnnnnnnn
Posts: 22
Joined: Mon Nov 18, 2019 1:36 pm
Full name: Mark Thellen

Re: Crystal Update

Post by nnnnnnnn » Fri Apr 17, 2020 10:41 pm

In this position, White wins with 1. g6+ Kh8 2. a:b7 Bg8 3. a8=R! (3. a3=Q is a fortress draw).

On my machine, after 6 minutes Crystal is seeing the correct move 1. g6+ and the correct evaluation, but its principal variation is incorrect: it sees 3. a8=Q instead of 3. a8=R. Crystal does quickly find the fortress and the winning plan if started on the position following 2...Bg8, which Stockfish 11 is unable to do.

jhellis3
Posts: 460
Joined: Fri Aug 16, 2013 10:36 pm

Re: Crystal Update

Post by jhellis3 » Fri Apr 17, 2020 11:34 pm

Ok, this one I was suspicious of and rightfully so. Unless there is some obvious immediate reason to under promote then queen should be fine. My clue was the fact that the only difference between queen and rook is the diagonal attack which is blocked by its own pawn. So I played out the winning line with the rook, and the rook does not do anything that the queen could not as well... there is no stalemate issue.

Thus promoting to queen is fine, the mate is still there and it will eventually see it if you play enough moves. The likely reason it is showing a draw is a pruning issue. Because white now has a queen instead of a rook, it thinks losing that piece is worse and thus the key move gets pruned more severely than it does as a rook. Missing the key move results in an incorrect draw score.

Honestly, this one is going to be much more difficult to fix because generally speaking losing a queen is much worse than losing a rook, and it does make sense to score such moves worse and prune those moves more severely. But I will keep it in mind and see what I can do...

Post Reply