Chess.com 2018 computer chess championship

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.
JJJ
Posts: 1286
Joined: Sat Apr 19, 2014 11:47 am

Re: Chess.com 2018 computer chess championship

Post by JJJ » Mon Sep 10, 2018 6:41 pm

Most test doesn't show Leela as strong as Stockfish dev at all and even below Komodo / Houdini. I find the result of this CCCC very accurate when I compare to the other test I saw.

Paloma
Posts: 847
Joined: Thu Dec 25, 2008 8:07 pm

Re: Chess.com 2018 computer chess championship

Post by Paloma » Tue Sep 11, 2018 12:15 am

Hello folks,

i know this link to reach the Chess.com 2018 computer chess championship.
https://www.chess.com/computer-chess-championship

But how to reach this page via menu inside chess.com,
neither about Home, Play, learn or share i find this link.

Nay Lin Tun
Posts: 519
Joined: Mon Jan 16, 2012 5:34 am

Re: Chess.com 2018 computer chess championship

Post by Nay Lin Tun » Tue Sep 11, 2018 2:26 am

JJJ wrote:
Mon Sep 10, 2018 6:41 pm
Most test doesn't show Leela as strong as Stockfish dev at all and even below Komodo / Houdini. I find the result of this CCCC very accurate when I compare to the other test I saw.
There is another thread, ? Leela official better than SF 9.

Nay Lin Tun
Posts: 519
Joined: Mon Jan 16, 2012 5:34 am

Re: Chess.com 2018 computer chess championship

Post by Nay Lin Tun » Tue Sep 11, 2018 2:28 am

Paloma wrote:
Tue Sep 11, 2018 12:15 am
Hello folks,

i know this link to reach the Chess.com 2018 computer chess championship.
https://www.chess.com/computer-chess-championship

But how to reach this page via menu inside chess.com,
neither about Home, Play, learn or share i find this link.
Type, "chess.com/cccc"

It is their website code, no menu link in their v3 interface.

JJJ
Posts: 1286
Joined: Sat Apr 19, 2014 11:47 am

Re: Chess.com 2018 computer chess championship

Post by JJJ » Tue Sep 11, 2018 4:24 am

Nay Lin Tun wrote:
Tue Sep 11, 2018 2:26 am
JJJ wrote:
Mon Sep 10, 2018 6:41 pm
Most test doesn't show Leela as strong as Stockfish dev at all and even below Komodo / Houdini. I find the result of this CCCC very accurate when I compare to the other test I saw.
There is another thread, ? Leela official better than SF 9.
some isolate test doesn't prove much. It wil mean something when most test show Leela stronger than SF 9.

DrCliche
Posts: 8
Joined: Sun Aug 19, 2018 8:57 pm
Full name: Nickolas Reynolds

Re: Chess.com 2018 computer chess championship

Post by DrCliche » Wed Sep 12, 2018 12:03 pm

kranium wrote:
Tue Sep 04, 2018 1:11 pm
Please contact me privately with your email for the full debug file, or more info.
Regards,
Norm

I think many viewers would appreciate public logs, post mortem, and explanation for how you are addressing whatever hardware failure occurred in game 395 "Ethereal vs. Arasan" on move 59. I've examined the position with Ethereals 10.55, 10.88, and 11.00, and can't reproduce the bizarre move selection that occurred on your hardware, with or without tablebases, with or without ponder, on one thread or eight.

Ethereal 11.00 considers 59. Nxd5 until depth 4, at which point it's never considered again from depth 5 to depth 45 (where I stopped). Depth 5 should be a matter of microseconds on your hardware, and Ethereal had the benefit of pondering on its opponent's turn. I can't figure out how it's possible for Ethereal to have actually communicated that move to your tournament interface.

Also bizarre is that evaluation, depth, nodes, time, speed, and tablebase hit statistics for Ethereal are all missing from your game archives for both 58. ... Qc8 and the fatal and irreproducible 59. Nxd5. Even for a tournament with an unexpected number of crashes, glitches, and irregularities, this seems highly irregular.

User avatar
Guenther
Posts: 3040
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Chess.com 2018 computer chess championship

Post by Guenther » Wed Sep 12, 2018 12:53 pm

DrCliche wrote:
Wed Sep 12, 2018 12:03 pm
kranium wrote:
Tue Sep 04, 2018 1:11 pm
Please contact me privately with your email for the full debug file, or more info.
Regards,
Norm

I think many viewers would appreciate public logs, post mortem, and explanation for how you are addressing whatever hardware failure occurred in game 395 "Ethereal vs. Arasan" on move 59. I've examined the position with Ethereals 10.55, 10.88, and 11.00, and can't reproduce the bizarre move selection that occurred on your hardware, with or without tablebases, with or without ponder, on one thread or eight.

Ethereal 11.00 considers 59. Nxd5 until depth 4, at which point it's never considered again from depth 5 to depth 45 (where I stopped). Depth 5 should be a matter of microseconds on your hardware, and Ethereal had the benefit of pondering on its opponent's turn. I can't figure out how it's possible for Ethereal to have actually communicated that move to your tournament interface.

Also bizarre is that evaluation, depth, nodes, time, speed, and tablebase hit statistics for Ethereal are all missing from your game archives for both 58. ... Qc8 and the fatal and irreproducible 59. Nxd5. Even for a tournament with an unexpected number of crashes, glitches, and irregularities, this seems highly irregular.
I could easily reproduce the move (59. Nxd5??) with exact depth and score too ;-)

Seroiusly, this was a surprise for me - normally I expect a depth 1 move to be a so called easy move,
but the first indication against it, is the consumed time of 6 seconds (for easy move normally 0).

The pgn and the analysis done by me obviously show that there was a hardware or tournament environment problem
and teh game shouldn't count at all...

Hopefully there aren't other such strange 'hickups' in already played games.
I could run a macro which looks for depth < x, while time > y...
(this is a check for irregular tournament conditions I had done already last year)



Ethereal 10.88 with 4 cores on my old machine (stepped through all moves in analysis to fill the hash):
(you need to scroll down to the bottom for lowest depth)

Code: Select all

exclude: none best +tail                                          
dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 35	+0.32 	78.2M  	0:23.01	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7c7 b2c1 c7b7 c1c2 b7b6 c2d3 b6a7 g1h1 a7d7 d3c3 d7c8 c3c2 c8b7 b3d3 
 34	+0.32 	55.9M  	0:16.28	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7c7 b2c1 c7b7 c1c2 b7b6 b3c3 b6b7 e3e4 b7a7 e4d5 c6d5 b4d5 a7d4 
 33	+0.32 	43.1M  	0:12.49	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7c7 b3d3 c7a5 b2c1 a5a8 g1h1 a8a7 h1h2 a7c7 h2g1 c7a5 c1c3 a5a7 
 32	+0.32 	29.5M  	0:08.54	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7d7 b2c1 d7b7 c1c2 b7a8 c2b2 a8a5 g1h2 a5c7 h2g1 c7b6 b2c2 
 31	+0.32 	19.7M  	0:05.74	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 c3b3 a4a8 b3d3 a8a4 d3c3 d6e7 c3b3 e7d7 h1g1 d7b7 c2c5 b7d7 g1h2 d7c7 h2g1 c7a5 
 30	+0.32 	13.1M  	0:03.88	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 h1g1 d6d7 c3d3 d7b7 d3b3 b7a8 c2d3 a8a5 d3c3 a5b6 c3c1 b6a7 c1b2 a7d7 
 29	+0.32 	9.78M  	0:02.90	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 h1g1 d6d7 c2d1 d7a7 d1d3 a7d7 d3b1 d7a7 b4c6 d8c6 c3c6 a4a3 g2g3 
 28	+0.32 	6.93M  	0:02.07	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 h1g1 d6e7 c3b3 e7b7 c2c3 b7b6 g1h1 b6b7 c3c2 b7a8 c2c1 a8a7 
 27	+0.32 	4.90M  	0:01.48	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1h2 e7d6 h2g1 d6e7 c3b3 e7b7 c2c3 b7b6 g1h1 b6b7 c3c2 b7a8 
 26	+0.32 	2.90M  	0:00.84	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1g1 e7a7 c2d3 a7a8 c3b3 a8a5 d3c3 a5b6 g1h1 b6b7 
 25	+0.32 	1.78M  	0:00.53	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1g1 e7d7 c2d1 d7a7 d1d3 a7b7 e3e4 d5e4 d3e4 
 24	+0.32 	1.10M  	0:00.32	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1g1 e7d7 c3d3 d7d6 d3b3 d6e6 c2c5 a4a8 
 23	+0.32 	677866	0:00.20	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 c3b3 a7a8 g1h2 a8a7 h2h1 a7c7 h1g1 c7b7 e3e4 d8e6 e4d5 c6d5 
 22	+0.32 	569281	0:00.17	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 c3b3 a7d7 b2c1 d7b7 c1c2 b7d7 b3c3 d7b7 c3d3 b7d7 g1h1 
 21	+0.32 	368770	0:00.12	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2c2 d7b7 c3d3 b7d7 d3b3 d7b7 c2c5 b7d7 g1h2 a4a7 c5c2 d7d6 h2g1 a7a4 c2c3 d6e6 
 20	+0.32 	99153  	0:00.04	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2c2 d7b7 c3d3 b7d7 d3b3 d7b7 c2b2 b7d7 b4d3 d8b7 b3c3 f7f6 b2c1 a4c4 
 19	+0.32 	81291  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7a8 c2c1 a8b7 c1b2 b7d7 b4d3 d7a7 d3b4 a7a8 g1h2 
 18	+0.32 	59432  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a5 c2b2 a5b6 b2c3 b6a7 c3b3 a7c7 
 17	+0.32 	47336  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a5 c2b2 a5c7 d3c3 c7b7 c3d3 
 16	+0.32 	41667  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a7 c2b3 a7c7 b3d1 c7a7 
 15	+0.32 	35069  	0:00.01	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a7 c2b3 a7c7 b3d1 
 14	+0.32 	28915  	0:00.01	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a7 c2b3 a7c7 
 13	+0.32 	21708  	0:00.01	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 g1h2 g8g7 b2c2 b7c7 h2h1 c7d7 h1g1 d7e7 
 12	+0.32 	15375  	0:00.01	b3c3 c8b7 c3b3 b7d7 h1g1 a4a7 c1c2 a7a4 b3c3 d7e7 c3d3 e7a7 
 11	+0.32 	11493  	0:00.01	b3c3 c8b7 c3b3 b7d7 c1c2 a4a7 b3c3 a7a4 h1g1 d7e7 c3d3 
 10	+0.32 	9186    	0:00.01	b3c3 c8b7 c3b3 b7d7 c1c2 a4a8 b3c3 a8a4 h1g1 d7e7 
  9	+0.32 	7763    	0:00.01	b3c3 c8b7 h1h2 b7a8 c1b2 g7h7 h2g1 a8b7 b4d3 
  8	+0.32 	5803    	0:00.01	b3c3 c8b7 c3d3 b7a7 h1g1 a7a5 c1b2 a5c7 
  7	+0.32 	5466    	0:00.01	b3c3 c8b7 c3d3 b7a7 h1g1 a7a5 c1b2 
  6	+0.98 	4552    	0:00.01	c1c5 c8b7 c5d6 b7a8 d6e5 g7g8 e5e7 a4a3 
  5	+0.88 	1357    	0:00.01	c1c5 a4a8 c5e7 d8e6 e7h4 
  4	+1.08 	615      	0:00.01	c1c5 d8e6 c5c6 c8c6 b4c6 
  3	+1.21 	293      	0:00.01	b3c3 c8c7 b4c6 
  2	+1.21 	197      	0:00.01	b3c3 c8c7 
  1	+0.44 	57        	0:00.01	b4d5 
Current foe list count : [97]
http://rwbc-chess.de/chronology.htm

AndrewGrant
Posts: 484
Joined: Tue Apr 19, 2016 4:08 am
Location: U.S.A
Full name: Andrew Grant
Contact:

Re: Chess.com 2018 computer chess championship

Post by AndrewGrant » Wed Sep 12, 2018 1:00 pm

Guenther wrote:
Wed Sep 12, 2018 12:53 pm
DrCliche wrote:
Wed Sep 12, 2018 12:03 pm
kranium wrote:
Tue Sep 04, 2018 1:11 pm
Please contact me privately with your email for the full debug file, or more info.
Regards,
Norm

I think many viewers would appreciate public logs, post mortem, and explanation for how you are addressing whatever hardware failure occurred in game 395 "Ethereal vs. Arasan" on move 59. I've examined the position with Ethereals 10.55, 10.88, and 11.00, and can't reproduce the bizarre move selection that occurred on your hardware, with or without tablebases, with or without ponder, on one thread or eight.

Ethereal 11.00 considers 59. Nxd5 until depth 4, at which point it's never considered again from depth 5 to depth 45 (where I stopped). Depth 5 should be a matter of microseconds on your hardware, and Ethereal had the benefit of pondering on its opponent's turn. I can't figure out how it's possible for Ethereal to have actually communicated that move to your tournament interface.

Also bizarre is that evaluation, depth, nodes, time, speed, and tablebase hit statistics for Ethereal are all missing from your game archives for both 58. ... Qc8 and the fatal and irreproducible 59. Nxd5. Even for a tournament with an unexpected number of crashes, glitches, and irregularities, this seems highly irregular.
I could easily reproduce the move (59. Nxd5??) with exact depth and score too ;-)

Seroiusly, this was a surprise for me - normally I expect a depth 1 move to be a so called easy move,
but the first indication against it, is the consumed time of 6 seconds (for easy move normally 0).

The pgn and the analysis done by me obviously show that there was a hardware or tournament environment problem
and teh game shouldn't count at all...

Hopefully there aren't other such strange 'hickups' in already played games.
I could run a macro which looks for depth < x, while time > y...
(this is a check for irregular tournament conditions I had done already last year)



Ethereal 10.88 with 4 cores on my old machine (stepped through all moves in analysis to fill the hash):
(you need to scroll down to the bottom for lowest depth)

Code: Select all

exclude: none best +tail                                          
dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 35	+0.32 	78.2M  	0:23.01	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7c7 b2c1 c7b7 c1c2 b7b6 c2d3 b6a7 g1h1 a7d7 d3c3 d7c8 c3c2 c8b7 b3d3 
 34	+0.32 	55.9M  	0:16.28	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7c7 b2c1 c7b7 c1c2 b7b6 b3c3 b6b7 e3e4 b7a7 e4d5 c6d5 b4d5 a7d4 
 33	+0.32 	43.1M  	0:12.49	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7c7 b3d3 c7a5 b2c1 a5a8 g1h1 a8a7 h1h2 a7c7 h2g1 c7a5 c1c3 a5a7 
 32	+0.32 	29.5M  	0:08.54	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 c3d3 b7a8 b3c3 a8b7 c3b2 b7a8 b2b3 a4a7 d3c3 a7a4 b3b2 a8a7 c3b3 a7d7 b2c1 d7b7 c1c2 b7a8 c2b2 a8a5 g1h2 a5c7 h2g1 c7b6 b2c2 
 31	+0.32 	19.7M  	0:05.74	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 c3b3 a4a8 b3d3 a8a4 d3c3 d6e7 c3b3 e7d7 h1g1 d7b7 c2c5 b7d7 g1h2 d7c7 h2g1 c7a5 
 30	+0.32 	13.1M  	0:03.88	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 h1g1 d6d7 c3d3 d7b7 d3b3 b7a8 c2d3 a8a5 d3c3 a5b6 c3c1 b6a7 c1b2 a7d7 
 29	+0.32 	9.78M  	0:02.90	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 h1g1 d6d7 c2d1 d7a7 d1d3 a7d7 d3b1 d7a7 b4c6 d8c6 c3c6 a4a3 g2g3 
 28	+0.32 	6.93M  	0:02.07	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7d6 h1g1 d6e7 c3b3 e7b7 c2c3 b7b6 g1h1 b6b7 c3c2 b7a8 c2c1 a8a7 
 27	+0.32 	4.90M  	0:01.48	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1h2 e7d6 h2g1 d6e7 c3b3 e7b7 c2c3 b7b6 g1h1 b6b7 c3c2 b7a8 
 26	+0.32 	2.90M  	0:00.84	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1g1 e7a7 c2d3 a7a8 c3b3 a8a5 d3c3 a5b6 g1h1 b6b7 
 25	+0.32 	1.78M  	0:00.53	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1g1 e7d7 c2d1 d7a7 d1d3 a7b7 e3e4 d5e4 d3e4 
 24	+0.32 	1.10M  	0:00.32	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 g1h1 a7d7 b2c2 d7e7 h1g1 e7d7 c3d3 d7d6 d3b3 d6e6 c2c5 a4a8 
 23	+0.32 	677866	0:00.20	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 c3b3 a7a8 g1h2 a8a7 h2h1 a7c7 h1g1 c7b7 e3e4 d8e6 e4d5 c6d5 
 22	+0.32 	569281	0:00.17	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2b3 d7b7 b3c2 b7a8 c2b2 a8a7 c3b3 a7d7 b2c1 d7b7 c1c2 b7d7 b3c3 d7b7 c3d3 b7d7 g1h1 
 21	+0.32 	368770	0:00.12	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2c2 d7b7 c3d3 b7d7 d3b3 d7b7 c2c5 b7d7 g1h2 a4a7 c5c2 d7d6 h2g1 a7a4 c2c3 d6e6 
 20	+0.32 	99153  	0:00.04	b3c3 c8d7 c1b2 d7a7 h1g1 a7d7 b2c2 d7b7 c3d3 b7d7 d3b3 d7b7 c2b2 b7d7 b4d3 d8b7 b3c3 f7f6 b2c1 a4c4 
 19	+0.32 	81291  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7a8 c2c1 a8b7 c1b2 b7d7 b4d3 d7a7 d3b4 a7a8 g1h2 
 18	+0.32 	59432  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a5 c2b2 a5b6 b2c3 b6a7 c3b3 a7c7 
 17	+0.32 	47336  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a5 c2b2 a5c7 d3c3 c7b7 c3d3 
 16	+0.32 	41667  	0:00.03	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a7 c2b3 a7c7 b3d1 c7a7 
 15	+0.32 	35069  	0:00.01	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a7 c2b3 a7c7 b3d1 
 14	+0.32 	28915  	0:00.01	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 c3b3 g8g7 b2c2 b7b6 b3d3 b6a7 c2b3 a7c7 
 13	+0.32 	21708  	0:00.01	b3c3 c8d7 c1b2 d7b7 h1g1 g7g8 g1h2 g8g7 b2c2 b7c7 h2h1 c7d7 h1g1 d7e7 
 12	+0.32 	15375  	0:00.01	b3c3 c8b7 c3b3 b7d7 h1g1 a4a7 c1c2 a7a4 b3c3 d7e7 c3d3 e7a7 
 11	+0.32 	11493  	0:00.01	b3c3 c8b7 c3b3 b7d7 c1c2 a4a7 b3c3 a7a4 h1g1 d7e7 c3d3 
 10	+0.32 	9186    	0:00.01	b3c3 c8b7 c3b3 b7d7 c1c2 a4a8 b3c3 a8a4 h1g1 d7e7 
  9	+0.32 	7763    	0:00.01	b3c3 c8b7 h1h2 b7a8 c1b2 g7h7 h2g1 a8b7 b4d3 
  8	+0.32 	5803    	0:00.01	b3c3 c8b7 c3d3 b7a7 h1g1 a7a5 c1b2 a5c7 
  7	+0.32 	5466    	0:00.01	b3c3 c8b7 c3d3 b7a7 h1g1 a7a5 c1b2 
  6	+0.98 	4552    	0:00.01	c1c5 c8b7 c5d6 b7a8 d6e5 g7g8 e5e7 a4a3 
  5	+0.88 	1357    	0:00.01	c1c5 a4a8 c5e7 d8e6 e7h4 
  4	+1.08 	615      	0:00.01	c1c5 d8e6 c5c6 c8c6 b4c6 
  3	+1.21 	293      	0:00.01	b3c3 c8c7 b4c6 
  2	+1.21 	197      	0:00.01	b3c3 c8c7 
  1	+0.44 	57        	0:00.01	b4d5 
Following right after, Nemorino vs Laser, Nemorino hangs a Queen for capture by a pawn, turning a 40+ply 0.00 into a mate in 6. Bizarre stuff.

AndrewGrant
Posts: 484
Joined: Tue Apr 19, 2016 4:08 am
Location: U.S.A
Full name: Andrew Grant
Contact:

Re: Chess.com 2018 computer chess championship

Post by AndrewGrant » Wed Sep 12, 2018 1:32 pm

DrCliche wrote:
Wed Sep 12, 2018 12:03 pm
kranium wrote:
Tue Sep 04, 2018 1:11 pm
Please contact me privately with your email for the full debug file, or more info.
Regards,
Norm

I think many viewers would appreciate public logs, post mortem, and explanation for how you are addressing whatever hardware failure occurred in game 395 "Ethereal vs. Arasan" on move 59. I've examined the position with Ethereals 10.55, 10.88, and 11.00, and can't reproduce the bizarre move selection that occurred on your hardware, with or without tablebases, with or without ponder, on one thread or eight.

Ethereal 11.00 considers 59. Nxd5 until depth 4, at which point it's never considered again from depth 5 to depth 45 (where I stopped). Depth 5 should be a matter of microseconds on your hardware, and Ethereal had the benefit of pondering on its opponent's turn. I can't figure out how it's possible for Ethereal to have actually communicated that move to your tournament interface.

Also bizarre is that evaluation, depth, nodes, time, speed, and tablebase hit statistics for Ethereal are all missing from your game archives for both 58. ... Qc8 and the fatal and irreproducible 59. Nxd5. Even for a tournament with an unexpected number of crashes, glitches, and irregularities, this seems highly irregular.
From the logs provided to me ...

info depth 1 seldepth 3 score cp 44 time 6093 nodes 202659512 nps 33255000 tbhits 2067 hashfull 53 pv b4d5

6093ms for a depth 1 search? It looks to me like the main thread got locked out of CPU time, and probably some of the helpers.

33mnps is < 50 % of the average NPS for the recent moves. A depth 1 search with seldepth 3 takes < 1ms

Naturally I blame pondering on hyperthreads. Windows is known to box out threads, and Intel is known to have poor hyperthreading overall.


Some Ethereal specifics ...
above output is reported as soon as the depth 1 search finishes. So that took 6 seconds. Its not like it finished and got hung up on something else, only to report as time ran out.
Seldepth is tracked through all search stages, so truly no line was explored more then 3 moves deep. Thus, there is no magic explosion of the search tree causing a massive slowdown or anything of the like.

DrCliche
Posts: 8
Joined: Sun Aug 19, 2018 8:57 pm
Full name: Nickolas Reynolds

Re: Chess.com 2018 computer chess championship

Post by DrCliche » Wed Sep 12, 2018 1:55 pm

AndrewGrant wrote:
Wed Sep 12, 2018 1:00 pm
Following right after, Nemorino vs Laser, Nemorino hangs a Queen for capture by a pawn, turning a 40+ply 0.00 into a mate in 6. Bizarre stuff.

I didn't think it possible, but CCCC has reached profound new heights of absurdity. Shall I save kranium the trouble of responding?

kranium wrote:
Tue Aug 21, 2018 3:43 pm
I do hope you can enjoy it instead of picking it apart, leveling accusations, and being so critical of the effort. I certainly realize there's a lot of 'haters', and egotistical people who believe they could do much better, but please don't forget: these events are being produced, funded, and broadcast for free by chess.com for the benefit of community...as the amount of new subscribers necessary to pay for it all is astronomical.
kranium wrote:
Wed Aug 22, 2018 10:26 pm
You know, there are are many types of people...ones that jump in, work hard, and make something happen...
ones that graciously accept what's been given for free, and make the best of it and enjoy it...
and ones that stand on the sidelines, waving their arms, complaining about everything.

Post Reply