Discussion of computer chess matches and engine tournaments.
Moderators: hgm , Rebel , chrisw
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Tue Dec 27, 2016 8:52 pm
It seemed to me, that Komodo 10.3 plays some better against SF with contempt 0
Of course I knew about it, but it's interesting to see how it does with default settings. I decided to check how it plays against SF with contempt -10.
After 40 games:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 18 55 53 40 52.5 % 0 75.0 %
2 Komodo 10.3 64-bit C-10 : 0 53 55 40 47.5 % 18 75.0 %
Individual statistics:
1 Stockfish 221216 64 POPCNT : 18 40 (+ 6,= 30,- 4), 52.5 %
2 Komodo 10.3 64-bit C-10 : 0 40 (+ 4,= 30,- 6), 47.5 %
Games
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Tue Dec 27, 2016 8:58 pm
Conditions:
TC 120"+2", 1 core per engine, hash - 2048 MB, each position will played with both colours. Komodo 10.3 plays with contempt -10, Stockfish 221216 - with default.
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Tue Dec 27, 2016 10:12 pm
After 70 games with C10 K10.3 has about
-100 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 : 102 54 51 70 64.3 % 0 57.1 %
2 Komodo 10.3 64-bit : 0 51 54 70 35.7 % 102 57.1 %
After 280 games with C0 K10.3 has about
-60 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 64 23 22 280 59.1 % 0 68.2 %
2 Komodo 10.3 64-bit : 0 22 23 280 40.9 % 64 68.2 %
After 52 games with C-10 K10.3 has
-14 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 14 46 45 52 51.9 % 0 76.9 %
2 Komodo 10.3 64-bit : 0 45 46 52 48.1 % 14 76.9 %
Let's see what will happen next!
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Wed Dec 28, 2016 12:51 pm
After 80 games:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 34 39 37 80 55.0 % 0 75.0 %
2 Komodo 10.3 64-bit : 0 37 39 80 45.0 % 34 75.0 %
Individual statistics:
1 Stockfish 221216 64 POPCNT : 34 80 (+ 14,= 60,- 6), 55.0 %
2 Komodo 10.3 64-bit : 0 80 (+ 6,= 60,- 14), 45.0 %
Games
Nay Lin Tun
Posts: 708 Joined: Mon Jan 16, 2012 6:34 am
Post
by Nay Lin Tun » Thu Dec 29, 2016 8:13 am
majortom wrote: After 70 games with C10 K10.3 has about
-100 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 : 102 54 51 70 64.3 % 0 57.1 %
2 Komodo 10.3 64-bit : 0 51 54 70 35.7 % 102 57.1 %
After 280 games with C0 K10.3 has about
-60 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 64 23 22 280 59.1 % 0 68.2 %
2 Komodo 10.3 64-bit : 0 22 23 280 40.9 % 64 68.2 %
After 52 games with C-10 K10.3 has
-14 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 14 46 45 52 51.9 % 0 76.9 %
2 Komodo 10.3 64-bit : 0 45 46 52 48.1 % 14 76.9 %
Let's see what will happen next!
So, if the contempt is C-20, Komodo might possibly win Stockfish.
Dr.Wael Deeb
Posts: 9773 Joined: Wed Mar 08, 2006 8:44 pm
Location: Amman,Jordan
Post
by Dr.Wael Deeb » Thu Dec 29, 2016 8:38 am
Nay Lin Tun wrote: majortom wrote: After 70 games with C10 K10.3 has about
-100 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 : 102 54 51 70 64.3 % 0 57.1 %
2 Komodo 10.3 64-bit : 0 51 54 70 35.7 % 102 57.1 %
After 280 games with C0 K10.3 has about
-60 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 64 23 22 280 59.1 % 0 68.2 %
2 Komodo 10.3 64-bit : 0 22 23 280 40.9 % 64 68.2 %
After 52 games with C-10 K10.3 has
-14 elo vs. SF:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 14 46 45 52 51.9 % 0 76.9 %
2 Komodo 10.3 64-bit : 0 45 46 52 48.1 % 14 76.9 %
Let's see what will happen next!
So, if the contempt is C-20, Komodo might possibly win Stockfish.
Most definitely
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Thu Dec 29, 2016 10:03 am
After 160 games:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 44 30 29 160 56.2 % 0 70.0 %
2 Komodo 10.3 64-bit C-10 : 0 29 30 160 43.8 % 44 70.0 %
Individual statistics:
1 Stockfish 221216 64 POPCNT : 44 160 (+ 34,=112,- 14), 56.2 %
2 Komodo 10.3 64-bit C-10 : 0 160 (+ 14,=112,- 34), 43.8 %
Games
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Fri Dec 30, 2016 3:59 pm
Match is finished:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 48 22 21 300 57.0 % 0 69.3 %
2 Komodo 10.3 64-bit C-10 : 0 21 22 300 43.0 % 48 69.3 %
Individual statistics:
1 Stockfish 221216 64 POPCNT : 48 300 (+ 67,=208,- 25), 57.0 %
2 Komodo 10.3 64-bit C-10 : 0 300 (+ 25,=208,- 67), 43.0 %
Games
mjlef
Posts: 1494 Joined: Thu Mar 30, 2006 2:08 pm
Post
by mjlef » Fri Dec 30, 2016 4:42 pm
majortom wrote: Match is finished:
Code: Select all
Program Elo + - Games Score Av.Op. Draws
1 Stockfish 221216 64 POPCNT : 48 22 21 300 57.0 % 0 69.3 %
2 Komodo 10.3 64-bit C-10 : 0 21 22 300 43.0 % 48 69.3 %
Individual statistics:
1 Stockfish 221216 64 POPCNT : 48 300 (+ 67,=208,- 25), 57.0 %
2 Komodo 10.3 64-bit C-10 : 0 300 (+ 25,=208,- 67), 43.0 %
Games
Thanks for running these. It seems Contempt does its job well at this time control where Stockfish is ahead.
If you have a chance to do any MP games, that would be very interesting. We think we have improved Komodo's MP scaling quite a bit, but it is hard to get enough games to know how much.
majortom
Posts: 669 Joined: Mon Nov 04, 2013 10:19 pm
Post
by majortom » Fri Dec 30, 2016 6:17 pm
mjlef wrote: If you have a chance to do any MP games, that would be very interesting. We think we have improved Komodo's MP scaling quite a bit, but it is hard to get enough games to know how much.
Yesterday I've ran the 2-core match with TC 2000"+15". New Komodo's contempt value is -20. Maybe 2 cores are not enouch to check MP scaling, but anyway let's see.
http://www.talkchess.com/forum/viewtopic.php?t=62660