Discussion of computer chess matches and engine tournaments.
Moderators: hgm , Rebel , chrisw
beram
Posts: 1187 Joined: Wed Jan 06, 2010 3:11 pm
Post
by beram » Sun Jul 30, 2017 9:39 am
I have played a match with Noomen2016 suite on my AMDRyzen 14t TC3m2s between latest Komodo 11.2.2 and 'old' asmFishW170522 to compare the results with Komodo 11.01 in this thread:
http://www.talkchess.com/forum/viewtopi ... 18&t=64176
Outcome now is a 58,5% win for asmFish where it was 62,5% so 4% better result for Komodo 11.2 (about 28ELO) +27 -10 =63
(against K11.01 it was +33 -8 =59)
Code: Select all
asmFishW170522-K11.2, Blitz 3m+2s
1 asmfishw_2017-05-22_popcnt +60 +27/=63/-10 58.50% 58.5/100
2 Komodo 11.2 64-bit -60 +10/=63/-27 41.50% 41.5/100
JJJ
Posts: 1346 Joined: Sat Apr 19, 2014 1:47 pm
Post
by JJJ » Sun Jul 30, 2017 4:17 pm
Funny how can 100 games show an accurate progress so often.
Lyudmil Tsvetkov
Posts: 6052 Joined: Tue Jun 12, 2012 12:41 pm
Post
by Lyudmil Tsvetkov » Sun Jul 30, 2017 6:18 pm
JJJ wrote: Funny how can 100 games show an accurate progress so often.
a small number of games might still be sufficiently accurate, if you include a wide variety of openings.
large number of games with fewer openings, on the other hand, might be well biassed.
Jeroen
Posts: 501 Joined: Wed Mar 08, 2006 9:49 pm
Post
by Jeroen » Sun Jul 30, 2017 6:18 pm
Hi Bram,
Can you post a link to the PGN? Thanks!
Jeroen
leavenfish
Posts: 282 Joined: Mon Sep 02, 2013 8:23 am
Post
by leavenfish » Sun Jul 30, 2017 6:29 pm
My thoughts are that the latest Komodo is always 30-50pts behind the latest Stockfish or derivative. Is that a fair statement?
Lyudmil Tsvetkov
Posts: 6052 Joined: Tue Jun 12, 2012 12:41 pm
Post
by Lyudmil Tsvetkov » Sun Jul 30, 2017 6:34 pm
leavenfish wrote: My thoughts are that the latest Komodo is always 30-50pts behind the latest Stockfish or derivative. Is that a fair statement?
that is one thing we don't know, because most rating lists will not test latest development SF vs latest released Komodo.
but I guess SF still has the edge over Komodo(with which Mark and Larry will promptly disagree, of course
)
Guenther
Posts: 4606 Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Post
by Guenther » Sun Jul 30, 2017 7:25 pm
Lyudmil Tsvetkov wrote: leavenfish wrote: My thoughts are that the latest Komodo is always 30-50pts behind the latest Stockfish or derivative. Is that a fair statement?
that is one thing we don't know, because most rating lists will not test latest development SF vs latest released Komodo.
wrong as usual:
Code: Select all
http://spcc.beepworld.de/long-thinkingtime.htm
JJJ
Posts: 1346 Joined: Sat Apr 19, 2014 1:47 pm
Post
by JJJ » Sun Jul 30, 2017 8:51 pm
leavenfish wrote: My thoughts are that the latest Komodo is always 30-50pts behind the latest Stockfish or derivative. Is that a fair statement?
In direct match I believe Komodo was more than 50 elo behind Stockfish dev. Now I believe Komodo has closed the gap to something around 40.
Lyudmil Tsvetkov
Posts: 6052 Joined: Tue Jun 12, 2012 12:41 pm
Post
by Lyudmil Tsvetkov » Sun Jul 30, 2017 8:56 pm
Guenther wrote: Lyudmil Tsvetkov wrote: leavenfish wrote: My thoughts are that the latest Komodo is always 30-50pts behind the latest Stockfish or derivative. Is that a fair statement?
that is one thing we don't know, because most rating lists will not test latest development SF vs latest released Komodo.
wrong as usual:
Code: Select all
http://spcc.beepworld.de/long-thinkingtime.htm
error bars for those tests?
widely published?
official?
you know what a general statement means?
beram
Posts: 1187 Joined: Wed Jan 06, 2010 3:11 pm
Post
by beram » Sun Jul 30, 2017 10:15 pm
Jeroen wrote: Hi Bram,
Can you post a link to the PGN? Thanks!
Jeroen
hereby the downloadlink
https://we.tl/vCExKOKwVz
I have played two other matches against Komodo 11.2 with latest Stockfish dev(ultimaq compile)170725
Result with Noomen suite 2016 58,5% and with my own testsuite (more Komodo friendly
) 2x50 games 53,5%
So overall over 200 games 56% which means 42 ELO difference
(Jean Baptiste's guess was about 40 ELO difference)
games link for thes games
https://we.tl/7cGcoTUcPf