Page 10 of 12

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sat Nov 07, 2020 7:40 am
by Marcus9
lkaufman wrote: Sat Nov 07, 2020 6:17 am
Gabor Szots wrote: Thu Nov 05, 2020 2:52 pm
pohl4711 wrote: Thu Nov 05, 2020 8:09 am Why changing the name?!? That is never a good idea IMHO. Komodo is a long-used and well known name with reputation. And because Komodo is commercial, it is a brand, too.
Whats so bad about "Komodo 15 NNUE" ? Or just Komodo 15 ? Stockfish is still named Stockfish, even though it is using nnue-nets, now and had a huge Elo gain by using nnue, too.
I agree. Simply Komodo 15. And better make NNUE default (or not a parameter at all) lest we have to use 'Komodo 15 NNUE on 64-bit' in the name which looks awful.
We can't call it Komodo 15 due to contractual commitments, and "NNUE" is too technical and obscure for ordinary chess players. Also chess.com (which had a major role in this) really likes the Dragon name. So the program will say "Dragon by Komodo Chess", but I expect the testing groups (but not chess.com) will just call it Komodo Dragon (or Dragon-Komodo). As long as the name includes both Komodo and Dragon, everyone should be happy, there is no confusion with either normal Komodo or with some other Dragon engine, and the Dragon name means that it is the NNUE version. It will have an embedded net which will be on by default (user can turn it off).
We plan to have some test games between Dragon and my son IM Raymond Kaufman on chess.com this afternoon (Saturday) at 5 pm Eastern Standard time, probably at knight odds. Look for "playkomodo" games. This is partly to look for any unexpected problems, and also to try to determine whether standard mode or MCTS mode is more appropriate for Dragon when giving handicaps to humans. This is not a formal event, just a test.
Thanks, I'll watch komodo matches against your son.
Where it will be specified if MCTS mode is active or not?

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sat Nov 07, 2020 6:04 pm
by lkaufman
Marcus9 wrote: Sat Nov 07, 2020 7:40 am
lkaufman wrote: Sat Nov 07, 2020 6:17 am
Gabor Szots wrote: Thu Nov 05, 2020 2:52 pm
pohl4711 wrote: Thu Nov 05, 2020 8:09 am Why changing the name?!? That is never a good idea IMHO. Komodo is a long-used and well known name with reputation. And because Komodo is commercial, it is a brand, too.
Whats so bad about "Komodo 15 NNUE" ? Or just Komodo 15 ? Stockfish is still named Stockfish, even though it is using nnue-nets, now and had a huge Elo gain by using nnue, too.
I agree. Simply Komodo 15. And better make NNUE default (or not a parameter at all) lest we have to use 'Komodo 15 NNUE on 64-bit' in the name which looks awful.
We can't call it Komodo 15 due to contractual commitments, and "NNUE" is too technical and obscure for ordinary chess players. Also chess.com (which had a major role in this) really likes the Dragon name. So the program will say "Dragon by Komodo Chess", but I expect the testing groups (but not chess.com) will just call it Komodo Dragon (or Dragon-Komodo). As long as the name includes both Komodo and Dragon, everyone should be happy, there is no confusion with either normal Komodo or with some other Dragon engine, and the Dragon name means that it is the NNUE version. It will have an embedded net which will be on by default (user can turn it off).
We plan to have some test games between Dragon and my son IM Raymond Kaufman on chess.com this afternoon (Saturday) at 5 pm Eastern Standard time, probably at knight odds. Look for "playkomodo" games. This is partly to look for any unexpected problems, and also to try to determine whether standard mode or MCTS mode is more appropriate for Dragon when giving handicaps to humans. This is not a formal event, just a test.
Thanks, I'll watch komodo matches against your son.
Where it will be specified if MCTS mode is active or not?
Maybe I'll post that info here.

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sat Nov 07, 2020 11:01 pm
by Ozymandias
lkaufman wrote: Sat Nov 07, 2020 6:17 amthe Dragon name means that it is the NNUE version. It will have an embedded net which will be on by default
Will it only use that net for evaluation or will it be mixed with classical, like SF?

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sat Nov 07, 2020 11:20 pm
by lkaufman
Ozymandias wrote: Sat Nov 07, 2020 11:01 pm
lkaufman wrote: Sat Nov 07, 2020 6:17 amthe Dragon name means that it is the NNUE version. It will have an embedded net which will be on by default
Will it only use that net for evaluation or will it be mixed with classical, like SF?
Well, Komodo is different from SF in how we do lazy eval, so the short answer is that technically it is mixed, but practically the role of classical Komodo eval is minimal in Dragon, it is much closer to pure net eval than is the case with SF.

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sat Nov 07, 2020 11:58 pm
by lkaufman
Marcus9 wrote: Sat Nov 07, 2020 7:40 am
lkaufman wrote: Sat Nov 07, 2020 6:17 am
Gabor Szots wrote: Thu Nov 05, 2020 2:52 pm
pohl4711 wrote: Thu Nov 05, 2020 8:09 am Why changing the name?!? That is never a good idea IMHO. Komodo is a long-used and well known name with reputation. And because Komodo is commercial, it is a brand, too.
Whats so bad about "Komodo 15 NNUE" ? Or just Komodo 15 ? Stockfish is still named Stockfish, even though it is using nnue-nets, now and had a huge Elo gain by using nnue, too.
I agree. Simply Komodo 15. And better make NNUE default (or not a parameter at all) lest we have to use 'Komodo 15 NNUE on 64-bit' in the name which looks awful.
We can't call it Komodo 15 due to contractual commitments, and "NNUE" is too technical and obscure for ordinary chess players. Also chess.com (which had a major role in this) really likes the Dragon name. So the program will say "Dragon by Komodo Chess", but I expect the testing groups (but not chess.com) will just call it Komodo Dragon (or Dragon-Komodo). As long as the name includes both Komodo and Dragon, everyone should be happy, there is no confusion with either normal Komodo or with some other Dragon engine, and the Dragon name means that it is the NNUE version. It will have an embedded net which will be on by default (user can turn it off).
We plan to have some test games between Dragon and my son IM Raymond Kaufman on chess.com this afternoon (Saturday) at 5 pm Eastern Standard time, probably at knight odds. Look for "playkomodo" games. This is partly to look for any unexpected problems, and also to try to determine whether standard mode or MCTS mode is more appropriate for Dragon when giving handicaps to humans. This is not a formal event, just a test.
Thanks, I'll watch komodo matches against your son.
Where it will be specified if MCTS mode is active or not?
Ray couldn't play tonite, so I'm playing it myself, first two games vs. standard Dragon, then after a break two games vs. MCTS Dragon (all knight odds). I lost the first one.

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sun Nov 08, 2020 2:43 am
by supersharp77
lkaufman wrote: Mon Nov 02, 2020 6:08 pm Komodo 14.1 Release and Dragon Announcement

Komodo 14.1 is released today at komodochess.com. It is a normal strength upgrade from Komodo 14,
about ten elo in normal mode and twelve in MCTS mode. It is free for subscribers and also this time for
anyone who bought Komodo 14. It will not run on processors older than the year 2008. Some of the elo gains
are from speedups, some from parameter tuning, LMR changes, revised aspiration, revised time management, etc.

We are also announcing our new "Dragon" version of Komodo, which is now playing in the chess.com CCC
tournament as "Mystery" and which we expect to release soon. It uses the new NNUE technology that was
developed for the game of shogi, but not the NNUE code. The search is Komodo search (with some parameters
tuned), and the nets we use are all trained on Komodo games and Komodo evals. The net is embedded so the
user need not do anything special to use it (though it can be turned off).



Special thanks to Dietrich Kappe for his expertise in training the nets, to Dmitry Pervov of chess.com
for writing the NNUE code we use and for many other enhancements to Dragon, and to our webmaster Jesse Gersenson
who unleashed hundreds of computers to generate the billions of positions needed to train the nets.
What compiles (how many) will be available for testing for your new engines........BMI?, Popcnt? Npopcnt? You have already stopped producing x32 compiles for your engines.. what is the reason for limiting the available compiles for the chess engine testing community? Thx AR :) :wink:

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sun Nov 08, 2020 4:50 am
by lkaufman
supersharp77 wrote: Sun Nov 08, 2020 2:43 am
lkaufman wrote: Mon Nov 02, 2020 6:08 pm Komodo 14.1 Release and Dragon Announcement

Komodo 14.1 is released today at komodochess.com. It is a normal strength upgrade from Komodo 14,
about ten elo in normal mode and twelve in MCTS mode. It is free for subscribers and also this time for
anyone who bought Komodo 14. It will not run on processors older than the year 2008. Some of the elo gains
are from speedups, some from parameter tuning, LMR changes, revised aspiration, revised time management, etc.

We are also announcing our new "Dragon" version of Komodo, which is now playing in the chess.com CCC
tournament as "Mystery" and which we expect to release soon. It uses the new NNUE technology that was
developed for the game of shogi, but not the NNUE code. The search is Komodo search (with some parameters
tuned), and the nets we use are all trained on Komodo games and Komodo evals. The net is embedded so the
user need not do anything special to use it (though it can be turned off).



Special thanks to Dietrich Kappe for his expertise in training the nets, to Dmitry Pervov of chess.com
for writing the NNUE code we use and for many other enhancements to Dragon, and to our webmaster Jesse Gersenson
who unleashed hundreds of computers to generate the billions of positions needed to train the nets.
What compiles (how many) will be available for testing for your new engines........BMI?, Popcnt? Npopcnt? You have already stopped producing x32 compiles for your engines.. what is the reason for limiting the available compiles for the chess engine testing community? Thx AR :) :wink:
For Windows, I think it will just be two: with/without avx2. We don't like to make extra compiles unless a significant number of people want them and will benefit from them meaningfully. avx2 appears to be critical, but BMI doesn't appear to offer much further benefit if any. x32 is too obsolete to be worth the trouble, almost no one requests it.

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sun Nov 08, 2020 6:40 am
by Frank Quisinsky
Hi Larry,

I will have it ... the good and old w32 compile from Komodo.

No, no ... I am joking!

Hiarcs is a good example:
Mark Uniacke made a x64 compile and goes back to w32 for many years.
Since a longer time Hiarcs engine isn't updated but Chess Explorer GUI.

Chess Explorer: A GUI can handle the PGN standards in perfection!
Since few days a newer version of Chess Explorer is available.
A nice work Mark do here, but it seems Mark is not very active in the last years.

Nibbler GUI is interesting for you ...
If you offer your Komodo as Komodo-Dragon NNUE or Dragon-Komodo NNUE version!

Best
Frank

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sun Nov 08, 2020 6:29 pm
by Hurnavich
Hi, I'm now running AMD Ryzen 3950x 16/32 and AMD Ryzen 3900x 12/24 and find AVX2 in most engines is much faster on my set ups.So would be nice to see AVX2 compiles for Ryzen...

Re: Komodo 14.1 Release and Dragon Announcement

Posted: Sun Nov 08, 2020 8:32 pm
by lkaufman
lkaufman wrote: Sat Nov 07, 2020 11:58 pm
Marcus9 wrote: Sat Nov 07, 2020 7:40 am
lkaufman wrote: Sat Nov 07, 2020 6:17 am
Gabor Szots wrote: Thu Nov 05, 2020 2:52 pm
pohl4711 wrote: Thu Nov 05, 2020 8:09 am Why changing the name?!? That is never a good idea IMHO. Komodo is a long-used and well known name with reputation. And because Komodo is commercial, it is a brand, too.
Whats so bad about "Komodo 15 NNUE" ? Or just Komodo 15 ? Stockfish is still named Stockfish, even though it is using nnue-nets, now and had a huge Elo gain by using nnue, too.
I agree. Simply Komodo 15. And better make NNUE default (or not a parameter at all) lest we have to use 'Komodo 15 NNUE on 64-bit' in the name which looks awful.
We can't call it Komodo 15 due to contractual commitments, and "NNUE" is too technical and obscure for ordinary chess players. Also chess.com (which had a major role in this) really likes the Dragon name. So the program will say "Dragon by Komodo Chess", but I expect the testing groups (but not chess.com) will just call it Komodo Dragon (or Dragon-Komodo). As long as the name includes both Komodo and Dragon, everyone should be happy, there is no confusion with either normal Komodo or with some other Dragon engine, and the Dragon name means that it is the NNUE version. It will have an embedded net which will be on by default (user can turn it off).
We plan to have some test games between Dragon and my son IM Raymond Kaufman on chess.com this afternoon (Saturday) at 5 pm Eastern Standard time, probably at knight odds. Look for "playkomodo" games. This is partly to look for any unexpected problems, and also to try to determine whether standard mode or MCTS mode is more appropriate for Dragon when giving handicaps to humans. This is not a formal event, just a test.
Thanks, I'll watch komodo matches against your son.
Where it will be specified if MCTS mode is active or not?
Ray couldn't play tonite, so I'm playing it myself, first two games vs. standard Dragon, then after a break two games vs. MCTS Dragon (all knight odds). I lost the first one.
I lost one and drew one vs. standard Dragon, and did the same vs. MCTS Dragon. These were 10' + 5"; I think I might break even at 15' + 10". The results don't help choose between standard and MCTS Dragon; my subjective impression is that MCTS was more challenging for me.