Komodo 9.2

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.
lkaufman
Posts: 3772
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Komodo 9.2

Post by lkaufman » Fri Aug 21, 2015 1:43 am

We have just released Komodo 9.2 at komodochess.com. As usual it is free to subscribers and available at a 20% discount to those who bought Komodo 8 or a later version. It's been about two months since our 9.1 release so we are well ahead of our quarterly target.
The estimated rating gain at fast levels is about 15 elo, about the same as the gain from 9.02 to 9.1, so the gain from 9.02 is about 30 elo and from Komodo 8 about 80 elo. This comes from re-tuning many parameters, from improved time management, from improved move ordering and from free speedups. However the most significant change introduced in Komodo 9.2 is "Contempt", which replaces "Drawscore". This is not just a name and sign change. Contempt now incorporates the concept of trying to keep more pieces on the board until you are winning, when playing a weaker opponent. The default value of 15 (centipawns) is probably optimum for playing engines rated around a hundred elo or so below Komodo 9.2. Against a close rival (of which there is currently only one other than Komodo 9.1) best results will be obtained with a smaller contempt value, perhaps 5. Against most other engines a larger value like 20 or 25 should work even better than the default. Against humans I suggest using 40 for Grandmasters, 50 for Masters, and 75 for strong amateurs. Even larger values are suitable for handicap games. Note that in game play Contempt will cause the score reported by Komodo to be optimistic; in analysis mode (infinite search) Contempt is automatically set to zero.
The release came too late to be used in the first stage of TCEC, but of course it will be available for the second stage. We anticipate using Komodo 9.2 in Sunday's two pawn handicap match with Grandmaster Petr.
Komodo rules!

TShackel
Posts: 313
Joined: Fri Apr 04, 2014 10:09 pm
Location: Neenah, WI, United States

Re: Komodo 9.2

Post by TShackel » Fri Aug 21, 2015 2:28 am

Hi Larry,

Thanks very much for this announcement. I'm very happy for this new form of contempt that I recommended all the top engines do (in the spirit of houdini and avoiding too many exchanges which typically lead into a drawn endgame). And I'm also happy you found increased elo with use of this contempt. It's possible that when you avoid exchanges that will make it easier to gain advantage later on in the game, whereas if you exchanged everything down early the chances of a good advantage would be nil.

Keep up the great work.

Sincerely,

Tim.

mbabigian
Posts: 106
Joined: Tue Oct 15, 2013 12:34 am
Location: US

Re: Komodo 9.2

Post by mbabigian » Fri Aug 21, 2015 3:57 am

Note that in game play Contempt will cause the score reported by Komodo to be optimistic; in analysis mode (infinite search) Contempt is automatically set to zero.
Is there any way to force contempt to apply during analysis mode?

Thanks,
Mike.

lkaufman
Posts: 3772
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo 9.2

Post by lkaufman » Fri Aug 21, 2015 4:06 am

mbabigian wrote:
Note that in game play Contempt will cause the score reported by Komodo to be optimistic; in analysis mode (infinite search) Contempt is automatically set to zero.
Is there any way to force contempt to apply during analysis mode?

Thanks,
Mike.
Not directly, but I suppose you could accomplish this by "playing" a game against it with a ridiculously long time control, tricking it into thinking it is not in infinite analysis mode.
Komodo rules!

mbabigian
Posts: 106
Joined: Tue Oct 15, 2013 12:34 am
Location: US

Re: Komodo 9.2

Post by mbabigian » Fri Aug 21, 2015 4:17 am

Ok. Can you add a checkbox to force it on during infinite analysis as an upcoming feature? Perhaps in 9.3?

On the same subject, allowing the change of contempt during analysis without risking a hash clear would be awesome too if it doesn't preserve it already.

Congrats to your team. Another nice set of improvements!
Mike

User avatar
Graham Banks
Posts: 33258
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: Komodo 9.2

Post by Graham Banks » Fri Aug 21, 2015 4:24 am

lkaufman wrote:We have just released Komodo 9.2 at komodochess.com........
Looking forward to testing it. :)
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz

JJJ
Posts: 1288
Joined: Sat Apr 19, 2014 11:47 am

Re: Komodo 9.2

Post by JJJ » Fri Aug 21, 2015 5:13 am

15 elo at fast time control.
Is this more at mid time or long time control ?

User avatar
Werner
Posts: 2429
Joined: Wed Mar 08, 2006 9:09 pm

Re: Komodo 9.2

Post by Werner » Fri Aug 21, 2015 5:52 am

Graham Banks wrote:
lkaufman wrote:We have just released Komodo 9.2 at komodochess.com........
Looking forward to testing it. :)
Me too :wink:
Werner

melajara
Posts: 213
Joined: Thu Dec 16, 2010 3:39 pm

Re: Komodo 9.2

Post by melajara » Fri Aug 21, 2015 8:11 am

Larry and Mark, are you aware that another Komodo, Komodo Activestate, has been updated to 9.2 too? Are you in touch with them for some coordination? LOL

Such a funny coincidence!

See http://komodoide.com/download/
Per ardua ad astra

shrapnel
Posts: 1248
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: Komodo 9.2

Post by shrapnel » Fri Aug 21, 2015 11:47 am

lkaufman wrote:Against a close rival (of which there is currently only one other than Komodo 9.1) best results will be obtained with a smaller contempt value, perhaps 5.
Hmm....there is a small, but perhaps significant difference between the figure stated here and that in the Official Readme of Komodo 9.2.
Which is more accurate ?
Kindly clarify.
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

Post Reply