Komodo 11 settings

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

cma6
Posts: 219
Joined: Thu May 29, 2014 5:58 pm

Komodo 11 settings

Post by cma6 »

For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
mjlef
Posts: 1494
Joined: Thu Mar 30, 2006 2:08 pm

Re: Komodo 11 settings

Post by mjlef »

cma6 wrote:For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
I do not really have any convincing evidence that anything other than the default settings are best for Komodo 11 at longer time controls. If someone can show an improvement at longer time controls with new values, please let us know.

Since parts of the dynamic evaluation, parts of king safety and parts of selectivity all changed from Komodo 10.4 to 11.01, then all three values might still matter.
cma6
Posts: 219
Joined: Thu May 29, 2014 5:58 pm

Re: Komodo 11 settings

Post by cma6 »

mjlef wrote:
cma6 wrote:For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
I do not really have any convincing evidence that anything other than the default settings are best for Komodo 11 at longer time controls. If someone can show an improvement at longer time controls with new values, please let us know.

Since parts of the dynamic evaluation, parts of king safety and parts of selectivity all changed from Komodo 10.4 to 11.01, then all three values might still matter.
OK, then I should go back to the default settings in K11 for these 3 parameters. Can you give me the defaults for these 3?
leavenfish
Posts: 282
Joined: Mon Sep 02, 2013 8:23 am

Re: Komodo 11 settings

Post by leavenfish »

cma6 wrote:For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
And for 11.1 for deep analysis of position - these settings would be 'optimal'?
cma6
Posts: 219
Joined: Thu May 29, 2014 5:58 pm

Re: Komodo 11 settings

Post by cma6 »

Hi Brian:
Were you answering "Yes" or ?
leavenfish
Posts: 282
Joined: Mon Sep 02, 2013 8:23 am

Re: Komodo 11 settings

Post by leavenfish »

I was asking a question of Mark.
lkaufman
Posts: 5960
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Komodo 11 settings

Post by lkaufman »

cma6 wrote:
mjlef wrote:
cma6 wrote:For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
I do not really have any convincing evidence that anything other than the default settings are best for Komodo 11 at longer time controls. If someone can show an improvement at longer time controls with new values, please let us know.

Since parts of the dynamic evaluation, parts of king safety and parts of selectivity all changed from Komodo 10.4 to 11.01, then all three values might still matter.
OK, then I should go back to the default settings in K11 for these 3 parameters. Can you give me the defaults for these 3?
The defaults are still 73 for King Safety and also (by coincidence) for Selectivity, and 120 for Dynamism. The numbers you mention are pretty far from these, and we would never have suggested them; do you recall where you got these recommended numbers? There is some slight evidence that longer time controls justify higher values for all three parameters, but we don't have enough games at long time controls to warrant any particular alternate recommendation to the default.
Komodo rules!
cma6
Posts: 219
Joined: Thu May 29, 2014 5:58 pm

Re: Komodo 11 settings

Post by cma6 »

lkaufman wrote:
cma6 wrote:
mjlef wrote:
cma6 wrote:For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
I do not really have any convincing evidence that anything other than the default settings are best for Komodo 11 at longer time controls. If someone can show an improvement at longer time controls with new values, please let us know.

Since parts of the dynamic evaluation, parts of king safety and parts of selectivity all changed from Komodo 10.4 to 11.01, then all three values might still matter.
OK, then I should go back to the default settings in K11 for these 3 parameters. Can you give me the defaults for these 3?
The defaults are still 73 for King Safety and also (by coincidence) for Selectivity, and 120 for Dynamism. The numbers you mention are pretty far from these, and we would never have suggested them; do you recall where you got these recommended numbers? There is some slight evidence that longer time controls justify higher values for all three parameters, but we don't have enough games at long time controls to warrant any particular alternate recommendation to the default.
My parameters are for correspondence deep analysis, e.g., 15-30 minutes/move. 131 for Selectivity must have been used in an earlier version of Komodo 10.
Dynamism of 93 was my compromise between recommended 110 and my inclination for a low value. My understanding, at least for v. 10, was that the higher Dynamism was above 100, the greater the loss in playing strength.
Aren't your defaults aimed for fast time controls?
mjlef
Posts: 1494
Joined: Thu Mar 30, 2006 2:08 pm

Re: Komodo 11 settings

Post by mjlef »

leavenfish wrote:
cma6 wrote:For Komodo 10,4 at long time controls, the recommended settings were
King Safety 105
Dynamism 93
Selectivity 131

Any changes here for v.11?
And for 11.1 for deep analysis of position - these settings would be 'optimal'?
We do not know. Long time control games just take too long to get the many thousands of games needed to confirm these parameter changes. Now if someone has a cluster free for a few months....
cma6
Posts: 219
Joined: Thu May 29, 2014 5:58 pm

Re: Komodo 11 settings

Post by cma6 »

The indications for the best settings for very long time controls would likely come from high level correspondence players who could, if so inclined, present anecdotal evidence of what worked best for them.