Komodo 13.2

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Jonathan003
Posts: 52
Joined: Fri Jul 06, 2018 2:23 pm
Full name: Jonathan Cremers

Re: Komodo 13.2

Post by Jonathan003 » Sun Nov 17, 2019 5:10 pm

Ovyron wrote:
Sun Nov 17, 2019 1:41 am
Jonathan003 wrote:
Wed Nov 13, 2019 7:36 pm
There is the same behavior later in a game, also in the late middle game. Komodo with the human personality seems to be very optimistic for the side ho has to play.
This is what happens with Contempt. For Komodo's analysis, it is recommended that you create one instance or White, and one instance for Black, and that you don't let them analyze the wrong side of the board. Then you will be able to use new entities to suggest moves and lines.
I checked white contempt and set contempt to +16 for the instance for white and to -16 for the instance for black.

That seems to give better results. I wonder if the recommended lines will be very different, or if there will only be a different evaluation score with these changes in contempt?

Do you recommend against using the default setting for contempt, (white contempt unchecked), with the Human personalty? I don't really care if the engine analysis say +0.8 or +0.2, as long the recommended variations are useful for opening preparation, it's ok for me.

Maybe it will be problematic for the stored engine analysis in the cloud with chessbase 15, as it would be misleading for other users.

I don't use MCTS because my hardware is limited and the variations are not not long enough if I use MCTS. I use a kibitzer in Chessbase 15 and copy the analysis to the notation for opening preparation.

User avatar
Ovyron
Posts: 2815
Joined: Tue Jul 03, 2007 2:30 am

Re: Komodo 13.2

Post by Ovyron » Mon Nov 18, 2019 1:31 am

Jonathan003 wrote:
Sun Nov 17, 2019 5:10 pm
Do you recommend against using the default setting for contempt, (white contempt unchecked), with the Human personalty? I don't really care if the engine analysis say +0.8 or +0.2, as long the recommended variations are useful for opening preparation, it's ok for me.
You want to use the default engine with default settings as your main analysis engine, as it will provide the most useful results for you. You'd use the Human Personality and Contempt to explore different ideas to play from default (but you need to know what are the best ideas and to know what are the alternatives, to see if you like the alternatives better, but knowing what are truly Komodo's preferred choices without being tweaked.) Note Human will swindle (playing weak moves and hoping that the opponent doesn't punish them, trying them on a game could be dangerous.)

Contempt is specially useful in certain positions, like quiet positions (where there's nothing to do so nothing is best, this is the best case to explore what personalities would play), or positions where you want to draw (where negative Contempt would aim for drawish positions.)

Disclaimer: I don't have Komodo 12-13 :mrgreen: but this is how Contempt and Personalities work so they should work like this with Komodo.

Jonathan003
Posts: 52
Joined: Fri Jul 06, 2018 2:23 pm
Full name: Jonathan Cremers

Re: Komodo 13.2

Post by Jonathan003 » Tue Nov 19, 2019 4:45 pm

Thanks for the recommendations.

I wan't to make a repertoire for otb games. And I want to avoid the problem, that many times you will never get your prepared opening lines actually on the board.
That's why I base my repertoire on human games, and engines with the human personalty.

But I realise this is probably only wishful thinking to. :D

I see here a big gab in existing chess software. There is no efficient way to build a comprehensive and durable openings repertoire, that fit's you playing style. In a non-10 year approach, at least.

User avatar
Ovyron
Posts: 2815
Joined: Tue Jul 03, 2007 2:30 am

Re: Komodo 13.2

Post by Ovyron » Mon Nov 25, 2019 10:38 pm

Jonathan003 wrote:
Tue Nov 19, 2019 4:45 pm
Thanks for the recommendations.

I wan't to make a repertoire for otb games. And I want to avoid the problem, that many times you will never get your prepared opening lines actually on the board.
That's why I base my repertoire on human games, and engines with the human personalty.

But I realise this is probably only wishful thinking to. :D

I see here a big gab in existing chess software. There is no efficient way to build a comprehensive and durable openings repertoire, that fit's you playing style. In a non-10 year approach, at least.
You don't need an opening repertoire. What you need is a system, a set of places where you want your pieces to end at, and emergency places where they end at if the opponent doesn't allow the first one. I have successfully developed such a system, and now I can play the opening really quick without caring about what the opponent plays, and always reaching comfortable positions.

It's funny that chess engines had nothing to do with it, so this is something that can be done without them.

leavenfish
Posts: 282
Joined: Mon Sep 02, 2013 6:23 am

Re: Komodo 13.2

Post by leavenfish » Sun Dec 01, 2019 4:43 am

Ovyron wrote:
Mon Nov 25, 2019 10:38 pm
Jonathan003 wrote:
Tue Nov 19, 2019 4:45 pm
Thanks for the recommendations.

I wan't to make a repertoire for otb games. And I want to avoid the problem, that many times you will never get your prepared opening lines actually on the board.
That's why I base my repertoire on human games, and engines with the human personalty.

But I realise this is probably only wishful thinking to. :D

I see here a big gab in existing chess software. There is no efficient way to build a comprehensive and durable openings repertoire, that fit's you playing style. In a non-10 year approach, at least.
You don't need an opening repertoire. What you need is a system, a set of places where you want your pieces to end at, and emergency places where they end at if the opponent doesn't allow the first one. I have successfully developed such a system, and now I can play the opening really quick without caring about what the opponent plays, and always reaching comfortable positions.

It's funny that chess engines had nothing to do with it, so this is something that can be done without them.
Hans Berliner would have been proud, I am sure....

User avatar
Ovyron
Posts: 2815
Joined: Tue Jul 03, 2007 2:30 am

Re: Komodo 13.2

Post by Ovyron » Sun Dec 01, 2019 8:05 am

What I have found is best is building a new system of your own that creates patterns that are very rare, then very strong opponents will trip up (because you'll get used to the patterns and will play those positions much better than them). I have been getting the most won positions against stronger opponents in my life like this! It doesn't matter because I miss the key moves and they still beat me, but it's cool to see I had cracked their code and could beat them if I was more used to playing won positions against stronger opponents (which is very difficult to practice), but I guess I can get there now.

Post Reply