Komodo 14 released.

[pohl4711]

Oh my gosh. The CCC needs to use that Armageddon mode.

What other engines have it?

None, as far as I know. But with my new released NBSC Advanced Armageddon openings and scoring system, all engines can play Armageddon (without knowing, that they are doing so). And the statistical quality of the results is just overwhelming.

Learn more on my website:
https://www.sp-cc.de/nbsc-armageddon-openings.htm

Or download the complete NBSC-Armageddon package (opening-files (pgn and epd), books (for FritzGUI, Arena and polyglot) and Advanced-Armageddon rescoring tools) just right here:
https://www.sp-cc.de/files/nbsc_advanced_armageddon.zip

Idea of NBSC Armageddon by Larry Kaufman (Komodo-team).
Advanced scoring idea and all work (and testings) done by me (Stefan Pohl, SPCC).

Hello! Would it be possible for you to port it to OBK format (Chessmaster)? Would love to try it there!

-JN

I use this Book converter tool to make the polyglot-books out of the Arena-book. It should also convert the Arena-book of my NBSC-Armageddon lines, which is in my NBSC-dowload-package (NBSC_30k_5mvs.abk), in the OBK-format.

https://goneill.co.nz/chess.php

(scroll down a little bit to:
"Opening book conversion (OBK, ABK, BIN)
The opening book editor in Chessmaster is very good but it only saves files in its own OBK format. These programs allow you to convert opening book files between Chessmaster OBK, Arena ABK and Polyglot BIN formats for use in other programs. During the conversion you can also output PGN and TXT versions of the files.")

[Cornfed]

This is to be expected. Take a position where the best move is to accept a 3-fold. Regular K will accept it and take the draw. Komodo Armageddon will score the 3-fold as extremely bad, and rather go for a worse position that it has a much higher likelihood of losing than winning. This is optimal from an Armageddon scoring perspective, but extremely bad from normal scoring.

Taking the regular score of an armageddon match, or the armageddon score of a regular match, produce misleading data.

This observation also means that, with engine that are made aware of Armageddon, awarding more points for black wins completely breaks the point of Armageddon, as white has then no incentive to play riskier than in regular chess.

That...actually kind of makes sense.

Armageddon strikes me as a bit of a gimmick in general to force a result in a single game. Chess at the higher levels is closer to a draw. Armageddon goes too far.

We are talking about blitz time controls at this point anyway, so simply let the normal 1,0,1/2 carry the day while letting time be the decisive factor: Give Black a slight time edge and progressively lower the time for each side after the first game (if it is not decisive) until a result occurs. Less time = more likelihood of errors and thus a quick decision.

That's normal Human chess where people have various strenghts/weaknesses and can pick the opening/defense according to their opponent or their situation.

Computer chess though, perhaps you need a gimmick if your engine of choice already calculates slower and thus 'plays worse' than its opponent the faster the time control. You would compound its weakness so you would naturally prefer and even promote something like armageddon.

[mbabigian]

Larry/Mark, is Armageddon active in infinite analysis? Can it be?

Thanks in advance,
Mike

[lkaufman]

It is active in infinite mode. Whether the Contempt adjustment is applied properly when you switch sides in infinite mode has to be checked out, it might even be GUI-dependent. You can always adjust Contempt yourself if needed.

[cc2150dx]

Larry/Mark, is Armageddon active in infinite analysis? Can it be?

Thanks in advance,
Mike

In addition to what Larry said. Some more information here: http://talkchess.com/forum3/viewtopic.p ... 93#p843793

[Jhoravi]

Is there a mode in Komodo to handle piece Odds better?

[lkaufman]

Is there a mode in Komodo to handle piece Odds better?

There's no mode for that, but setting Contempt to about half of the absolute value of the initial evaluation of the handicap position helps significantly. Whether MCTS mode helps or not in handicap play is not quite clear, today we'll learn more about that as we did not use MCTS vs Lenderman yesterday but will today.

[Lion]

@ Larry

Just purchased K14....
I was wondering if the MC version running around 4 kn/s is correct or is it because I am using my CPU (AMD Threadripper 1950x) instead of my GPU (2080Ti) ?

Thanks for the engines!

rgds

[lkaufman]

@ Larry

Just purchased K14....
I was wondering if the MC version running around 4 kn/s is correct or is it because I am using my CPU (AMD Threadripper 1950x) instead of my GPU (2080Ti) ?

Thanks for the engines!

rgds

Komodo (at least publicly available Komodo) does not yet use GPU, so yes, this is correct. Each MCTS node is based on hundreds or thousands of regular nodes, so that 4 ks/s really equates to millions of normal nodes per second.

[Lion]

Thanks for your response!

Reading "in between the lines", should we understand there may be at some point a GPU MC version? Should we exect an increase of speed with such a configuration/setup compared to CPU MC?

rgds