New engine releases & news 2021

[tmokonen]

DoctorB

V1.0.0

Thanks to the Author Hiblom slt

https://github.com/hiblom/DoctorB

The author is Hans Blom, whose previous engine was Haddock.

[Gabor Szots]

I tagged version 3.0 of Zahak, this comes with ~200 elo boost, thanks to the Texel optimization:
I suspect Zahak 3.0 to be around 2350-2380

Nice progress, Amanj. I'm looking forward to testing it. Only, I have problem with opponent selection: +200 gives about 2200 Elos which is way lower than the range you claim. What do I miss?

[amanjpro]

I tagged version 3.0 of Zahak, this comes with ~200 elo boost, thanks to the Texel optimization:
I suspect Zahak 3.0 to be around 2350-2380

Nice progress, Amanj. I'm looking forward to testing it. Only, I have problem with opponent selection: +200 gives about 2200 Elos which is way lower than the range you claim. What do I miss?

+200 over Zahak 2.0.0 (which is still not completely tested I believe), not Zahak 1.0.0 (which is around 2010). I believe the latest version (3.0) is slightly weaker or like laser v0.1

[Damir]

Fire_8.N
Thanks to the Author FireFather Norman Schmidt

NNUE support
uci option enginemode: choose classic, nnue(default), or random

https://github.com/FireFather/fire/rele ... g/Fire_8.N

Releasd with couple of other binaries as well....

Thx Damir

A couple notes:
To run this effectively, a processor that includes AVX2 (like BMI2) is almost mandatory, otherwise you could see an NPS decrease of up to 50%.
If you're not sure what instruction sets your CPU includes, I've included a complete of bmi2 and avx2 processors on the main page...see
https://github.com/FireFather/fire

ATM the nnue network must be named nn.bin and reside in the same directory as the Fire 8.N executable.

The NNUN implementation is completely different than SF, and utilizes a modified version of Daniel Shaw's/Cfish excellent nnue probe code:
https://github.com/dshawul/nnue-probe
Many thanks to him and Ronald de Man for that!

The download includes the top reinforcement learning NN (according to my extensive testing) trained by Sergio Viera last year.
I think you'll see that this, combined with Fire's search translates into some real interesting, unique, and strong engine chess.

Dear Norm,

Is it possible to improve the speed of Fire. I am testing it now on Playchess and is very slow, compared to the other engines I used...
Also is it possible to embeed the latest Stockfish nnue ( 47 mb) compared to the 20 mb one ?

[kranium]

Dear Norm,

Is it possible to improve the speed of Fire. I am testing it now on Playchess and is very slow, compared to the other engines I used...
Also is it possible to embeed the latest Stockfish nnue ( 47 mb) compared to the 20 mb one ?

Hi Damir,

I'm not very interested in embedding SF's nn, or using a larger size.
In the coming months, I'd like to work on creating a unique RL network.

I'd also like to work further on my attempts to implement MCTS...
I find that search method fascinating and the moves chosen via it unique and very interesting.
Everything I've tried so far is quite basic and very weak, and blunders material.
A decent MCTS looks to be a real challenge.
Kudos to the Komodo team for such a strong one.

But I'll do what I can concerning speed in the coming days/weeks.

[Damir]

Thanks Norm. Looking forward to new Fire with speed improvement.

[tmokonen]

Cinnamon 2.4 released:

https://github.com/gekomad/Cinnamon/releases/tag/v2.4

And somehow I missed Counter 3.8 from a few days ago:

https://github.com/ChizhovVadim/Counter ... ag/v1.38.0

[Rebel]

And somehow I missed Counter 3.8 from a few days ago:

https://github.com/ChizhovVadim/Counter ... ag/v1.38.0

Have Counter 3.8 running here.

[tmokonen]

Have Counter 3.8 running here.

Everyone has a 3000 rated engine except for me....

[mar]

Everyone has a 3000 rated engine except for me....

what's that supposed to mean, Tony?
I understand your frustration, but...

Cheng didn't become 3k overnight, no NNUE either. still waiting for my magical 100 elo bugfix, it never came somehow.
my first chess program dates back to 2006. There are some who've been working for decades (Bob?) and still haven't got their well deserved 3k on single core.

kids write 100 elo stronger programs than that these days, so I must be very stupid that it takes so much time to improve.
people want real improvements (100, 200, 300, 400, 500, 600 elo) quickly, somehow I'm unable to deliver that
what to do? shrug and move on

want elo boost? copy-paste NNUE probing code, grab strongest SF net and pat yourself on the back.
2k can become nealy 3k overnight, as has been demostrated by some.

PS still waiting for a CCRL rating! (and I'm not someone who releases each week or two)