Page 1 of 5
New uci engine: Rofchade
Posted: Mon Aug 27, 2018 3:11 pm
by Ronald
After the good results of Rofchade in the monthly HGM tournament yesterday it's time to release the first official version of Rofchade, "Rofchade 1.0".
The engine is written from scratch in C(++) and has an estimated elo rating between 2775 and 2800 on the CCRL 40/4 scale single threaded.
Rofchade 1.0 only uses Material Value and Piece Square Tables for it's evaluation and is multi threaded. On my threadripper 1950X the estimated elo rating is around 3050 with 30 threads (hyperthreading on though).
During development of the search routine it became a challenge how far Rofchade could reach with only Material an PST. The last period I couldn't find much improvement anymore, so now it's time to start working on the evaluation routine in version 2.0!
Rofchade executables can be downloaded from it's own website:
http://rofchade.nl. After restructuring and cleaning up I will make the sources available for download.
Thanks to everybody who has helped directly and indirectly with establishing a long wished goal!
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 3:22 pm
by flok
I would like a "I like this" emoticon in this forum.
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 5:18 pm
by Dariusz
Hi Ronald
Very promising engine, Thank you for mac version
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 5:19 pm
by Guenther
Ronald wrote: ↑Mon Aug 27, 2018 3:11 pm
After the good results of Rofchade in the monthly HGM tournament yesterday it's time to release the first official version of Rofchade, "Rofchade 1.0".
The engine is written from scratch in C(++) and has an estimated elo rating between 2775 and 2800 on the CCRL 40/4 scale single threaded.
Rofchade 1.0 only uses Material Value and Piece Square Tables for it's evaluation and is multi threaded. On my threadripper 1950X the estimated elo rating is around 3050 with 30 threads (hyperthreading on though).
During development of the search routine it became a challenge how far Rofchade could reach with only Material an PST. The last period I couldn't find much improvement anymore, so now it's time to start working on the evaluation routine in version 2.0!
Rofchade executables can be downloaded from it's own website:
http://rofchade.nl. After restructuring and cleaning up I will make the sources available for download.
Thanks to everybody who has helped directly and indirectly with establishing a long wished goal!
Congratulations for the first release and also for the no popcount compilation!
I noticed that it calls itself '1.0 b104' and currently have it listed with that version number in the XB/UCI chronology.
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 5:30 pm
by Gabor Szots
Excellent news! Can't wait to test it (and I will not).
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 7:23 pm
by jorose
Congratulations on the release!
Ill be honest, as someone who is struggling to write such a strong engine without the constraint you gave yourself it is honestly a bit depressing seeing people pull off what you have! Nice work!
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 9:40 pm
by CMCanavessi
Awesome! I'll include it in next season's CCLS
Re: New uci engine: Rofchade
Posted: Mon Aug 27, 2018 10:59 pm
by Graham Banks
Thanks Ronald. Looking forward to testing Rofchade.
Re: New uci engine: Rofchade
Posted: Tue Aug 28, 2018 11:41 am
by Daniel Anulliero
Hi
Amazing (and disapointing for me
) to see such à very strong enfine with "no eval" (just material and psqt !)
My weak engine Isa have a lot of knowledges in her eval and it is just 2350 ccrl lol
So I can say "there is something wrong in my search !"
(or in my eval too ?
)
Anyway , congratulations for your great work on RofChade !
Dany
Re: New uci engine: Rofchade
Posted: Tue Aug 28, 2018 11:51 am
by zenpawn
This is fascinating, reaching such a high rating with just material and PST's. Your search must be awesome!
Could you say a little about your piece square tables? For example, are they symmetrical or were they "Texel tuned" and left alone (which often results in higher weights on the kingside since that's usually where the enemy king winds up)? Thanks for any insight on this accomplishment.
) to see such à very strong enfine with "no eval" (just material and psqt !) 
(or in my eval too ?