Rofchade - no testing and rating

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Uri Blass
Posts: 8942
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Rofchade - no testing and rating

Post by Uri Blass » Sun Dec 09, 2018 6:33 am

lucasart wrote:
Sat Dec 08, 2018 3:15 am
Wolfgang wrote:
Sat Dec 08, 2018 1:50 am
and CEGT 40/4 (blitz) ... /1096.html
Impressive for an engine that has only Material + PST eval. Next version will come with hundreds of elo of eval, and steamroll the competition.
I will not be surprised if modified stockfish with PST eval can do better.
I believe the strong programs are strong usually mainly because of superior search.

I remember that in the past I tried to modify strelka's evaluation to piece square table evaluation and it did very well and beat Joker convincingly
and I even did not try to optimize the piece square table of strelka so basically it did not have correct value of pieces.

Joker has only 2301 ccrl rating list but piece square table not optimized strelka could have at least 2400 and with some optimization I am sure it could get more than 2500 elo.

stockfish is many hundrends of elo better than strelka so I expect modified stockfish to get at least 2900 elo by piece square table evaluation and tuning of evaluation and search parameters.

Edit:looking again in the CEGT it seems that the rating of Joker in CEGT is only slightly more than 2100 elo and CCRL is not equivalent to CEGT.

Still I expect more than 2700 elo from modified stockfish also with CEGT rating list.

I will also not be suprised if you get more than 2900 with piece square table CEGT rating list.

2900 is still not close to 3447 that stockfish 10 1 cpu has there.

Post Reply