OpenChess

Blue Marlin - Stockfish derivative that provides sharp and attractive games. The engine is suitable for playing and analyzing, it supports Polyglot books.

Blue Marlin 15.5 is kind of special release because it is perhaps the last one. Blue Marlin 15.5 still uses the old SFNNv4 architecture with an old net from the 18th of April 2022 which is great for sacrifices and attacks. All problems regarding analysis I found in previous releases are fixed. Tactics is improved, strength is improved, evaluation is improved, the style remains unchanged.

Download:
https://www.mediafire.com/folder/zszqb3 ... arlin_15.5

One of the best attacking engines, Thank you D

Thanks for the Blue Marlin engine releases - they are always welcome for their more attacking and enjoyable playing style. I hope you will decide to keep them going

STUDENT_MASTER wrote: ↑

Fri Dec 30, 2022 2:03 am

One of the best attacking engines, Thank you D

Have fun!

Eric Hallsworth wrote: ↑

Fri Dec 30, 2022 12:56 pm

Thanks for the Blue Marlin engine releases - they are always welcome for their more attacking and enjoyable playing style. I hope you will decide to keep them going

I use engines often for analysis, sometimes a different perspective from Stockfish is needed. Engine tournaments are fun too when there are interesting games, but most people are only interested in Elo. Until yesterday, I thought the only person using Blue Marlin was me. If it's not what I thought... we'll see.

I think it's good that you're using an older NNUE. New coach Linmiao Xu's latest 3 nets are so boring and tactically weaker. I will take these new nets out of my engine again.

Yes, they are so strong and so boring. Hopefully new nets will appear soon.

Sorry but I may haf uncover a bug in Blue Marlin 15.5

U can download the pgn file here, https://pixeldrain.com/u/wsJHRftt

Blue Marlin lost a few times using the Caro-Kann Advance variation, Short variation, I suspect it is due to some evaluation issue.

Thanks for posting it. I don't see a bug here. BM lost because it didn't play the best moves in several cases. I checked it on older versions of the engine and it looks very similar. It is probably an imperfection of the network used. BM is tactically strong, positionally weaker, it can lose sometimes.

I said that I would rather not make a new BM, but I do, the tests look quite good.