TalkChess.com

There is 4 winning moves at least. Analysis by Stockfish 110121: 1. +- (5.27): 1.Qxd2 Ba6 2.Ng4 Rac8 3.Re1 Rfe8 4.Qg5 Kf8 5.Ne5 Rc7 6.Ng6+ Kg8 7.Nf4 Bd3 8.Nxd3 Qd8 9.Qg3 Qxf6 10.Qxc7 cxd4 11.Rf1 Qe7 12.Qxe7 Rxe7 13.cxd4 Rc7 14.Rc1 Rc4 15.Rxc4 dxc4 16.Nb4 b5 17.Nc6 a6 18.Kf1 Kh7 19.Ne5 f6 20.Nd7 Kg8 ...

Yes You read right! Just for fun and curiosity I ran Stockfish dev version with NNUE against classic version with 3 GB of syzygy tables. Result: Games Completed = 1000 of 1000 (Avg game length = 80.946 sec) Settings = Gauntlet/64MB/30000ms+300ms/M 450cp for 5 moves, D 100 moves/EPD:endgames.epd(3000...

The goal: every position has exactly one winning/drawing move to find! How many are then left? 50?

Just one wish: Name engine Crystal 3.1 not Crystal 140121!

NCM has peak rating 8.1. with "Update copyright years". Sadly not possible before 2022 again .

With "clean search" I got mostly better results in testsuites with Fritz. Arena does always engine re-start in every position so not needed? But I noticed really weird thing: with one core and "clean search" Crystal crashed always in Fritz! Sooner or later engine speed goes to 1-5 knps.

One more.I have read, that Lc0 needs a lot memory to run well. But how do You increase memory? Even if I increase hash to XX GB and NN cache to zillion system monitor says always about 2 GB GPU memory used and 15% of GPU power. System monitor sucks I quess?

My GTX card never reached 7 plies to output something in Fritz GUI. I use one big net.

"Clean Search" makes engine to crash in Fritz GUI.

Fully agree Dann. There should be warning in Crystal page: NOT FOR PLAYING!