TalkChess.com

LittleBlitzer accepts MultiPV=x command without problems! My short test with SF12 has ELO decrease as follows:

MultiPV=2 -73
MultiPV=3 -119
MultiPV=4 -191

Hard position 2008 :D . Crystal took whole 2 seconds: Also quad. Analysis by Crystal 110920: ... 1.Bxe6 Nxe6 2.Rd6 Rae8 3.Bc3 Re7 4.Red1 Qc8 5.Qg4 Nf4 6.Qxc8 Rxc8 7.Be1 Ne6 8.Rd7 Kg8 9.Rxe7 Nxe7 10.Rd7 Rc7 11.Rxc7 = (0.30) Depth: 16/28 00:00:00 2141kN 1.Qxf4 exf4 2.Rxe6 Ne5 3.Rxe5 fxe5 4.Rd7 Qb8 5.N...

I mean AUTOMATICALLY. So You can play matches and run testsuites in MV mode. Not possible?

I mean automatically and always. I haven't found any so far. If I insert MultiPV= 3 to *****.uci file no effect. The GUI is not using that parameter at all?

SF 3 best moves reveals fast the most obvious errors. Your example: Analysis by Stockfish 051020: 1. +- (1.95): 1.c3 Re8 2.Bxe7 Qxe7 3.h5 a5 4.Nxc5 Nxc5 5.dxc5 Qxc5 6.h6 Ba6 7.Re1 g6 8.Qa4 Rab8 9.b3 Rbc8 10.Re3 Qb5 11.Qd4 Qc5 12.Qd2 Qf8 13.Rh3 Kf7 14.Re1 Re7 15.Rd1 a4 16.Rh1 Qe8 17.Nh3 Rec7 18.Ng5+ ...

Tactics - interesting always! But I have looked short time ago for CSS-WM test with current engines = SF. I think 10-20 of positions are buggy. Rest are solved instantly in most cases. From your 100 positions I found already 12,21,23,35,46 dubious.

I only have 4 cores, so I repeat match giving SF11 also 20% time advantage. Result: exactly same! With these conditions You need 6-8 cores to draw SF12.

Now NNUE is +80 to classical with 172 cores. Big difference to CCC tournament were they have almost same score with 90 cores. The truth is somewere between I quess .

I don't remember so far direct test. But here's one 1m + 1s level and shortlines openings. 1 th vs 4 th. That is real progress .

Now TCEC has scaling test with 172 cores. Currently +90 to classic.