TalkChess.com

10% less nps than SF 16.1. Solves 70/110 in HTC110. SF 16.1 got 80. Lost 60 + 0.6 match vs SF 16.1. But good start for further development? I already removed from disc .

Jouni wrote: ↑Tue Apr 02, 2024 7:41 pm 10% less nps than SF 16.1. Solves 70/110 in HTC110. SF 16.1 got 80. Lost 60 + 0.6 match vs SF 16.1. But good start for further development? I already removed from disc .

You are so funny, tested with HTC and already removed from disc

As Eduard already mentioned:
Top Chess Engines Testsuite 2024 v2
https://www.mediafire.com/file/cypaz2t0 ... 2.pgn/file
-Stockfish 16.1 (20%) 23/115
-Stockfish 02042024 (45%) 52/115

HTC is very very old and much much weaker.
It's like:
HTC for ~3000 elo engines
TCETv2 for ~4000 elo engines.

For a dumb small sample test suite, the 2021 revision of Hard Talkchess is... ok. Top Chess Engines Testsuite v2 is not much better in comparison.
Just combine all (actually correct) positions and solutions from test suites into a big one with thousands of positions instead of a single small one and you guys might actually get usable results

Of course, tactical ability is not the same as game play ability.
I don't think it's possible to project one measure from the other accurately.
I will say that I think tactical ability is more important in correspondence chess.
That is because the quiet opening positions have been studied to death and the game actually begins after both engines are fully developed.

Its tactical ability is poor

EDIT: though some positions it gets much faster than previously...