==========
Disclaimer
==========
The architecture of a chess engine is a choice of its programmer(s). If chess engines can (or can't) use:
-endgame databases (Syzygy, Gaviota, Nalimov, various bitbases...);
-evaluation based on 1-n data accumulators (NNUEs, NNs, DCNNs) or HCE;
-experience files;
-internal or external books;
-etc, etc, etc......
does not imply unfairness in matches/tournaments. Each chess engine programmer achieves what he wants, given the skills he has.
====================================================================================================================
Stockfish_23030515_x64_bmi2
==========================
Match conditions:
-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: SuperGM_4mvs.abk (8 plies book) (only for Stockfish_23030515_x64_bmi2)
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-TBS: 6-men Syzygy bases for both engines
-NNs: defaults (Stockfish_23030515_x64_bmi2 uses nn-52471d67216a.nnue net; SpectralByLOBO-bmi2 uses an old SF net: nn-3c0054ea9860.nnue)
-SpectralByLOBO-bmi2 uses a special experience file
-Ponder: off
-OS: Windows 10 Home.
The final result:
+0 -9 =91
-28 Elo points in 100 games (+35 for SpectralByLOBO-bmi2)
A new sample:
======================================================================
Well, a few months off (until some progress is made) would not hurt...I will play with some HCE engines.... some private ones.....there are also some beautiful festivals around here........
======================================================================