Deeds on Chess Engines Testing

[Homayoun]

Ok. Thanks. I understood the point.

[Homayoun]

I understood that an experience file can help as a great assistant to modify lines and variants in chess openings. It can be a great complement for books too , because you can creat innovative moves with experience file. So a match between a book and experience file is meaningless. Because an experience file can be a superior complement for a book or a bookmaker too. Also we shouldn’t apply an engine with its experience file( on mode) in tournaments for measuring it’s strength , only the content and quality of an experience file is important. These are true?

[Martin_1969]

Creating an experience file takes a lot of computer time, a lot of energy.
You have to play the same line many times. I let Eman play 2000 games at (TC 3+1) against the same engine (C_Dragon v3.1) with a single line (rnbqkbnr/pppppppppp/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1). This is a UHO opening, which was also used on chess.com.The engines both used 2 cores each. Not only did it take a long time, as you can only play one game at a time, I'm sure Eman didn't get any real advantage from it even after 2000 games. Me, on the other hand, I created a bin-book from the 2000 games in only a few minutes, which performed better in a match with a TC of 3+1 against C-Dragon. For the book I used SF-FB from 09/28/24.
Creating a strong experience file is disproportionate between the benefit and the time needed.
And everyone can test this himself. I certainly won't do it again.
An independent tester can certainly confirm this.

[Martin_1969]

link to the games:
https://uploadnow.io/files/bHhTzT2

[Homayoun]

Is it possible for engine in learning mode to find a brilliant move that nobody else has never noticed it brfore( innovation)? If the answer is yes, I think it has the worth to spend time for training and making experience files than just making books from only existed moves. Although making books is also worthy but in recent years chess needs more to innovations.
Note: when I say nobody else I mean even other chess engines too.

[Homayoun]

Let me ask my question in other way:
Suppose an engine like clover ( not stockfish family). Can the version 6 of clover with experience file and also skilled trainer be even stronger than version 8 of clover which uses a different and new NNUE net( but without exp file).
Thanks in advance
Note: stronger here : means having more positional knowledge and better performance.

[Martin_1969]

[/quote]
An independent tester haven't the tool to use SF-FB from 09/28/24 + the EXP file in order to compare with SF-FB from 09/28/24 + the BIN file.
[/quote]

you can find all the tools here in this forum and creating a bin book is very easy.

[Martin_1969]

you can find all the tools here in this forum and creating a bin book is very easy.

[Martin_1969]

An independent tester haven't the tool to use SF-FB from 09/28/24 + the EXP file in order to compare with SF-FB from 09/28/24 + the BIN file.

you can find all the tools here in this forum and creating a bin book is very easy.

[deeds]

Is it possible for engine in learning mode to find a brilliant move that nobody else has never noticed it brfore( innovation) ?

Yes, it often happen at training, it is the goal too !

When non-learning engines plays against non-learning engines, nothing allows/force them to select an another move than their bestmoves. So they often produce the same lines, the same moves in the same key positions, etc. They are very predictable.

When learning engines plays against learning engines, they try the 3rd or 4th bestmove until they find a line that is more effective than the 1st or 2nd bestmove. They rarely produce the same lines, the same moves in the same key positions, etc. They are increasingly unpredictable.