OpenChess

inheritance :

From


Then


To

renaming :

From


To

editing :


From start position, it takes 25 days 24/7 on 36 threads until D70, imagine until D100+ without a cluster loool

Oh! terrible, only good machines can achieve such depth. At 104 depth for d4 and e4 is the best move, but only +0.04 and +0.03 are also balanced

And when the hardware isn't big editing of experience data

When there isn't experience data trained by the own engine renaming another exp file

When there wasn't training merging several exp files

Etc.

At least since the November 5, 2022 patch, these people have either corrupted their exp file or started from scratch

hey deeds. If I remember correctly, you once wrote not to use multipv in Eman because he gets too much knowledge about one position and it makes him play worse? slower? I don't remember exactly but I think you said something like that. Has anything changed so far? Is multipv now advisable to enable this option?

Regards

Yes, I confirm that is still the case.

I guess you told about this feature :


With Experience MultiPV set to true (default), few of moves with shallow depth are written to the experience file :


And as you can see, these optional moves :
- don't help improve the quality of the most efficient move
- they just make the file heavier
- it takes longer to load it
- it takes longer to defrag it
- it unnecessarily slows down the engine

[for engine trainers] if we reinforce this position, the experience data become more consistent :

Dear Deed!
It is true that, choosing Experience Mulpv will appear bad moves. In my settings, I click on "Experience Book" and "Experience Book best Move" - that's good, right?

A bit of literature :
https://cumnor-chess-server.com/phpbb/v ... .php?t=558
https://www.open-chess.org/viewtopic.ph ... =20#p26652
https://www.open-chess.org/viewtopic.ph ... =10#p27325

DeeDs

merging :

From



To


Then