On you testing no dev build has been as highly rated as the 20220207-1630 dev build.
For every new build this one goes a step further towards the end of the selection list.
What if it goes out of the list with no other dev build being stronger? The strongest dev build from your testing should always be on that list.
I think you should always use the latest dev build reason being it has the latest patches even though it lost elo since that feb. build, Maybe someone can better explain it than I do.
Always use the new dev buld in my opinion, ncm test 30 second bullet test against stockfish 7, how reliable is it? check Stockfish Development Versions
This is because every test is against the dev bulid behind it.
Q: When we use NCM, we must use dev build which ncm showing high elo ?
Only NCM person mr CHENDRY can give a correct answer to this ,.waiting his replay.
I think @puttutathy is right in that the dip in the graph probably has a lot to do with NCM using short time controls for the tests. I’m guessing there changes that strengthened Stockfish for normal time controls, but weakened it for extremely short time controls.
I would expect that on the seconds-per-move scale, the latest dev builds are going to outperform 20220207-1630.
To put that to the test, I’ve scheduled 1000 games of 20220207-1630 vs 20220602-0543 at 1s/move:
I’ll make sure that 20220207-1630 doesn’t drop off for now, but as a longer term solution I’ll add links to each dev build on https://nextchessmove.com/dev-builds so that you can use any dev build no matter how old it is.
I got answer ,I contacted sf dev builder MR . condirfish, and Disservin through Discord
using the latest commit should give u the strongest version