Change SPRT bounds for the "only tune patches" LTC tests | Stockfish | Page 1

warm stone Jan 19, 2023, 12:05 PM

#

I suggest changing the LTC bounds for the patches that only change the values of parameters, without touching anything else, from <0.5, 2.5> to <0, 2>
As there is no complexity addition in these types of patches.
And if it proves to pass STC and LTC both of them at <0, 2> bounds, this should confirm that the patch is good enough to be merged.
This should also help in getting out local max/min

plucky saffron Jan 19, 2023, 6:18 PM

#

i agree with the general principle that we can be more lenient about LOS95% type patches if they are not subject to simplification, exactly like a paramater-tweak-only patch

warm stone Jan 20, 2023, 9:39 PM

#

Since I see positive feedback, how can I proceed to suggest to maintainers to make it official?
I never opened a PR for bounds change.

glacial kite Jan 23, 2023, 4:18 PM

#

warm stone Since I see positive feedback, how can I proceed to suggest to maintainers to ma...

U would need to open an issue in the fishtest repo and/or create a pr that adds a "Tune bounds"

warm stone Jan 23, 2023, 7:19 PM

#

glacial kite U would need to open an issue in the fishtest repo and/or create a pr that adds ...

Thank you david, I would prefer the second one, but I'm not sure in what file of the code there are the drafts of standard bounds, do you know?

glacial kite Jan 23, 2023, 7:23 PM