Questions about optimize_gh_cut

[zhipzhang]

As mentioned in the doc, if we use hadroness-like score, we should use lesser operator here.
However, why the fill_value of this situation is the minimum of score?
If we don't have enough statistics in this situation, I think we should use a very loose cut the maximum of the score(the larger the more like proton).