If you followed this thread, you might be interested by this independent post that proposes a formula that goes in the same direction than I did in this thread. It proposes a parametrizable formula, that for some parameters value ends up being equivalent to the one I propose, but is more general: