>>106851239
feel free to run huge test series to get your statistical evaluation, but it'll probably kind-of vary by model anyhow

simply use some of the samplers you think are good?