Quasispecies pointed me at this strategy - in a comment.
I don’t know if it’s significant - but there appears to be a temperature at which the centroid differs from the MFE - and the temperature required to have that occur seems to vary.
so perhaps a [Strategy Market] might be to work out what temperature is required to make the centroid differ.
I’m guessing those that require higher temperatures would generally have less variation in the ensemble, and that would probably be better.
Order them in temperature order when the centroid differs (I don’t know if how much it needs to differ is significant - not really understanding centroids, but if it is then I guess that could be a way to optimise this strategy)