My idea is basically that the melt plot must be flat in the first (and possibly first 2) squares (as that means the right number of bonds have formed), but that tends to christmas trees - so once that is fulfilled and the dotplot is roughly lined up lets count low energy design as too sticky to fold and thus bad.
[Strategy Market]
the melt plot must be flat in 1st square, and dotplot must have dark grey (or darker) in all target areas (otherwise the wrong shape is being made), or minus 100 from the design’s free energy value
Then order by modified energy value closest to zero = top
=Note=
I managed to break my plan by building a design in backward C with a flat melt plot for the first two squares, but with the dotplot not lining up with the target areas, and a poorish free energy result.