Part of analyzing this data that has not been talked about is that fact that, for most characters, three of the stats won't matter most of the time. Since the defenses are the greater of two mods, then sorting the columns in decending order by stat1, stat2, and stat3 could give a better idea of playability than just total stat mods. Sure, things like stat raises, feat thresholds, and total stat mods will come into play, as well as weighted stat mods over levels, but the fundamental mechanic is the greater of two stat mods for defenses, and the primary/secondary stat mods for attacks. If you have primary and secondary stats that do not overlap, then you can get the best of both worlds. You get something like this.
17 14 14 10 10 8
16 16 13 10 10 9
16 16 13 11 10 8
16 16 12 12 10 8
16 16 12 10 10 10
16 16 12 11 10 9
16 16 12 11 11 8
16 15 14 10 10 9
16 15 14 11 10 8
16 14 14 12 10 9
16 14 14 12 11 8
16 14 14 13 10 8
16 14 14 11 10 10
16 14 14 11 11 9
Two have no penalties, most have 13s or higher in the top three, and one has a 17, which if you think about going to 30th means that counting the odd number of possible stat raises means that you get the highest possible stat mod at 28th lv, and take biggest advantage of the starting points. Notice no 18s. Interesting.
This dataset also includes only the 12.5 to 11 rated weighted averages from the original dataset that I got off this thread. It doesn't even contain all the 11s.
Condensing this data down even further, it seems you could easily get just a couple of choices.
1) Do you want to overweight your primary stat?
2) Do you want to overweight one of your secondary stats?
3) Is one secondary stat more important than the other?
4) Are you OK with a penalty in a stat?