I ran 10 battles with each of those, with units spread evenly between 3 flanks. So that's 30 times a flank faced another flank and it should average out fairly well.
Not sure about that. One lucky flank can snowball on the other flanks and avoid another flank to be wiped out after an unluckily triggering a bad tactic. Plus, when collecting the results, you take the win percentage of a battle, not of each flanks individually (which would be bad anyway since a winning flank snowballs an adjacent flank). Thus, you have only 10 independant battles.
Thanks for the information ^^
notes:
(I don't contest your choice of dividing the troops in 3 flanks tho... I think this is the way to go since in real game, you will use 3 flanks whenever possible)
(I don't contest the fact that you do only 10 battles. I know how doing these tests can be boring, and I know this is a work in progress

)