Interesting results, Roy.
My tests assumed that it made no difference what size and composition the enemy army was, and so I had nothing but the general unit and one other of exactly the same type. This definitely could have skewed my results.
I'll give it another go with non-general unit types in the enemy army and see what sorts of results I get.
edit: also, I'm using custom battles rather than campaign battles, so that's another factor to consider. I guess I could savegame and play out the same battle over and over to control test conditions a bit.
Bookmarks