It's a poor design choice if that is based on pure random principle....
If you don't loose 1 out of every 5 '80% chance to win' battles, then the initial estimate is simply wrong, regardless of whether the battle results are purely random or based on a simulation.
