We use a formula for our baseline testing, that is to get the units reasonably balanced and fix anything that is really out of whack. We fine tune the numbers based on playing a bazillion games, though. Our formulae include things like cost, range and unit speed, but they are still based on groups of single units vs. groups of single units. In a game where one Toxote backing up your Hoplites can make a difference, there is no way to represent that in any kind of simple formula. The only answer is to play a lot of games and see if it feels right.