[WraithLord]

It's difficult to explain in a single sentence. Since this conclusion is based on many other concepts, facts and underlying strategic principals. The essence of it is basically an attempt to inspect every possible action/commander/spell-cast as an investment of a resource, be it gems/gold or more subtle (like time) that yields X gain over the next turn, 2 turns, 10 turns etc. So you buy the assassin, equip it and then have it lie in ambush or move towards ambush but most of the time it's idle. This is already bad for ROI. Then there are so many counters to assassinations once your enemy is alert and skillful that you're most likely to lose it.

I'm not saying it's bad. Sometimes you'll score a significant hit and/or make your opponent paranoid (so he will "waste" resources on counters) but I am saying that you could most likely put your money on a better horse. This is my point of view - based on experience and my own judgment, not on pure number crunching.