In this article, he describes the formulaic Voltron episode: the heroes fight the evil robot individually and get whupped; the heroes form Voltron and fight, but get whupped; the heroes then pull out the Sword of Arus and save the day. He goes on to say this:
From a gaming standpoint, though, there's just one slight, itsy-bitsy, teeny-weeny problem with it:
It would never work.
No, at the first sign of danger, any group of PCs would immediately form Voltron and have that sword pulled out. No questions asked. An inhuman shriek from above? Form Voltron, draw the sword, enter battle. The sound of an explosion from over the hill? Form Voltron, draw the sword, go investigate. Telephone rings? Form Voltron, draw the sword, answer the phone.
And so it is with this planning; the PCs are trying to make a perfect plan, using all their special abilities. (It's pretty clear that even if they win, after the battle is over they'll have expended so many of their spells and special abilities that they'd be hard-pressed to fight a housecat before resting.)
They hope they'll win. But if they fail, it's not easy for them to improve and try again; they've eliminated some room to do that.
On another note, I've become more willing to consider the players' plans in my counter-planning. My logic is this: the vampire has centuries of experience and very good intelligence and tactics. So in general, he's considered all plans that might be used against him, and prepared counter-strategies. But I don't have nearly as much experience or cunning--so I compensate by limiting my attention to the plans that actually might get used by these players.
The same principle is in effect when I prepare only the NPCs and locations that people might encounter.