Rules light magic that doesn't get wacky?

[Monkeys]

I would treat it as "what are you trying to do by non-magical means", and "how many points of magic do you want to use to add to the relevant roll" - in this case what you want to do is "beat 100 men in combat", so the spell would be however you normally do combat rolls, with whatever bonus comes from the amount of magic you 'spend' on it. Depending on the system, perhaps "beat 100 men in combat" has several parts, eg "I want to have 100 attacks per turn", "I want them all to hit", "I want all their attacks to miss or do no damage".

For magic to be balanced with anything else, something like beating 100 men in combat would have to be impossible, unless non-magical characters can do similar things.

I see what you're saying Monkeys.  I like that.

Modifiers to all actions:
+2 You have an ideal tool for the task
+0 You have an appropriate tool for the task
-2 You have the wrong tool for the task

Magic is like always having the appropriate tools for the task.

What about this approach?

Player: "I want to hurt those 100 men.  The appropriate tool is a big bomb on the end of a short-range precision-guided rocket.  So that's easy." 
GM: "Actually, the appropriate materials are bomb-making materials and a big sling to toss it with.  So your spell is as difficult as making a bomb and aiming that sling in 10 seconds.  That's daunting."

Too arbitrary?

[Valamir]

Its clear there's a serious break down in communication here...I'm not sure sure where that is, so lets start over with an example.  I'm moderately familier with PDQ.  Its not my favorite system since its about one step away from being "resolution by GM fiat" but lets go with it.

You're heroes are being chased by a gang of mercenaries.  They've done a bunch of rolls to shoot arrows, set traps, try to cover their trail...whatever but now their pursuers are closing in and they find themselves at the brink of a big chasm.  Maybe you just invented that chasm, maybe it was already on the map* whatever.

A player sorcerer decides they're going to conjure a bridge to cross the chasm.  Great.  Make the Roll.  2d6 + Conjuration. 

Lets say you set up your rules for magic so that magic starts at Average plus a level for additional circumstances.  In this case you decide that the size of the bridge necessary is more than the standard size range for conjuration, and that things need to be done in a hurry so you bump things up to "Difficult"

The player rolls and fails by a good bit.

Why'd he fail?  Did the mercenaries arrive before the spell was completed.  Was the bridge too large and the sorcerer couldn't summon enough energy?  Was the bridge conjured in time, but collapsed before the heroes could cross?  The only thing the dice decide is that the goal of conjuring a crossable bridge over the chasm didn't happen.  That's all FiTM is.  Same logic if the roll had succeeded. 


*As another example of FiTM at work, lets say that earlier one of the PCs made a 2d6 + Orienteering (or whatever) roll to try and find a short cut to help escape from the mercenaries.  He failed it miserably.  As a result, you, the GM decides that failing such a roll by such a margin means the heroes are in a worse scrape than they were before.  That's when the chasm got invented.  Wasn't there before, but it was a result of interpreting the result of the previous failure.

[Christoffer Lernö]

Another variant:

A player wants to make a bridge across a chasm, and rolls 2d6 + Conjuration.

Result is 11 which is a success-level "Challening"

The GM says that this is enough to make a bridge across by doing a ritual for 1 minute.

The player can now negotiate perhaps making a weaker bridge in less time, a more grand-looking bridge by taking more time etc.

The player wants to do the spell in 5 seconds, and the GM says this will make the bridge really weak and dangerous to cross.

The player agrees with this, and narrates the effect of the magic.

Afterwards, the GM might require skill rolls for players crossing or whatever else the GM feels is appropriate.