I've run into this concept a few times over the past couple of months so I thought that I should make some kind of reference to the idea in my weekly mechanism blog.
The basic gist is that you choose a bunch of objectives, then roll a bunch of dice. Upon seeing the result of the dice, the player chooses to allocate them across their objectives, allowing the effects they really want to have a better chance of occurring, while allowing less important aspects can fail.
The example given uses 3 objectives for an event, and then rolls three six sided dice. For the three objectives you pick two or three degrees of outcome (Bad[1-3]/Good[4-6] or Bad[1-2]/Neutral[3-4]/Good[5-6]), and once you roll the dice you allocate them across the 3 objectives.
The degree of the good effect should be roughly proportional to the degree of the bad effect. Good, my opponent is humiliated; versus Bad, I am humiliated. Good, I win the race; Neutral, an unrelated competitor wins the race; Bad, my nemesis wins the race.
You basically pick a bunch of mutually exclusive outcomes that can be derived from the action.
I haven't quite defined the concept in the same way that the original author did,but I'm trying to show another way of looking at the notion.
The original concept has a primary objective and two relevant outcomes. The primary objective is divided into "No Success" [1-2], "Partial Success" [3-4], "Full Success" [5-6]; while the relevant outcomes are related to typically bad events "Does Not Happen" [1-3], "Happens" [4-6].
You roll the three dice, if one gets high, do you use it to complete the task and allow the other dice to apply a pair of penalties to you, or do you sacrifice completing the task in order to prevent the worst from happening to you?
If you roll two dice high, and the third low, do you use the good results to prevent anything bad from happening and accept the failure, or do you push for the success in exchange for a backlash from somewhere?
It's the kind of system that really gets a player thinking about their priorities with respect to their character and the game world.
I've actually played with similar concepts before, and the notion is even similar to the hit location mechanism I proposed a while back (where the lowest of two dice applies one effect (like hit location), while the other die determines something else (like base weapon damage).
One of the interesting comments noted about this mechanism is that it seems to favour the main event occuring, because most players will automatically tend to place their highest die in the event outcome slot, unless the event stakes are REALLY bad.
Task: Climbing Fence...relevant outcome 1: die from getting impaled on the fence...relevant outcome 2: a dog starts barking as you climb the fence, alerting security guards.
Suddenly you want at least one decent result, and that will immediately be used to prevent the death. The second good result (of it's there), will go on prevent the dog barking. Climbing the fence is a bit less important.
You need to make sure the negative events are suitably weighted against the activity being accomplished.
The mechanism can be easily expanded with a few simple additions. A simple skill system can be used to modify the die applied to the primary task (+1 to the task for basic skill expertise, +2 to the task for higher skill expertise).
The flipside is a modification to the die applied to the other outcomes, modifiers that might kick in under certain circumstances...a berserker has a modifer that makes it more likely they will lose their composure if this becomes a risk factor, a cursed mystic has a better chance of incurring the wrath of angry spirits.
Another way to expand the mechanism would be to prevent certain dice from being allocated under certain circumstances. A reckless character might be banned from placing their highest die in a secondary issue, therefore increasing the tendency of something bad going wrong, but also increasing the chance of success. An overly cautious character might have the exact opposite problem, they may not place their highest die in their primary task, but this tends to mean that they'll have a good die to avoid at least one of the secondary issues.
Still, the system has a simple elegance to it, and I wouldn't want to mess with it too much.