I've blogged about this ages ago, but apparently I never made myself understood.
Imagine you have a few characters, probably lovable misfits, a tough one, a hacker, a disguise artist, etc. The GM is adjudicating something simple and in theme, say the hacker needs to bypass an electronic keypad and then the tough one needs to spring through the door and take down half a dozen guards.
Traditional resolution mechanics, used commonly in all of simulationist games, tactical games, and lightweight narrative-focused games, go something like:
* Decide how hard each of those are for a typical human
* Each character gets a bonus for how much better than a typical human they are
* Then you resolve it.
It's important that the players and GM all have a similar idea how difficult these things actually are for the players, or they'll get into an argument about the resolution. But in truth... most of them will have watched a LOT of movies about tough ones who take down rooms full of guards, and never ever seen it in real life. So when you get to the "estimate difficulty" part, it's easier to estimate "for the tough one, taking out six surprised and lightly armed guards is of moderate difficulty" than to estimate "for a typical human, is this challenging? extreme? superhuman? something else?"
I'm considering an alternative, something like:
* Look at the obstacle as described by the GM
* Look at the character's ability
* Adjudicate "OK, for your specific character, that's easy/medium/hard/nigh-impossible", and roll a die that says "you succeed on an easy/medium/hard/impossible" challenge.
If you have a simulationist system, the traditional method is almost necessary. It's also a lot more practical if you have lots of different small bonuses, because adding those to the player's achievement is easier than subtracting them from the difficulty. But outside those situations, in theory, that system has some advantages: the GM doesn't need to model the characters abilities, just how hard the situation is; it means players usually get big numbers or lots of dice which is fun. But I'm not sure I actually believe those.
In practice, in creating a fun experience, the GM probably has a better idea of "I want to provide the players with this much of a challenge" than of "I want the situation to be this challenging in the abstract". Especially if there's modifiers being thrown around, it's easy for a "choose a difficulty, and then the players get bonuses" model to end up with "whoops, the player can just always/never succeed at this".
For instance, the players try to bribe a guard. Everyone expects that to happen in heroic fantasy all the time, so the GM gives it a fairly low difficulty. Now the players want to disguise themselves as laundry attendants to escape the castle. The GM does the same thing. But it turns out there's a mechanic for bribing but not disguise, or vice versa, so the players get a whacking great bonus to one of them and not the other, despite both being what you'd expect from the genre. It means the GM and player's instinctive knowledge of what the characters can do can work against them if the mechanics don't perfectly line up.
But with the new system, appropriate difficulties happen automatically if people forget themselves, but you can still calculate them in detail when you feel the need. The GM can always just assume that as long as the hacker does the hacking and the tough one does the bruising, most challenges will be "medium", but they can throw an "easy" or "hard" in there if they want. And if they DO want to make things more objective they can use a rule-of-thumb of "for every notch above typical human you are, you reduce the difficulty by one level" without wiring it into the rules of the universe.
What are the advantages of that system?
One is, as I said, it's easier to adjudicate difficulty on the fly if everyone has a good idea what the characters can do but not what a normal human can do.
Also, if characters want to work outside their specialities it also works better. Maybe "jumping a gap", anyone can try even if only the athlete can be assumed to succeed, but "picking a lock" you can't do at all unless you know. Most systems force you to pick one or the other of those for all possible tasks (or choose two possible levels, as with DnD's "take 10/take 20" system and restrictions on some skills without training). In this system, the GM can adjudicate on the fly what obviously makes sense in the situation at hand, even if it means some tasks which are medium for the hacker are hard for other characters and some are impossible. Whereas with a traditional resolution, if two different players want to try the same thing, it's easy to have the results break everyone's expectation of what the characters can achieve.
And, it implicitly puts the variance under the GM's command, not only the mean. If one character has a special ability that makes routine something that is usually far out of the reach of other characters, the flavour might still suggest that they some of those tasks are easy and some are hard for them. In a traditional resolution mechanic, you *also* need to make sure those difficulties are out of reach of other characters, except for the times they actually should be able to do it with sufficient effort. With the new system, you can simply assign difficulties for the character with the special ability, and worry about the other characters only if they try something like that.
I'm not sure if there's actually any use for this system, but thinking it through helped me think how abilities and difficulties work.
And I'm still confused by the responses I got when I talked about this before, which were mostly, "If you think that, you should try FUDGE" which I mean, sure, a popular widely used system probably is a lot better than one person's random idea, but it seems so irrelevant, since FUDGE uses exactly the same traditional resolution order as DnD, so I wasn't sure what they were trying to say.