Echelon, being derived primarily from the Revised System Reference Document, uses a d20-based task resolution system.
Pretty much all character-initiated actions that can pass or fail have the decision made with a single die roll:
d20 + modifiers >= Difficulty Class (DC)
This is about as simple as it gets, really. A consistent mechanic, easily remembered, and I aim to keep the number of modifiers somewhat lower than typically seen in D&D 3.x or Pathfinder. I have noticed that players often try to acquire multiple small bonuses because the aggregate cost (in whatever resource is being measured, be it gold, experience points, or spell slots) is generally lower than paying for one big bonus.
I can’t be bothered, so a small number of larger bonuses, for much the same net effect, will satisfy the purely numeric part.
Degrees of Success and Failure
Dungeon & Dragons 3.x and Pathfinder appear to run quite handily on a pass/fail mechanic, almost entirely. There are a few places that consider degrees of success and failure, but they are pretty uncommon. Critical hits and Climb checks are about the only ones that come immediately to mind. A percentage of successful hits can do extra damage, and a Climb check that fails by five or more points means the climber falls.
While most checks are pass/fail, some allow you to try for better results by changing the check DC, or taking a penalty to the check, which is much the same difference most of the time. This is not the same as degrees of success because you will either pass or fail, just with lower chance of success but better results.
It seems degrees of success or failure depend on how much higher or lower you rolled than the DC. If the DC is 15 and your check totals 20, you have succeeded by 5. If the DC is 15 and your check totals 10, you have failed by 5 (and if climbing… well, you aren’t any more).
This is not a bad rule and can be extended, with success by five, ten, fifteen, and technically more, if your check modifier is more than the DC itself.
This mechanism requires totally the d20 roll and modifiers, then subtracting the check DC to find the specific degree of success or failure. This isn’t usually difficult, but it can be after a few hours of gaming (and frankly, drinking, in at least some of the groups I play with). However, there is a quicker and easier way.
Degrees of Success by Examination
I have noticed that often in play success or failure is evident almost immediately on rolling. If my fighter is trying to hit an opponent and I roll an 18, I have almost certainly hit. If I have to check carefully it means that we probably shouldn’t be trying to hit it with weapons, and should either fall back on magic, bug out, or suddenly get very clever. Similarly, a 4 almost never hits, regardless of modifiers, because that means we have horribly outmatched our opponents.
Now, in both cases I could work out the total check result and compare to the DC (Armor Class in the case of an attack), then subtract one from the other (the check minus the DC for the successful one, or the DC minus the check for the failure, unless I’m willing to deal with negative numbers). Annoying.
Consider for a moment, if I need a natural 20 to hit my target, I can only ever barely succeed (succeed by 0). If I can succeed on a natural 18-20, I can succeed by 0-2. Similarly, if I fail, I will fail by 1-19 or 1-17 (depending whether my DC was 20 or 18, respectively).
Say, that looks very suggestive.
Instead of working out exactly how much I succeeded by, why not just look at the natural roll and work with that?
A natural 1 will almost always fail, so it should be a minor failure. If I fail on a natural 19, though, I am either desperate or foolish to try it, because a natural 19 being a failure almost never happens. In fact, it happens exactly as often as rolling a natural 1 when I need a natural 20 to succeed. In exactly the same way, a success on a natural 2 will almost never happen; it means I could have succeeded on a natural 1 or lower.
Consider the following table.
This allows for up to eight levels of success and failure, and it probably looks exactly backward, I am sure. When you succeed, you want to succeed with the lowest possible roll on the die. 18-20 gives you pretty certain success, but the effect is likely to be small. However, it is mathematically the same as rolling 0-2 points higher than you needed to beat your check DC. In the same way, if you’re going to fail you probably want to fail with the lowest roll possible in order to minimize the effect of failure. Rolling a natural one and failing is in this model the equivalent of rolling one point less than you needed.
A modifier still has the same overall effect, since we look only at the natural die roll. A +1 improves my chance of success, so I might succeed on a natural 13 where previously I would have failed… but now it’s the difference between a natural failure and a natural success. If I could previously have succeeded only on a natural 8 or higher, that +1 means I could succeed on a natural 7 now and possibly get a better result… just as under the common subtractive model that +1 might move me from a total check result of 23 to 24 and thus move me from “DC+5″ result to “DC+10″ result.
I am inclined to drop rules such as “20 always hits” and “1 always fails”. If you need more than a natural 20 to do something you probably shouldn’t be doing it in the first place, and if you can only fail on a natural 1 because the rule says so you should be allowed to succeed without the check.
Basic task resolution is just that: basic. It is entirely serviceable, but many people like to see a little more differentiation in degrees of success.
To do that, the obvious mechanism is to compare the total check result against the Difficulty Class of the check. Easily explained, but I find it annoying in practice because common cases, obvious success and obvious failure, still require a lot of work to be done. Depending on the mental state (fatigue and inebriation) this could be more trouble than people are willing to do, and error prone.
Instead, going to an examination method allows the obvious success and failure cases to be handled very quickly using a standardized table. The only time it is necessary to work out the exactly check result is when it is a borderline case… in which case the additional effort determines whether it is a greater success or greater failure.
This does take away from the “20 is always best!” idea, and I admit I’m not really excited about that because I like rolling natural 20s as much as the next guy. I think in the end that if I want to include degrees of success in a convenient and functional way, I can live without “20 is always best!” pretty easily.