The factors at play in the overall feel (and objective results) of a check are fairly numerous, considering how simple the check concept is. They include at least the following, in my opinion:
- The probability distribution of the check. In particular, it is easy to forget that just as important to the game as the expected value is the variance. Higher order "moments" of the distribution are usually less important, but not always. Positive skewness, for example, can impart a "long tail" to successful checks and a small tail to failed checks, which can drastically change how the game feels.
- The offset (i.e. check bonus minus DC in a D&D-like system, and really any system where resolution boils down to a random variable plus a value) has no impact on outcomes for a given fixed value, and in that sense whether a check bonus is +0 or +20 doesn't matter at all to a single check if the offset remains the same. However, for comparisons between creatures (or the same creature at different times/circumstances) where the offset might vary for exactly the same task (usually because the DC is fixed but the check bonus is different), the offset is important to comparisons in feel. In essence, it is a property that is only noticeable when comparing different checks, never within a single check where the offset is, by definition, fixed. As such it cannot be ignored when thinking about feel at a gamewide level, where characters or circumstances may differ, except where it is defined to be fixed for all checks or perhaps in the Solipsism RPG.
- Are checks contests (i.e. opposed checks) or a one-sided affair? This inherently changes the underlying probability distribution of action resolution, as above, in addition to gameflow and, for some, how "active" characters feel with respect to their fates. That is primarily why I put it under a separate point. D&D uses a bit of both, and sometimes which one can feel pretty ad hoc.
- Are checks binary or degree-of-success (dos)? D&D is very muddled on this point. They are usually presented as binary, but the reality is quite a bit more complex. Even the attack roll, the apparent paragon of binary outcomes, isn't quite that because basic degree-of-success is handed off to damage dice, except when it depends on a natural 20 or natural 1, except when beating the target's AC by 10 causes an extra effect, etc. Skill checks are also all over the place: some are truly binary, some give better results on a higher roll, and so on. RangerWickett's earlier post gave some excellent examples of essentially changing what "success" and "failure" mean depending on the situation (including the character performing the action), but at a system level in D&D there hasn't been much in the way of guidelines for how or when to do something like this. That makes it a DM art (yay!) but without much support for helping DMs learn the art (boo!) or apply it with consistency.
- How simple/complex and slow/fast the sytem is in play. A check is the "elementary particle" of mechanical resolution, and at a basic level much of the game is a composite of these. Besides the practical impact on time, I think these can color our perspectives of what a check means within the game's world, its degree of abstraction, etc. For example, in my opinion a d20 system where 11+ is success, 1 a critical failure, and 20 a critical success has a different feel than one that is d20 + mod vs. DC beyond just the different probability distribution. The former is probably more abstract, and definitely less sensitive to in-game circumstances, than the latter. A more interesting example is the Jenga-based system of Dread, where the actual complexity of physically performing a "check" increases (or at least becomes more fraught) as the narrative progresses. A great deal of Dread's unique feel comes from the dynamic complexity and tension of the physical performance, and not just its mechanical similarity to increasing the DC for a die roll.
With respect to the d20 system and bounded accuracy, as currently implemented in the playtest, I think it will pay to draw closer attention to, and write better guidelines for, degrees of success. (Tuning the details of the boundedness itself is also, of course, critical.) That lets one explode the tighter range of offsets into a still rich variety of outcomes. The increased mechanical weight placed on damage in D&D next is an example of this, but it doesn't really address any area other than hit points or those that effectively utilize hit points, like expertise dice and spell hp thresholds in the current playtest. I'd like 5e to find a clearer voice for dos, one that gives some pragmatic structure while still paying a bit of homage to its schizophrenic past. I think we will see this leveraged a bit more fully using advantage/disadvantage, and that's a good start. That advantage/disadvantage seems to be widely popular so far is also, I think, a sign the community is willing to accept a few new organizing principles when the benefit is fairly clear. It would have to be carefully done, however. For example, what is the difference between a crit vs. succeeding by 10 or more? In order to keep both (if D&D drops crits there will be riots) they would probably each need their own story that feels natural. For example, a crit represents a creature's absolute best effort (it depends only on the self), and I could appreciate that having a different impact than a creature who simply succeeds very well because the task was easy. An alternative, which I don't really like, would be to add "damage" rolls to other checks for determining dos. Many other mechanical options are possible, I'm sure.
The longer I have gamed the more I have come to appreciate resolution mechanics where degrees-of-success have an elegant implementation, because it provides so much mechanical leverage, and helpful prompts, for the DM to introduce interesting descriptions and details that are still grounded in the mechanical proceedings. (That is, obviously, my own taste for one way mechanics should interact with narrative. I know some people like binary results because it places the least constraint on how they get to describe things in a game with randomly determined outcomes, although I think this is essentially on a continuum.) That said, I find the d20 system, and really any game relying primarily on a uniform distribution for resolution, will tend to work the least easily with this design notion because the most obvious way to do so (i.e. the amount the check exceeded or missed the DC) has all the quirks of the d20's extremely large variance relative to its absolute range of outcomes, and can even emphasize them in ungainly ways. This is particularly true with things like skill training, where I've come to prefer that training quickly gives reliability for easy checks, but only slowly adds reliability for harder checks, while retaining a "long tail" that can produce amazing success but with ever-decreasing probability. Systems like 3d6, dice pools, and others with a more bell-shaped distribution can be adapted more naturally to those preferences. Given the heavy abstraction associated with almost every check in an RPG (i.e. the resolution mechanic summarizes the mean effect of potentially many random influences within the game world) such mechanics almost (almost
!) approach forming an objective rational model for resolution, due to the central limit theorem
, given a dice-based RPG. But if I'm running or playing a game, darn it, I'm going to run with the mechanical and genre conceits on their own terms as best I can.