Unthinkable? Ditching Auto-Success and Failure

For an age, indeed probably since the pre-history of D&D, it has been a convention that "a 20 always hits". Okay, this was not actually quite true, because in the 1st Ed DMG there were actually combat tables that required you to roll a 21 on d20 to score a hit but these were really extreme - and that oddity was removed for 2nd Edition and beyond.

So, a natural-20 always hits and a natural-1 always fails. And that is good, that is the way it has always been, and that is the way it always shall be.

But maybe it shouldn't.

See, here's the thing: the rule about a natural-20 always hitting only matters in those cases where a 20 wouldn't be enough to score a hit - if you have +4 on your attack rolls and the enemy has AC 25 or more. In which case that nat-20 represents the hugely lucky hit against all odds. Which is cool enough, I guess... except that in that case it almost certainly means that the PCs are badly overmatched anyway, and really need to run away - and scoring one lucky hit won't change that. (And if the target has hit points appropriate to such a high AC, the PCs really aren't going to whittle it down with those one-in-twenty lucky hits.)

Likewise, the rule that a natural-1 always misses only applies in those cases where a 1 wouldn't otherwise miss - in which case the challenge is trivial anyway, so might as well be skipped.

In fact, Blizzard (makers of WoW) did some extensive analysis of success thresholds, and they concluded that a success rate of ~68% (IIRC) gave rise to the most satisfying game experience - that is, a character who is 'good' at something should succeed at level-appropriate challenges on a roll of 7+ on the d20. That probably means that a 'hard' challenge for such a PC should succeed on a 12+ (45%) while an 'easy' challenge for such a PC should succeed on a 2+ (95%). A character who is merely 'average' (not good) at the task should change those thresholds to 7/12/17 for easy/average/hard tasks, while one who is 'poor' should change them to 12/17/22.

In other words, the only time (in the normal order of things) that the auto-success/failure rule should apply anyway would be when a character who is 'poor' at something attempts a 'hard' task - and, in that case, I have no problem whatsoever in simply ruling that the task is beyond his capabilities.

(Of course, in actual use there will be more variation of success values than I've indicated above. After all, 'good', 'average', and 'poor'; and 'easy', 'average' and, 'hard' are all rather nebulous and ill-defined terms. But the principle remains - if you're good enough, and the task easy enough, that you'd succeed on a '1' anyway, there's probably very little interest in that 5% chance of failure; conversely, if you're sufficiently outmatched that even a '20' isn't really enough to score a success, why bother rolling?)

So, I'm inclined to think that that's actually a rule that would be better consigned to history.