Quote Originally Posted by AgentPaper View Post
So your issue is that ability scores matter?
My issue is that high ability scores (which you either roll randomly or buy with diminishing returns at character creation and that do not advance as you level--that we know of yet, at least) outweigh skill training (which you have full control over and that does advance as you level) at the low levels and compose a large proportion of your bonus at high levels. This means that (A) choices you make at character creation can be more important than choices as you level and (B) the overspecialization this encourages is a bad thing.

Point A is the "locked-in builds" problem in 3e that was so derided during the 4e transition: if a fighter picks some bad feats at low levels and wants to change his style at mid levels, or if a ranger rolls a low Wis and can't boost it high enough to cast spells, oh well, tough luck, you're stuck with it. Point-buy for ability scores and static HP were made the standard methods for 4e for the same reason (being at a disadvantage at 20th for some bad rolls at 1st is not a good thing), and yet now they're making ability scores more important than they ever were there.

Point B is the "fighter feat trees" and the SAD vs. MAD problem: you can pick a bunch of synergistic options and focus your resources, or pick a bunch of non-synergistic options and spread your resources around, and the former is almost always better. Given the choice between being able to do godlike things in one of six areas or merely difficult things in three or four of those six areas, the logical choice is the former, because D&D rewards specialization already and you're going to be doing stuff within your niche most of the time anyway.

Besides that, you got into a ton of stuff about scaling and the range of randomness, which we've already covered. If you want unexpected results to not happen, then you can't use a random element for your determiner. If you really don't like randomness, then just houserule that everyone takes 10 on all skill checks. Now the fighter will jump the same height every time, the wizard won't perform super-human feats some of the time, and all that other stuff that you mentioned.

Of course, you'll likely also have a player revolt on your hands, because all of the stuff you mentioned is secondary in importance to having a fun, balanced game.
Please don't put words in my mouth. Randomness is a good thing, and constant success is boring. However, having a low range of randomness on a RNG only twenty points long doesn't work. In 3e, static skill DCs go from "easy" at DC 5 to "nearly impossible" at DC 40 and opposed checks can have you rolling against DCs in the 50s or 60s, which means there are cutoff points: someone with a +25 bonus is incapable of failing some checks that someone with a +0 bonus is incapable of succeeding at, while they in turn are incapable of succeeding at checks that someone with a +50 bonus is incapable of failing. If DC 40 is "nearly impossible" for anyone who isn't superhuman, than it's actually nearly impossible for normal people with less than a +20 to hit; conversely, once you get good at some skill, you're actually good at it and don't fail simple tasks anymore.

The 5e system tries to achieve the same effect by merely shrinking the range of numbers proportionally...but the d20 is still 20 points wide, so the skill range has gone from being roughly 2 RNGs wide for static DCs and 3 wide for opposed checks to being 1.25 RNGs wide for static DCs and 1.5 wide for opposed checks. On the high end, there's no differentiation between high level and low level, expert and non-expert: it's possible for a character of any class with a high stat or a low stat and training to hit the top of the scale from 1st level, and someone who maxes out training in a certain skill is less than 50% away from someone with the same stat with no training. On the low end, it's no longer guaranteed to succeed at normal tasks; like crit fumbles or a dice pool, no matter how skilled you get the chance for failure is always there.

That can and does work in other, "grittier" games, but not in a game where "good at skills" is one of the three protected niches and the main schtick of one of the four iconic classes, and not in a game where you're traditionally superhuman by the mid levels, whether explicitly (4e's explicit paragon and epic tiers) or implicitly (3e's implicit tiers and AD&D's "normal people are 0th level" thing) and thus better in your area of focus than mere mortals.

If you have a better system you'd like to propose, that's fine, but I warn you that it'll likely get ripped apart by everyone here just like you've ripped apart the bounded accuracy system.
As far as a better system goes, I point you to the BAB/skills/etc. of 3e and 4e, and any variants thereof like PF, SWSE, etc. There was no need to condense the numerical range like that; if for whatever reason you want a single orc to still be a meaningful challenge to someone who should be slaying armies of them, just use the 4e system and drop the +1/2 level from things. It may have had a level treadmill where your chances of success don't really change as you level, but at least it was a treadmill against on-level opponents, not against everything.

The problems with those systems were never the systems themselves, but the spells, items, and such attached to them. Use some method to remove the reliance on stat-boosters (inherent bonuses in 4e, a simple houserule in 3e) and remove the stupidly-high bonuses in 3e like glibness and such--two trivial things to do if you're making a new edition based on one of them--and the problem basically goes away. The only difference between the two, really, is whether you want someone who doesn't specifically train at something to still have a chance of success because you assume he's been practicing off-camera (the SWSE/4e way) or not to have a chance of success because being trained means you're better than a normal person and only training gets you the extraordinary results (the 3e/d20 way), and switching between one and the other requires literally less than a paragraph of rules, maybe one sentence if you're being concise.

The object of good design isn't to eliminate all flaws, but to ensure that the flaws you have detract from the game as little as possible.
Having the numerical basis for your whole noncombat minigame be unsound is kinda flawed, don't you think? I don't expect them to abandon the idea, or even take steps to fix it, since we already know WotC is bad at math--how many skill challenge iterations did the 4e dev team go through, again?