Joseph - thank you for taking the time to share your philosophy. I absolutely agree with you.
I think the challenge is whether or not to scale challenges. Metagame theory would suggest that all things being equal, a 10th level Thief should be challenged to the same degree as a 1st level one. Otherwise, what's the point? Sure, if the same thief goes into the dungeon of his 1st level days to finally overcome the lock the stymied him he should be able to pop it without issue, right? Because it's a first level challenge, right?
Or is it?
Perhaps the narrative reason your 1st level thief was unable to pick the lock is because only a 10th level thief could actually succeed.
This poses a bigger question. Why should character's abilities improve at all? In real life there is not an unending progression. You can't simply progress and keep adding skill levels. However many levels you define, there is a limit. Out at a ski school in Colorado they identified about 8 levels of ability, which I thought was pretty fair. These levels were quantifiable, and I could easily recognize where I and my students fell. Through that spectrum, there are about 5 orders of magnitude of expertise (Untrained, Trained, Professional, Expert, Master).
Each of these orders of magnitude is, IMO, best modeled by mapping to the standard deviation of whatever dice mechanic you use. Standard Deviation for d20 is 6 (5.91). So, using a very simple progression DC 11 is Untrained, DC 17 is Trained, DC 23 is Expert, and DC 29 is Master.
Using your No Skills guideline, then somehow something characteristic to the class should progress from +0 to +18 over the course of the character's career (6 points per std. deviation).
There are a large number of assumptions here, but that is the gist of the math and assumes that you do want to scale the challenges. Why? Because what's the point of character with +18 making a test against a DC 11 challenge other than comedic value or if they are under extreme conditions such that the challenge is, in fact, challenging - thus making my point. IF the Judge is creating an adventure, planned or ad-hoc, the challenges are going to be such that it will be interesting regardless of the level of the character. The Judge will rule that "due to the age of the blah blah, and the stormy conditions, and the alien construction the Thief has a DC 10 pick locks". Clearly if this is a 10th level adventure a 1st level Thief could not possibly hope to do it.
How objective do you want the system to be? If you want it to be completely objective, then consistency is king, math is your friend, and you can very neatly define a mechanic that will work.
If you want it to be subjective, which you clearly do, then I suggest the following:
Define challenges in terms of DC and who can attempt. Any is viable, and implies untrained attempts.
If the action is part of your background, roll d20. Otherwise roll d14 (note the 6 point difference - std deviation).
Perception is always d20. Everything else is subjective.
The challenges assume a certain level of adventurer. If there is a difference, adjust accordingly one for one (i.e. a 1st level Thief in a 5th level adventure will find all the DCs adjusted by +4).
- Tough Lock (Thief:Agility DC 17 to pick, Any:Strength DC 20 to break)
- Rough Cliff (Thief:Agility DC 11 to climb, Any:Strength DC 17 to climb)
- Find Clue (Thief/Elf:Intelligence DC 17 to spot, Halfling:Luck DC 11 to find, Any:Luck DC 20 to find)
This encourages the Judge to evaluate each adventure and the challenge it presents rather than trying to fit it into a specific progression. Will this create situations that some characters have no chance? Yes. A Wizard should not be able to break down a door unless he burns luck. A warrior should not be able to sneak unless he does the same.
And if you do this, then Saving Throws should probably follow the same model. Vary the bonuses from +0 to +3, flat (no change with level), and then the DC of the save is either set by the Judge (for traps and such) or the level of the spellcaster (relative to the target).
I think that's closer to the spirit of what you are looking for, based on the adventures that were written (which led me here as I was trying to understand why in the first room of the first adventure it said DC 15 Pick Locks for a Thief when in the beta it has % skills). After I got over the confusion about the lack of a % chance, I asked myself why you even needed to specify Pick Locks. If the Thief is the only one that can do it, why not just say DC 15 Thief. If, based on alignment, some thieves are better at certain things than others then why not simplify and say Law thieves are +1 at X and Chaos thieves are +1 at Y.
However, when you look at it this way... combat must follow the same idea. Essentially, all monsters/traps/challenges become relative to the adventure level, following the same math rules. Most challenges would be +/- 3 (6 point spread), tougher/easier would be +/-6, and extreme would be +/- 9.
A Dragon therefore might be written like:
Dragon (AC 20, HD 9, HP 45, Attack: Bite +6 (2d8+4)...). Since it is HD 9, it assumes a level 9 adventure. If placed in a level 6 adventure, all rolls it makes would be at +3.
Such a relative system might seem okay at first, but can be challenging, especially when you mix elements (such as a horde of orcs attacking a 6th level party). Objective systems really shine here.
I think what you have put together is somewhat of a hybrid approach - clearly more objective for spellcasting and combat, more subjective for skills. I think that's okay. There is a relatively clear line that I think can clearly work.
Is there a simple answer? Obviously not, but hopefully this helps offer some perspective.