This was a subject matter that was covered on my site recently but one I wanted to pitch on this forum because it's quite possible it's a stupid idea.

I've been recently play-testing with my gaming group a game I've been working on that is something of a cross between The Invisible College meets Hunter the Reckoning original, mechanics and of invisible college, but the themes and lore more inspired by Reckoning.

One of the ideas I discussed and which my players thought was kind of cool was, what if every mundane skill worked like Attainment in Invisible College? mainly what if skills had a range of 0-100, the bonus for skill checks being 1/10th the skill score.

And at the end of every session you tally every skill you rolled with success and roll a percentile equal or greater, it goes up by +1.

This to me seems like it'd have a much more natural skill progression than simply assigning skill points every level up for mundane stuff, you get better at using skills, by literally using those skills.

Sounds like a mix of Star Frontiers and RuneQuest.

Collage =/= College

It could work. I would also add two successes for every critical failure rolled. You learn as much, or more from mistakes than you do from success.

I see some problems with this approach.

1) Why does it matter whether you succeeded or not? You can learn just as much from failure. And as a starting character, you'll be failing a lot, and as GM there is good reason to reward failure.

2) Having to roll to see if your skills advance completely sucks. I say this from experience as a former Runequest player. It's hard enough to be a starting character without being unable to spend your xp. Especially when you then lose that xp.

3) By its nature, this will cause some players to advance faster than others, for no reason other than sheer dumb luck. That's not good. And if you say "over time it will all even out," you're wrong. The odds are exactly the same for each roll. The dice have no memory, and someone who pulls ahead because of a streak of lucky rolls is likely to stay ahead. And someone who falls behind is unlikely to catch up.

4) If you think it's bad when people get shafted once on advancement, wait until it happens twice right out of the gate. And it will happen twice to somebody. I wouldn't expect that player to come back.

5) How do you buy up new skills that you don't already have? What if nobody bought Cartography?

6) In Runequest, this encouraged what was called the "golf bag" approach, where players carried around a (figurative) golf bag full of different weapons, each of which they would use exactly once per session, to maximize their chances of getting a successful advancement check.

7) Some find the extra bookkeeping during play to be distracting. And the time spent rolling skill advancement checks comes out of game time.

Basically, this is point buy with extra steps. And those extra steps are problematic.

I wouldn't use it for normal skill progression but there are two ways I think it might be useful.

First is getting a new skill.  IRL, there are several skills that have a first success barrier that can frustrate novices.  A good example is learning to ride a bike.

The second is overcoming a developmental plateau.  In your example of 0-100, what if 70 is the normal max someone can obtain through the usual skill development but you can break through this ceiling through intense training and pushing yourself to the limit until you finally overcome the limit. 

I would also add that even a failure shouldn't result in no positive results.  If a player rolls and fails to achieve attainment, then perhaps a point could be added to attempt counter that accumulates until the success happens.  Maybe this can be a roll modifier that makes the next attempt a little easier or maybe it can give an automatic success once a player has made 10 or so attempts.

This is similar to how Call of Cthulhu does it.  It works but there are some problems.

It encourages players to roll as many checks for as many skills as possible hoping to get a success.  Characters will do things just to get rolls so they can maybe advance. 

Certain skills will go up much faster than others.  It will be whatever skills the GM calls for the most rolls with.  In CoC this is usually spot hidden, listen and library use.  Skills that rarely come into play will rarely if ever advance.   

If you're interested in keeping players at least potentially in sync with each other, the way I came up with was to have one skill advance per session but you got to roll in your choice of order based on skills you meaningfully used (whether successfully or unsuccessfully) in game.   You kept rolling until you ran out or succeeding in upping a skill.   Any sessions where no roll was successful was "banked" for the next session(s) until it succeeded.  For example, if you failed two games in a row, you'd be able to roll potentially three increases on different skills if you got lucky after the third game.  It's not perfect as you'd still only increase an individual skill once in that third session (as opposed to potentially multiple times if repeatedly successful over multiple games) but at least the attempts we're lost completely and characters stayed roughly on par with each other.   This was a d20 system though so for percentile it may be better to have multiple successful rolls allowable per session due to the greater granularity there.