It seems much more intuitive (as someone put it, "more of a game"), and I feel the old way of doing Art advancement would keep magi power levels to a slightly less mind-bending power level.
Before I go to "try it and find out", has someone done this, and would someone with greater knowledge of the system comment whether it has unforeseen consequences?
How far are you borrowing from 2nd ed ?
If you borrow both Art progressions and book writing from 2nd ed, you will see that there is "soft" cap about Art progression between 20 to 30 (broad range). In 2nd ed, a written Summa has a single characteristic, its level. And a author can only write book up to half his/her skill level, whether hermetic or mundane.
So archmage Piros with 40 in Ignem can only write a book up to level 20.
So once a mage reaches 20 by reading Piros' book, the only way to progress is through vis experiment, very slow since the mage need to generate a study total resulting from dstress x pawns of vis used (from 1 to 3 max). You can easily how statistically, it becomes very slow to progress above 20 - and you can wonder if it is really possible to achieve 40 in one Art. And definitely not in two Arts.
Some Twilight can result in an increased understanding allowing the mage to write up to 2/3 of his Art, possibly 3/4, 5/6. But again statistically very rare since if you play without fudging dice, the mage need to have a positive outcome and that the positive outcome is this increased understanding.
If you want a low-medium power Saga, it is a good choice as it will simplify a lot book keeping, book trading and such. You will have to tweak some virtues from 5ed though - any virtue granting Art XP need to be tweaked to give level instead of XP or simply removed (I won't go through all virtues to check which one work or not).
It means the progression curve is flatter and cap much earlier than in 5th ed. In 5th ed, there is no real cap, only slow progression as the mage is going through lower and lower quality tractatus to gather XP to progress from 49 to 50. No such thing in 2nd ed: you will need to run an exploding dice of 18 (1 and 9) x3 to reach a 54, that will grant you a single level. I let you figure out the statistic of this happening versus the number of Twilight episodes (and I let you imagine the amount of vis burnt just to reach 49, and the quantity of Twilight episodes...).
It also means that high level spells (above level 40 are much rarer since there are few magi competent enough to cast them).
I have played ArM3 for quite a bit many years ago, and recently played a short ArM3 saga with my troupe out of nostalgia. As Ezechiel3571 writes, the two big things with ArM3 Art advancement are:
Books and Vis were better balanced against each other; in particular, books (barring V&Fs) were safer and cheaper - and good books just better - than vis, until you could no longer learn from them. But after that, they were of no use at all (which makes sense: you can learn from others until a certain point, then you must learn on your own). Canonically, it was easy to find books that would get you to level 15, hard to find books taking you to level 20, and exceptionally hard to find books taking you beyond that (but ArM3 specifically wrote they existed, and indeed, you could have them in your starting Covenant library, if your Covenant was an Autumn or Winter one).
Whether by books or vis, growing your Arts beyond your low 20s was sufficiently hard that it was rarely worth it. But it was possible, and it was quite possible to go beyond 30 too - though a score of 30 in an Art was much more exceptional in ArM3 than in ArM5; and a Level 70 Ritual (with an extra requisite!) like The Shadow of Human Life Renewed something really awe-inspiring. This (and the much lower oomph of Art Affinities) made generalists more appealing than in 5th edition, and starting magi more competent compared to their elders.
A few things to keep in mind. Some Familiar "Bond Qualities" could boost your Art scores (in particular, you could get up to +6 to a single Art in that way). Rolling Discovery when experimenting had a pretty large chance of yielding a 1-level increase in one Art and, with a Risk Modifier of 2 or 3, a small chance of seeing a 2-level increase. So, no matter how high your Art score was, you could always experiment for an expected increase of 1/15 of an Art Level. Finally, even without Twilight experiences, someone with Strong Writer could write a book to 2/3 his own Art level.
The biggest difference is that advancement stops being guaranteed once you exceed the books. There's no exp so there is no partial advancement. If your stress + aura multiplied by the 1-3 Vis you are using does not exceed your current Art, you get nothing.
All the time I spent playing 3rd is the reason I favor Generalist. Books actually felt even more valuable while you could use them because of the lower soft cap and guaranteed increase, meaning the guys who studied the best book every season would end up much more powerful overall than the guys who just focused on an Art or two (switching to Vis study). Once you exceed all the books, increases in power by Arts becomes a very expensive gamble with low returns.
For the Art limits you are lowering the "soft cap" and "hard(ish) cap" by about 10 each, from 30's and 50ish in 5th to 20's and 40ish.
Just be prepared to deal with the player who has a ton of roll bad luck, burns through a few years and couple dozen pawns, and doesn't increase an Art a single point. Seen it happen in a single session and it was a close thing that the table didn't get flipped.
You may want to look up 4ed, which first introduced xp for arts but retained an optional full-level advancement system following 3ed. There may be some comparison there.
From my experience, that's not such a big deal in the long run.
Even in ArM5 you have a lot of uncertainty in terms of xp gained from vis study (especially because a botch might blow up your lab). If I have to gain 10 levels in an Art, it makes little difference if I am gaining 0.1 x 1d10 levels per season, or if every season I have a 55% chance of advancing 1 level and a 45% chance of not advancing at all. The latter has a little more variance in the time taken, and it can certainly feel more frustrating to gain nothing at all (then again, so does falling just 0,1 levels short of a full advancement in the first system) but the difference is minor.
I really think that a much bigger issue is that, in an Aura of 5 and barring any V&Fs, to advance from 39 to 40 in an Art in ArM3 via vis study you had to spend an average of 1/0.059 seasons, i.e. almost 17 (in fact, roughly at this point just performing lab experiments became a better choice) while in ArM5 you may be spending 4 or so.
That matters a little bit, but the "you get nothing" is the thing that makes it unfun. The guaranteed advancement does cause things to be faster, but that's a side effect of fixing the glaring problem with gameplay. A player might or might not choose to get 2xp/season for 20 seasons to get to lvl 40. But that's an informed decision where they are making the decision that suits them. But in ArM3, they don't know if it will take 1 season or 100 hundred seasons. And they will typically have many, many seasons where they literally do nothing but throw away vis without any benefit.
That's not good gameplay.
... In my experience, the system from 2nd/3rd tended to discourage specialists.
No, that's not exactly the case. You'd often start out as a specialist. But after some years, once libraries couldn't help you (as) much anymore, people would start widening their base of Arts.
Eventually, you might be forced to study purely from Vis anyway. But elder magi would generally be fairly widely competent. Unlike some of the magi I've seen in ArM5 that just kept bulling forward with the same few Arts. Which I have tended to see as a benefit.
I was pondering how to improve slightly the 2nd system, without making it completely wide. So far, I came with a simple implement that I call Cumulative Learning: when learning from vis, if the mage fails, he gains a +1 bonus to his dice next vis study roll for the same Art. During his failed season, the magi took notes, improved slightly his understanding, but he still needs to validate some hypothesis and explore further to solidify his knowledge.
Without having tested it, I cannot decide if the bonus applies only on consecutive seasons of research, or if season on other topics (like vis extraction) won't disrupt the research. This is one area for tweaking if it seems too generous.
Without doing a lot of math, a magi who reached 20 (assuming the best book he could find in 2nd ed), will need on average 4 to 5 seasons of vis study to improve by one level: after 3 seasons of failing, he has a +3 to his vis dice, so a 4 will give him a study total of 7 x 3 (3 pawns) =21 (plus aura). Let's say one level every 10 seasons, considering other activities. So 100 seasons or 25 years to reach 30.
Once he reaches 30, he should still be able to progress every 6-7 seasons using the same parameters. If we take in account vis extraction and other activities, he might still progress every 3-4 years. So another 40 years to reach 40.
Assuming that he reached level 20 at 50, he will be 120 when he reaches 40 (without much optimisation), he would have used roughly 300 paws. Of course, there is likely several Twilight episodes in between which will throw a wrench, but it becomes too random to assess the impact.
That's all I have for now.
Why would you want to do this?
To me, it looks like a more complex way to achieve similar long-term progression as 5ed.
The most significant difference is that you do not allow Affinity with Art, but we could just remove that 