It was an imteresting concept that basically vanished from my radar until the appearance of the Cortex games (in my case, I first saw it again with the Serenity RPG).
I played Serenity a few times, so I'm more familiar with the way step dice work here.
There's typically a difficulty, and the player rolls a die for their attribute (with size determined by how good their attribute is), plus a die for their skill (again reflected by skill expertise), and maybe another die due to equipment or some quirky advantage. Add the total, and if the dice beat the success, then the attempt succeeds...if not, failure.
System 4 has been working on a similar premise. You roll a bunch of dice and if any of them beat a '4' they count as a success...characters weak in a specific field might only be rolling a d4 in that attempt, average characters might be rolling a d6 or d8, good characters might roll d10 or might have multiple dice to roll. If you have multiple dice, each one rolling greater than a 4 might contribute successes to the task. For those dice capable of rolling far greater than a 4, multiple successes might come from a single die...rolling an 8 or higher contributes 2 successes to the pool, rolling a 12 contributes 3 successes.
I've just spent over a week trying to get this working in my mind...aiming to get the dice balanced, making things a natural progression.
I was wondering whether to allow one character to buy a bunch of d4s as they improve their character, while another might keep a single die and improve that...then wondered how the numbers might stack up.
I started developing intricate systems where a player might have a limited range of choices in their progression paths. Perhaps they cannot have more dice of one size than the next...ie. At least as many d4s as d6s, at least as many d6s as d8s, etc... but every way I wrote these rules, they just seemed clunky and unnecessarily complicated.
I like the core essence of this system, you roll a bunch of dice, anything that meets a 4 (or better) earns a success...odds add a penalty/"but" result to the outcome, while evens add a bonus/"and" result. D6 has a 50/50 chance of succeeding, D4 is worse, D8 is better.
My current thoughts for step dice involve a basic chart that all characters progress through, level by level. Which is basically the way Earthdawn handled things, with some differences in the way results were read from those dice...and I'm not following the specific progression they use.
Lvl 1 - d4
Lvl 2 - d4 + d4
Lvl 3 - d4 + d6
Lvl 4 - d4 + d6 + d4
Lvl 5 - d4 + d6 + d6
Lvl 6 - d4 + d6 + d8
Lvl 7 - d4 + d6 + d8 + d4
Lvl 8 - d4 + d6 + d8 + d6
Lvl 9 - d4 + d6 + d8 + d8
Lvl 10 - d4 + d6 + d8 + d10
Running the numbers on this set up, each level has a better chance of producing more successs, so that's a good thing, the progression is also pretty consistent. The only down side, is that the players won't be rolling as many dice as I had envisioned. In turn, this means that there won't be as many bonus or penalty results applied to the outcome. When more dice are rolled, there are more bonuses/penalties floating around, and more chance to do imteresting things.
I also thought of the option where you'd simply roll a number of d6s equal to your skill level. This means lots of dice hitting the table and a lot of potential bonuses and penalties (but if you cancel positive and negatives the bell curve effect pulls the average result to zero). Maybe that's a better way to handle things.
If a d4 has a 25% chance of success, and a d6 has a 50% chance of success, maybe we could increase the number of dice more slowly by adding a d4 to the pool on odd levels, then bumping it up to a d6 on even levels. It's a halfway point between the options described so far.
More thinking to do...