A recent hangup in my work on the Avatar RPG has led me to an idea for an alternate bending sub-mechanic that I may well use for all kinds of Tricks in d20 Rethought in general. The reason for this is that it winds up providing the overall system with a curious flexibility regarding which kinds of interaction a given GM/designer wants the system to focus on- combat, social interaction, magic, etc.
This advantage is enough to override my concerns about the disadvantage this subsytem would bring; namely, a loss of elegance due to it deviating from the core mechanic. How bad is this deviation, you ask? well, the mechanic's about half a step away from being a die pool. (GASP)
First, say that any Trick (i.e. any base effect) has an associated "Trick DC". Trick DCs will work on a different scale than normal DCs, because rather than making an normal check you make an unmodified, exploding* die roll that may or may not use a different kind of die. If the initial roll fails, you have a number of rerolls available equal to the number of relevant feats you've taken- "relevant" meaning they pertain to your particular skill or subskill. Once you roll high enough, you can spend the unused rerolls to upgrade your Trick's result beyond the base effect.
So let's say, hypothetically, that have a Wallrun acrobatics trick that lets me run up or along vertical surfaces for short distances. The base effect lets me move 5 feet upwards or (with a 10-foot running start) 10 feet horizontally with no decrease in my move speed. Each upgrade can be spent to make another 5 ft. of vertical or horizontal movement (limit 2), or to make a jump check perpendicular to the wall with a small penalty (no limit, but penalty on jump check stacks). Wallrun has a fairly easy Trick DC; if we assume that in this particular system Trick Dice are d10s, then the DC would probably be 4. So if I have a pool of 3 Acrobatics trick dice total, and my first roll is a success (it gets a result of 4 or higher), that leaves me with 2 "dice" to spend on upgrades, meaning I could run horizontally along a wall for a length of 20 ft. before having to drop or grab hold of something, or run 15 ft. horizontally and then jump away.
Now, how does this afford an rpg system the "curious flexibility" I was talking about earlier? Let's look at the Avatar RPG for an example. In this RPG each of the four types of "bending" (elemental control) has more feats than a single character could possibly take. At a high level, you could feasibly have 10 bending feats and thus a pool of 10 dice, letting you significantly upgrade the effectiveness of your basic moves and reliably pull of advanced moves with a Trick DC of 8 or 9. Meanwhile, the Manipulation skill probably has only a single 3-feat chain associated, providing a number of basic moves with low Trick DCs. So I can invest a small part of my character into being socially manipulative, and have as many as 4 trick dice on hand when using my small array of Manipulation Tricks. If a splatbook full of social interaction feats comes out, suddenly I can instead play a character who's 100% focused on becoming a master social puppeteer. At high levels this character will instead have a full 10 Trick dice on hand and a large array of Manipulation tricks.
In other words, the "power" of a given area of expertise depends on how much the character can invest in it. A single Acrobatics feat chain allows for moves on par with a real-life le parkour runner. Introduce more acrobatics feat chains and you can have a character who vaults and tumbles through the air like they're in the Matrix- but only if they invest a corresponding amount of their own ability, keeping things balanced. A GM who wants their game to focus on kung-fu action rather than politics can just advise players that the majority of the encounters in this game will be resolved with kung-fu battles and that investing too much unrelated feat chains probably won't pay off.
Though I can't yet speak for d20 Rethought, I know that in the Avatar RPG bending attacks will involve first rolling your Trick Dice and then making a normal check to attack your opponent with said effect. This approach is necessary because it's the best way to allow the attacker create an attack that's extra-effective against a given type of defense, before the target chooses whether to defend and how they'll do it.
As always, thoughts and feedback are appreciated.
*: Whenever an exploding die gets the maximum result, you reroll and add a number equal to 1 less than that result. Roll a 6 on a d6, and you reroll and add 5. The end result is that an exploding die can theoretically get any result, no matter how high, so long as you keep getting the maximum result on your die roll.