I'm working on a dice mechanic/resolution system based off of the Ghost/Echo (hereafter shortened to G/E) tabletop RPG. Specifically, since G/E can be a little harsh with dealing out consequences and failure, I was hoping to soften the system and add a little more player control, as well as offer the chance for players to evolve their characters into something unique, right from creation. So, here's the mechanic: Players roll 2d12 against the two statistics for their character (each is a number from 2-11, and may be rolled above or below depending on the nature of the action attempted, rolling your stat exactly always fails). Depending on the success for that roll, they add dice to the pool rolled for a modified G/E style action.

The acting player gets two dice anyhow, and I am debating offering a bonus die for each success, or a single bonus die for succeeding on both of the statistic-compared rolls.

One the size of the dice pool is set, the entire pool is rolled, and the players are allowed to assign rolled dice to a goal and a danger. Assigned results are judged as follows:

1-4 means the attempted goal fails, or the danger comes true.

5-8 is a partial success at the goal, or partially avoiding the danger.

9-12 means the goal is achieved, or the danger avoided.

My concerns are twofold:

Firstly, that the two-stage action is too complicated, with two rolls to judge separately before anything can happen.

Secondly, that the statistics involved go too far in softening the game. I've run some basic simulations, and the approximate statistics follow:

                 2 dice     (up to) 3 dice      (up to) 4 dice
failure          ~33%            ~25%               ~20%

partial          ~33%            ~35%               ~35%          

success          ~33%            ~40%               ~45%

I'd appreciate any advice that addresses my concerns or offers to refine my simulation (right now the first roll is statistically modeled as sign(1d12-1d12), where >0 is a success).