It's nice to have a more-granular form of advantage/disadvantage. But as a statistician, I think the "reroll if your first roll is too high/low" dynamic seems bad; the distribution has funny kinks in it, and it feels as if it "punishes/rewards" you for rolling too well/poorly the first time.
I've been thinking of easy-to-implement ideas that are "cleaner", and here's what I have:
Roll two d20 of different colors; a "main" and "auxiliary" one. If the auxiliary die shows 6-15, use the result from the main one. Otherwise, use disadvantage/advantage as normal (that is, for disadvantage, use the lower of the two, and for advantage, the higher of the two.)
I've been thinking of easy-to-implement ideas that are "cleaner", and here's what I have:
Roll two d20 of different colors; a "main" and "auxiliary" one. If the auxiliary die shows 6-15, use the result from the main one. Otherwise, use disadvantage/advantage as normal (that is, for disadvantage, use the lower of the two, and for advantage, the higher of the two.)