Mustrum_Ridcully
Legend
Obfuscating describes it well, I think.One point we haven't made yet is that, sure, the 3d6 distribution is almost a bell curve, but any one test -- to-hit roll, save, whatever -- is a binary test, yes-no, pass-fail.
It doesn't matter whether you roll 1d20, 3d6, or percentile dice, if you have an N% chance of success, that's your chance of success. For the most part, rolling 3d6 simply obfuscates the probabilities.

Modifiers are likewise "obfuscated". A +1 modifer against a DC 10 task changes the probability differently then a +1 modifier on a DC 15 task.
I am not sure that Gnomeworks is using much of D&D or the d20 System as baseline.Look at a GURPS probability table sometime to see what 3d6 looks like. In short, it won't work properly unless you plan on rewriting the whole game. While the bell curve is often thought of as "flattening," in this case, the net result would be that between opponents of very similar ability, small differences would become larger. A +2 point difference is of far greater utility when you can be more confident of die rolls favoring the expected outcome (i.e., the person with the +2 winning).
2AtomicPope:
That rule system sounds really interesting!