Being an old-school AD&Der by training, I've been dissatisfied by 3e, and I'm even more dissatisfied by my own DM's attempts to "fix" certain things in it. One of his complaints is from using a d20. A roll of 20 results in automatic success in combat, meaning that an untrained serf with a pitchfork can do damage a knight in shining (or not) armor 5% of the time. In skills, though, a natural 20 is the best of a character's exertions, meaning that there was a fairly low limit to a character's abilities. In other words, he believes that the limitations of the linear probability curve of a single d20 stinks. His solution, however, was to expand the roll by having the player roll again and add the new roll to the 20. I have this sinking feeling that this really only makes things crazier than they need to be, since, statistically, the second roll can land on _anything_. However, seen as a roll of 2d20, the most probable result is only 21. So my question is: on the whole, how drastic of an effect does exploding the roll like this actually have on the probabilities of a roll? I figure that the answer is actually somewhere between "an average of 21" and "all over the place", but where? I'm asking this question because I'm trying to put together my own "plug'n'play" system based on d20, and I'm wondering whether or not I to keep the d20 roll with a modified form or go for another type of task roll (preferably one with a built-in bell-curve; possibly 3d6 or 2d10). -J. Jensen