Failures of D20

Matt Miller 

Part of what makes the D20 work is the variable DC. Yet, the use of DC increments on 5 (such as for the falling tables) suggests that d20 is excessively granular.

Rob Hicks 

The d20 system, as I saw it, was a way of doing a percentile in 5% increments. It makes it easy from a game design perspective to understand how much of a bonus you are giving across the board.

 Matt Miller 

That is one of the great virtues of d20: A +1 is a +1. But a +1 just isn't a big enough deal. Our minds can't grasp changes in probability that smaller (Our minds max out at about a 10:1 ratio). So get an effect we can intuit as significant, DnD requires math; it requires 'picking up the pennies' of +1 bonuses to get a significant effect. +1 on a d10 works, +1 on a d12 might also.

Rob Hicks

[D20] has too much variance. You need two to four categories of variance. Either "success vs failure," or "success vs failure with critical success and critical failure". The d20 added too much variation on that, which led to a system that was more crunch than flavor.

Matt Miller

You can lower the variance on a d20 by using multiple dice. DnD originally used 2d6, some games use 3d6, FATE uses 4dF. All generate bell-curves that make the outliers possible but uncommon. (As 7 is the most common in Settlers). But for counted variance, the range matters. 2d6 provides 2-12, 3d6 3-18, Fate -4 to +4.  

Matt Miller

Using -4 to +4 provides a range of 9, and that may simply be enough. OD&D abbreviated into -1,0,+1, and 3e made it into -5 to +5. But there is a difference between possible outcomes from the dice, and possible outcomes for attributes. For the former, so large a range may not be necessary. But for the latter, a larger range is required...