When I began my RPGing career, I was a big fan of percentile-based systems for task resolutions. It seemed the most rational: you had X percent chance of making it, so you rolled against that. Seemingly simple and straightforward. Later d10 or d20 based systems (much more sane – 5% granularity is more than enough, thank you) carried that forward, with quick-to-resolve results.
But… there’s a thing about those results. They’re linear, and they’re swingy.
For the most part, once people become proficient at something, they perform relatively reliably, consistently, at that task. Their results cluster together. Sometimes they do a bit better, sometimes a bit worse, and rarely will the needle move much beyond with a horrible fork up or an astounding success. If they’re not proficient, their results will be more varied, but again, they’ll somewhat cluster around a certain point. Because of this, if the difficulty of the task increases, the results don’t have a gradual even decline, they tend to drop precipitously.
So, in game terms, a system that would want to model this would work it out so that a PC’s skill check results would cluster together, and modifiers would be geometrically cumulative.
Let me back up for a second and say, there’s no denying the blatant sense of if you have a 75% chance to hit, then you need a 75% on your percentile die (or 16 or better on a d20). It’s about as blam accurate as you can get. The key, however, is in determining that 75%… especially if you want to model clustering, and if you want to incorporate a Margin of Success system in the game. Because to generate that 75% with a linear die, you’ll have to have some sort of chart or formula to calculate that target number, and that is quite the pain.
Fortunately, of course, we have the beauty of the bell curve to the rescue, in the form of multiple dice. Even 2d10, or 3d6, or the FUDGE dice, makes for something that clusters quite nicely, and adding or subtracting a modifier to the target number there not only varies the actual chance of success based on how hard or easy the chance already was, but multiple modifiers it also cumulatively add up in a non-linear fashion.
And I say that not only more closely models what happens in “reality”, but also makes for an easier story to build and a greater grasp of your character’s abilities. You know about where things will land, most of the time. And as a player, while the big wins are awesome, the times you bork up a supposedly easy roll just feels frustrating and weird.
This, all in all, is another reason why I’m going for a multi-die system in this system design. While curve is a lot different than a pure XdX die system, it shares many of the same advantages of clustering and of modifier variance (along with the visceral feel and ease of die add/subtracting).