It’s unfortunate that most people are so math-challenged, that you needed to run brute force calculations and provide 3 graphs to prove that if you have a 50% chance of success, it doesn’t matter if you roll d20, 3d6, or d100 – you’re odds of success, are still 50%. I suspect people who think 3d6 is less “swingy” than d20 because they are much more likely to roll a 1 on d20 than a 3 on 3d6 won’t be convinced by your graphs – if they could interpret them, they wouldn’t need convincing in the first place… That said, dice with a linear distribution are “swingy” in that they are terrible at handling modifiers in situations when the base odds of succeeding are over 80% or under 20%. This probably doesn’t matter in fantasy genres or heroic games, but it is a particularly thorny problem in “gritty” “realistic modern” or “hard SF” RPGs. Those games are often “crunchy”, with lots of combat. With the advent of automatic weapons, firefights basically became pitched battles between opposing sides hiding in cover (nobody wants to die), exchanging low probability of success gunfire. Once one side is suppressed or breaks, the other side leaves cover with relative impunity, to close in for a high percentage point-blank kill. This has been SOP for almost a century now and with no end in sight. It’s very difficult to model this with linear dice distributions – frankly, using d20 or d100 in that instance, is a terrible mistake and frankly, just asking for a trainwreck. A -10 DM for light cover and -20 DM for heavy cover is often meaningless when units have an 80% chance of hitting. But if your base chance to hit is 20%, those DMs are massive. So do you start halving, quartering, doubling odds? Those are very coarse modifiers that are also grossly inaccurate. The only way to attain believable results is to use decimal multipliers, like x0.8. The beauty of bell curve dice is that +1 or +2 DM smoothly scales along with the odds of success or failure. We observe this behavior when you shift the median time and time again in the real world. In statistics, almost all random samplings of populations follow a normal distribution – whether your histogram is for North American consumer spending habits or the air-speed velocity of African swallows… So why not use a dice system that provides a normal distribution? Even as you shift the median result (+ or – DM), it remains a normal distribution – just as in real life…