3d6 vs 1d20

You'll notice the major difference when you compare degrees of success instead of flat possibility of pass/fail. In 3d6 majority of results cluster around 11, so if you're rolling against 10 then most of your rolls are going to be one or two points away from it. Meanwhile on d20 you're as likely to succeed by one as fail by ten. This makes 3d6 favor the more skilled opponent in contested rolls, where d20 is more swingy.

My group and I find that combat becomes more interesting or dynamic when you can plan for exchanges rather than there being an equal chance that you'll stab yourself or curbstomp the opponent immediately.
Maybe we're just weirdos.

In a vacuum, there's no reason to pick either. Are we discussing 3d6 + modifier vs target number? 1d20 roll under? What other mechanics are built on this? How are crits determined, if they exist at all? All of these questions carry more weight than simply the die choice.

I like the 3d20 take middle, that looks great!

>But that required doubling everything
Yeah. Because d20 is more granular. It's just like how if you compare a d20 system and a d100 system, you don't dismiss the d100 system because "a +5 bonus hardly does anything!" They're clearly operating on different scales.

So if your point is that you don't need the granularity of a d20 over 3d6, that's legitimate. But when people talk about how 3d6 is "less swingy" or "more consistent", they tend to conflate the normal distribution and the different effective spreads.

The standard deviation on a d20 is 1.95 times that of the standard deviation on 3d6, and you really have to take that into account. You can't directly compare a +1 bonus on 3d6 to a +1 bonus on a d20 just like you can't directly compare a +1 bonus on a d20 to a +1 bonus on a d100.

So these what & are doing is trying to remove the scaling issue from the equation and just leaving us with a comparison between the flat distribution of a d20 and the bell-curve (normal) distribution of 3d6.

But but muh bell curve!

Given a range of 1-20 vs 3-18, doubling everything to try and say they're similar seems much less solid than multiplying by 5 to get from 1-20 to 5-100.

See

>looks

Now create an actual system with all the bells and whistles and sit down and calculate the probabilities you get out of that before asking yourself if those are the probabilities you want in each given scenario. Because that's what you'll notice playing, not the curve of the die results.

Having a 1/108 chance of an edge result compared to a 1/10 chance is a difference. As is having a 1/4 chance of an average (+- .5) result compared to 1/10. This is what people are looking at when they say things like "less random", because most people cannot into statistics and as such don't know how to discuss them.