Pretty straightforward question; a lot of people complain about the swingyness of the D20, so what if you replaced it with the sum of D10s (counting zeroes as 10). You could still have critical successes (double zeroes) and critical fails (double ones). Has this already been tried in some game I’ve never heard of? Will the mere suggestion of this call down the wrath of Gygax from beyond the grave?
Tell what you guys think, or how I’m a dumb idiot that doesn’t understand bell curves.
1) Don't confuse x for f(x);
2) And as for What Would Gary Gygax Do, be mindful of calibration.
3) There are better ways to do a bell curve.
1) Don't confuse x for f(x) refers to the fact that the number generated on the die does not exist in-game. Further, even out of game, the number has no units attached. It's a completely meaningless abstraction. The rules of the game is what gives it meaning by creating a function: roll this number or higher and you hit, otherwise you miss. That's what f(x). It's the only thing that's meaningful. Complaining that the d20 as used in D&D "swingy" is an example of confusing x for f(x).
2) I can't prove what was in Gary's head, but enough things line up just right that lead me to believe the train of thought was something along these lines.
The base character in 1E is the 0th level human which needs a 20 to hit AC 1, for which there is a 1 in 20 chance. A 19 is needed to hit AC 2, for a 2 in 20 chance. An 18 to hit AC 3 for a 3 in 20 chance. You see the pattern? The game is calibrated so that descending AC that nobody seems to understands makes perfect sense if you understand this is your chance in 20 of being hit by an average human. The upshot of which, if you're running a mass melee, particularly on a 20:1 scale, it's really easy to determine number of hits landed. It's just equal to the defending units AC.
Now as long as you're adopting "modern" concepts like "ascending AC" and a "bell curve mechanic" you're going to lose this functionality with no hope of ever getting it back. And since most people don't do mass combat or platoons of 0th level mercenary hirelings anymore, that part is not really a big loss. Fans of "modern mechanics" thus view these updates as objectively superior when it's really a subjective opinion and the only thing objective about it is the loss of functionality.
But the rest of Gygaxian calibration can be useful.
As for why a given AC should have a given hit probability, I think the calibration goes something like this. The worst base chance to hit in the game is the 0th level human versus the worst base AC is 10 (unarmored) is calibrated at 50/50 chance. We have no particular reason to believe attacker or defender has any advantage in this base case. The base armor in the game is chainmail (like the combat system's predecessor's namesake), and so that is calibrated to bisect the base hit probability between no protection (50%) and total protection (0%), so it's set to 25%, or AC 5 under the d20 system. It wouldn't be fun if the best (non-magical) armor in the game made you impervious. So it's calibrated to protect you from 90% of all attacks, or AC 2 for Platemail & Shield.
So here's what you'd have to do for 2d10. The baseline character (however you're defining that: THAC1 = 20, THAC0 = 20, or BAB = 0, etc) should have a 10% chance of hitting the best armor offered up on your equipment list. In the 1E PHB that was platemail & shield. But full plate and field plate appeared in the DMG and later in UA. Whatever the exact armor is, it will take a 17 or better to hit by a zeroling, since that's 10% likely on 2d10. For no armor, the probability should be around 50%. One of the things I hate about 2d10 specifically is it has no 50/50 point. Fucking even numbers of dice. So you're just going to have to make a call as to whether you want it to be 11 or 12.
I'd probably lean towards 11, because you might be starting to notice, ACs are going to fall on a tighter range under this system than the original. And so then you want to bisect the probability. If you chose 11, that's 55%, or 27.5% bisected, if you chose 12, that's 45% or 22.5% bisected. Meaning to hit the midling armor (chainmail or whatever) it's going to mean needing a 14 or 15 respectively. 3 places better than no armor.
So if you're doing modern d20, you'd recalibrate the ACs as follows:
No armor = AC 11
Chainmail = AC 14
Plate & Shield = AC 17
If you're doing AD&D 1E (with THAC1 = 20), then your ACs will be
No armor = AC 10
Chainmail = AC 7
Plate & Shield = AC 4
In either case, you'll have to rank the remaining armors and assign them to the in-between numbers.
3) There are better ways to do a bell curve. You can get a bell-curve mechanic while still retaining the 1 to 20 range by rolling 3d20 and dropping the highest and lowest. You can always pull this trick whenever the game provides you a linear mechanic. Which means if designers are smart they'd limit themselves to linear mechanics with the full knowledge that GMs can flip this switch according to their own tastes--there's no reason for the designer to force their tastes into the equation.