Dice change

[Alexvrahr]

In my group trying out D&D 5e (plenty of experience with other RPGs, just not this one) the disturbing unpredictability led to the GM introducing a house rule that 2d10 replaces the 1d20, with advantage meaning 3d10 keeping the 2 highest, disadvantage meaning 3d10 keeping the 2 lowest. Natural 18-20 becomes the new natural 20. Is there any consequence down the line of doing this that you can think of?

[Composer99]

(1) Medium and higher results are more likely overall, with results being weighted towards the middle. For instance, a result of 9 or better on a single d20 occurs 60% of the time, while the same result set on 2d10 occurs 72% of the time. The "middling" results of 9 through 13 occur 44% of the time with 2d10 as compared to 25% of the time when rolling 1d20.

(2) Very low results are much less likely. (A roll of 1 through 3 happens 15% of the time on a single d20, compared to a roll of 2 through 4 happening 6% of the time with 2d10.) However, very high results are correspondingly less likely as well.

(3) With that same thought in mind, on average attack rolls, saving throws, and ability checks are more likely to succeed as long as you only need a middling roll. They're less likely to succeed if you need a high roll, perhaps very much less likely.

I'm not as familiar with how advantage or disadvantage would work. Getting both good die roll modifiers and advantage would be important if you need to roll very high on the dice to succeed. I definitely recommend working out what advantage or disadvantage bring to the table.

Edit to add: Also worth thinking about are any implications for abilities that set die roll values, such as Reliable Talent or Portent. They might not be quite as good (and Portent will on average be less exciting).

[Grod_The_Giant]

I've never tried it in 5e, but I've made the substition in another d20 system, Mutants and Masterminds. 2d10 makes for a nice balance between bell curved and linear results.

[Yakk]

The standard deviation of 1d20 is sqrt(399/12), the mean is 10.5. The standard deviation of 2d20 is sqrt( 2 * 99/12 ), or about 30% lower. Or, in other words, the SD of 1d20 is about 50% higher (well, 41%) than the SD of 2d10.

90% of the effect of this rule change is to boost the size of modifiers, and DC-distance-from-10, by about 50%.

So this is similar to making Plate be baseline 21 AC, and shields be +3 AC. Attributes, instead of scaling from 8-20, scale from 8-24. Your proficiency bonus starts at +3 and scales to +8. Magic item bonuses go from +1 to +4 instead of +1 to +3. The Shield spell grants a +7 to AC. Etc.

It also requires a bit more math at play-time, as you have to add up 2 numbers.

---

This massively reduces the bounded accuracy system of 5e. Lower level monsters will be far more trivial than 5e assumes, and higher level monsters far harder.

Stacking bonuses will become more important; a Paladin gets more important, as that cha bonus to saves is going to be needed more.

SS/GWM gets weaker, as it is akin to a -7 for +10 instead of -5 for +10.

Advantage gets weaker, both because modifiers will render it moot more often, and because 3d10 pick best 2 isn't as much better than 2d10 as 2d20 pick best is better than 1d20. Disadvantage will matter less.

---

Is the unpredictability out of combat or within it?

If out of combat, the GM can be more generous with situational advantage for attribute checks. Many out of combat attribute checks should be performed with advantage (with someone helping the more skilled person), which shrinks the standard deviation a lot.

---

Note that talk of a "bell curve" is a bit misleading. There is a bell curve, but if you calculate the cumulative distribution function of 2d10 vs 1d20, and correct for mean (average) and standard deviation (horizontal scaling of DCs/modifiers), the two curves end up being very close to each other.

The bell curve is most visible when viewing *roll exactly* values; but *roll exactly* values are the first derivative of the value you care about (can you roll X or higher), and derivatives can make minor differences look large.

[Zman]

It is great for skills, but really boinks the math for saves, to hit, and AC to the point it is extremely problematic. I've ran it pretty thoroughly in modified 5e as the skills variant and found it far superior to the standard d20 roll in all but one respect, it is an added layer of complexity.

[Kane0]

Its a significant but not catastrophic change. Give it a go over a few sessions and you’ll soon figure out if it works for your table or not.