D&D 5E Average damage or rolled damage?

[Rushmik]

I split it fairly precisely;

Anything vs Player? We're rolling it. That arrow that hit you might graze by or land center mass, and the dice tell us which.
NPC vs NPC? Average damage it is. I don't want to waste my players time while I "play with myself." A single attack roll will suffice. If we're talking massed NPC - an army or something - I forgo the attack roll and only do the damage roll, but that's rather another topic. The point is, my players get the game's attention, and whenever I must focus on anything else I'm going to take the fast road so I can get back to the players.

 

[Lanefan]

I split it fairly precisely;

Anything vs Player? We're rolling it. That arrow that hit you might graze by or land center mass, and the dice tell us which.

Absolutely.

NPC vs NPC? Average damage it is.

Now here it's a bit more nuanced. If one of the NPCs is in fact a party member, hench, or anyone else the party might have to patch up afterwards, the damage gets rolled. If the party is watching two individual NPCs fight and intend to take out (or recruit) the winner such that I need to know not just who wins but what shape the winner is in afterwards, damage gets rolled*. But in most other cases...

If we're talking massed NPC - an army or something

...such as this...

I forgo the attack roll and only do the damage roll, but that's rather another topic.

I go the other way around, and every so often roll what amounts to a d20 attack roll for each side (which kinda builds in the general damage amount) to vaguely determine how well each is doing relative to average. A good example of this from recent play: a party was getting creamed by a demon until a wild magic surge summoned a dragon; dragon and demon turned on each other while (what was left of) the party crept away into the night, after which I just made some general d20 rolls to see who won between the demon and dragon (as it turned out, the dragon won but was too badly hurt to survive; so both died) as that would tell me what the party might find if they returned to the site later.

But, if I end up in a wargame situation where the PCs become part of an army fighting another army I've got rules for that; and if not I can always fall back on Battlesystem or similar.

* - if all the party is doing is watch (as opposed to trying to interact with the combat in any other way) I'll sometimes get a couple of players to volunteer to roll attacks and damage for the combatants; they just won't know what they have for hit points beyond what the party could see i.e. good shape, nicked, winded, hurting, bleeding, dying, etc.

Lanefan

 

[dmnqwk]

I think if the DM is either very lucky or unlucky for damage, it sometimes pays to simply take the average. You're also welcome to reduce the dice involved so the average has a shorter range.

The initial 13-68 (or 11d6+2 to simplify) is a great example - the average is sitting on 40.5 mean compared to the average, you're looking at 33% to 150% as a range
Reduced Range - 2d10+29 puts us down to 67% to 125%, a much easier range to cope with but still fairly random
Low Range - 2d6+33 pulls it down even further, so we're dealing with 87.5% to 112.5%

Looking at more normal rolls on numbers, let's look at something such as a Young White Dragon, whose breath weapon is normally 10d8.
Normal Range - 10d8 goes from 10 to 80, average of 45. This means we're going from 22% to 178%, quite a bit of pot luck going on there.
Reduced Range - We could go 2d12+32 and the average remains the same, but the range is now 34 to 56 (or 75% to 125% ish)
Low Range doesn't seem necessary since 75 to 125 is very reasonable. It makes the breath weapon dangerous every time, without allowing a bit of luck to decimate the party.

Obviously when you're dealing with normal low damage attacks such as a 3d10 necrotic touch, or just a 2d8+2 you could prevent them being capable of dealing 18 points, and work on the average 11, letting them do 1d6+8 (or 1d4+9) to prevent misfortunate screwing the PCs over. Basically I'd look to roll no more than 2 dice, and try to get the % between about 67% and 150% for a range where possible, to really make things a bit less random without removing the danger entirely.