This is not particularly compelling.
In short, you're making the argument that "Good things happen when you roll well." If you're going to factor in the benefits of rolling well (a fair chance of taking out a weak creature), you also need to factor in the drawbacks of rolling poorly (a fair chance of doing nothing at all). Otherwise, your analysis is incomplete.
This is generally why average results are used.
But it
is relevant, particularly at lower levels (which is where most cantrip use is found). Let's say my party is fighting some goblins, with 7 hp each. I have cantrip-
magic missile and
firebolt available. The party fighter uses a longsword + shield and has the duelist fighting style. The rogue is using a rapier.
If the fighter hits a goblin, the fighter deals 1d8+5 damage (Str 16 + duelist). The fighter kills the goblin unless he or she rolls a 1 - a 1 in 8 chance.
If the rogue hits a goblin and gets sneak attack in (and he or she should), that's 1d8+1d6+3 damage. The rogue kills the goblin unless he or she rolls 2-3 on 1d8+1d6 - a 1 in 16 chance.
If I cast cantrip-
magic missile at a goblin, I will deal 1d4+1 damage to it. That won't kill it - there is
no chance of that happening. The next round, that goblin will still be around to make an attack - unless one of my comrades attacks it as well, and in that case the goblin would almost certainly have died anyway. That means that my action was almost certainly wasted (unless I fire one one of the goblins in the back that my buddies aren't hitting).
But if I cast
firebolt, I at least have a chance of dealing 7 points of damage and taking the goblin out.
Now, all of this is an artifact of how damage and hit points scale at lower levels. If we're fighting a 27 hp bugbear instead, none of us is going to take that thing down in a single shot, so then averages become more relevant.