Correction: that's 5 per short rest. It's 0-5 per combat because the use is deliberate. Players use these abilities when it matters, not when an encounter occurs just to spread them out.
There's no randomness in the use forcing a fighter to average the use out. That makes the uses always significant when they occur.
It's not only the first hit against an enemy that matters, though. Any time the enemy is within a potential average hit of dying, a character with a low chance of doing a very high amount of damage is not as effective as a character with a high chance of doing just enough damage, if the averages are close.
And that's ignoring the possibility that the higher damage lower chance to hit character already has the enemy dead at this point.
It amazes me that everyone that does that comparison never takes that scenario into account. It's always well the enemy is up with 1 hit remaining. NO. The enemy may very well already be dead by the higher damage lower chance to hit PC's previous hit.
I'm not talking about comparing two PCs with very different average DPR, I'm talking about comparing two PCs with similar DPRs, but different variances. The bigger the difference in variance, the bigger the difference in DPR has to be for the high-mean-high-variance character to be more effective.
Take an extreme example, much more pronounced than any difference you'd see in reality, but presented as a proof-of-concept. Character 1 only hits on a natural 20 but does 20d6 damage when they do hit. Character 2 always hits, but only does 1d6 damage on a hit. So both have 3.5 DPR but achieved in very different ways. Suppose each character is going one-on-one with an enemy with 70 HP. No real battle has an expected length of 20 rounds, but again, we're taking an extreme which is far, far away from the one-to-two-shot enemy scenario.
Character 1 has a chance to one-shot it, and is virtually guaranteed not to need more than two hits, but also has good chance they'll whiff round after round. Meanwhile, Character 2 has almost no chance to kill the enemy in fewer than 15 hits, but makes steady progress every round. Who would you bet on finishing the job faster?
Character 2 finishes first about 53% of the time; Character 1 about 45% of the time, with the other 2% being ties.
In a long battle like this, the matched DPR characters are pretty well matched for who finishes first, though the low variance character has a non-trivial edge. So who would you rather be?
If the odds are stacked against you such that you're likely to die "on average", then you'd take the Hail Mary and hope for that early crit. But most encounters are balanced such that the PCs are unlikely to die unless things go badly for them. So typically, guarding against tail risk is more important than trying to blow through encounters faster than expected. In that typical case, you stand to lose more by taking longer than average than you stand to gain by finishing sooner than average.
In this toy example, Character 1 has about a 1/4 chance of needing more than 40 rounds to finish the enemy, whereas Character 2 is nearly guaranteed to finish in 25 rounds or fewer.
So in terms of guarding against tail risk, you'd much rather be the low variance character.
That has nothing to do with overkill. It's not about doing more overkill, it's about less risky options being better in a game that already favors the PC's winning. That's a phenomenon I agree with. But it's not overkill.
Let's say the non-SS PC has a 60% chance to do 1d8+4, whereas the SS-PC has a 35% chance to do 1d8+14. So the regular archer has a DPR of 5.1, compared with the SS archer's DPR of 6.5. So the SS archer is about 25% better from an average DPR perspective. And indeed, against a foe with exactly 15 HP, such that on a hit the SS archer is guaranteed to kill them, they're very likely to do the job faster, and also less likely to take a dangerously long time.
But suppose the enemy has 18 HP, so the SS archer still has a better than 50% chance of one-shotting them if they hit, but no longer a guarantee.
Sure, Consistency has benefits. 2 important considerations:
1. Other PC's will be attacking your enemy as well. So the fight will really never hit that dangerously long territory because of that.
2. Extra attacks greatly reduce the overall variance. Considering a fighter can have SS + CE by level 4 and be making 3 attacks with it by level 5, I think that mitigates the issue enough to not worry about.
The SS archer has a lower expected number of turns needed to finish the enemy (3.9 vs 4.3), and a much better chance of doing it within two rounds (41% vs 16%), but they also have nearly double the chance of taking a "dangerously long time" (say 8 rounds or more; about twice the average), at 12% vs 6.5%. Even if your threshold for "dangerously long" is more liberal, at 1.5 times the average, or 6 rounds, the non-SS character has a slightly better chance of staying clear of that danger zone.
So if you're optimizing for the odds of impressive breezy battles, yes, the extra 1.4 DPR is worth the added variance. But if you're hedging against catastrophe, you'd rather not use the -5/+10.
I agree with the general premise, but I think that in play there are more mitigating factors than you've provided. That said, none of this still has anything to do with overkill. We are talking damage distribution and variance and kill rates. Not wasting damage on variance.
We are looking at a class comparison based on damage. Action surge's most common use by far is a damage spike.
We cannot ignore a class spike damage ability in a damage comparison. And just like superiority dice, the effect is always relevant because it's deliberate in it's use instead of controlked by random probability.
And that's ignoring the possibility that the higher damage lower chance to hit character already has the enemy dead at this point.
It amazes me that everyone that does that comparison never takes that scenario into account. It's always well the enemy is up with 1 hit remaining. NO. The enemy may very well already be dead by the higher damage lower chance to hit PC's previous hit.
They might, and in scenarios where the high-variance character's damage is just enough to providing good odds of a one-shot kill, that works in their favor (hence the example about 15 HP being a really good number for the 1d8+14 character). But falling just short of a kill happens too. That's why we should look at probability distributions instead of just isolated scenarios.
That has nothing to do with overkill. It's not about doing more overkill, it's about less risky options being better in a game that already favors the PC's winning. That's a phenomenon I agree with. But it's not overkill.
It's not only about overkill, but it's partly about overkill. It's also about "underkill". Any time you fall within an average amount of damage short of a kill such that it takes you another hit to finish the job, the reason the reliable medium damage character has the edge is that their results consist of less overkill and less underkill.
1. Other PC's will be attacking your enemy as well. So the fight will really never hit that dangerously long territory because of that.
It sounds like you're saying PCs never fail to finish off monsters before monsters finish them off. I could also argue that the fact that the other PCs are there is reason to make sure you do your part, so the party as a whole can finish off the monster before their turn comes. Which means that the "one shot kill" edge case that you dismissed should arguably be about the "four shot kill" case.
2. Extra attacks greatly reduce the overall variance. Considering a fighter can have SS + CE by level 4 and be making 3 attacks with it by level 5, I think that mitigates the issue enough to not worry about.
I'm not sure exactly what distinction you're making. Overkill is part of the right tail that balances out the left-tail misses when looking at average damage.
So the way I would "tweak" the fighter in a no-feats game is replace the extra ASI at level 6 with "choose one feat". This would be the only way to gain a feat.
