JackOfAllTirades
Explorer
OK, I'm out. There's no point debating with someone who plays a game with dice and insists that probability distributions are "meaningless."
I wouldn't say its trivial to compute the optimal time to use Savage Attacker in multiattack situations. However, it is doable. But algorithms aside that's not the point I was making.
My post was in relation to JackofTirades where he suggested that "the whole story" would include looking at the damage distributions. Right now about the only thing knowing the damage distributions allows is to do is to calculate DPR, which is rather handy because DPR is strongly correlated with being "better" at combat.
So basically the point I am trying to make is that when we talk about DPR with savage attacker we are talking about the only meaningful stat that knowing the damage distribution currently let's us compute. So it's almost wrong to talk about the whole picture and bring in damage distributions when they are currently meaningless in relation to combat effectiveness outside their DPR.
Oh there is one other thing knowing a damage distribution can do for us at the moment, knowing our minimum and maximum damages which has very little to do with being "better" at combat. So basically trivially obvious and meaningless information that seems to get pointed out by way to many people because those people apparently think everyone else is too dumb to realize that a reroll that causes +1 DPR doesn't actually have change your minimum or maximum damage values.
The reminder given would actually be a useful reminder if there actually existed something besides DPR that allowed comparing different damage distribtuions in such a way that the comparison was a useful predictor of combat effectiveness.
OK, I'm out. There's no point debating with someone who plays a game with dice and insists that probability distributions are "meaningless."
Actually it wouldn't. The more dice you role, the less the feat does for you. You have to re-role all your dice with this feat, you can't cherry pick and re-role only some.I understand it not working with all damage dice (it would be an auto take for Rogue for Sneak Attack),
If I understand you correctly, you're pointing out that the "whole story" is way more than DPR. While it's pretty easy[1] to calculate the DPR boost for Savage Attacks, what is not easy to do is calculate the value of the opportunity cost (other feats, which are much harder to analyze), especially since many of them affect things other than simple DPR.
Am I understanding you correctly? If so I agree. DPR is an interesting metric in some comparisons but it's far from the whole story.
[1] Okay, you're right, it's not quite trivial to compute the optimal Savage Attacks algorithm. It would take me a computer and fifteen or twenty minutes to write a calculator for giving the optimal algorithm vs. AC X with Y number of attacks, and I might make mistakes while writing the code... so it's not trivial in the same way that computing average damage boost using Sharpshooter is.
You are getting close and all you said is true even if it wasn't exactly what I was trying to say. Yes, the whole story is more than DPR. However, damage distributions really don't help tell us any meaningful part of the story at the moment that DPR values do not.
Oh, I get it! You're saying that DPR is a useful summary statistic[1]. Yes, it is useful. It's not the best or only statistic out there (loss ratios are better IMO), but it has the virtue of simplicity.
-Max
[1] I'm unsure whether you're referring to DPR as an actual Damage-per-round metric against a particular AC under particular circumstances, or abstract DPR in a "assume all attacks hit" scenario, but I think both metrics can be useful.
Actually it wouldn't. The more dice you role, the less the feat does for you. You have to re-role all your dice with this feat, you can't cherry pick and re-role only some.
The more dice you role, the more average your total result, the less chance of the re-role being much different than your first roll.
I'm not sure what loss ratio is?