Does Xanathar saves Savage Attacker?

[Waazraath]

Or maybe better: does the spell Shadow Blade?

For context: Savage Attacker is generally considered to be one of the weaker feats (reroll damage for a melee weapon attack 1 / turn and pick the highest). Normall, rerolling 1d8, or even d12 once per turn isn't going to be worth a feat. The sole exeption were some beast forms that a Moon Druid can change in, that make 1 big attack/turn.

But Shadow Blade does 2d8, when upcast up to 5d8. Rerolling 4d8 or 5d8 every turn and picking the highest might be worth it? What do you folks think?

[strangebloke]

No.

Savage Attacker actually gets less good as you add more die. Basically it lets you choose the best of 2 rolls, and if you have a lot of dice in each roll they're more likely to be similar to each other. It's best with a small number of large dice. For example, on a single d12, Savage attack adds ~2 damage. That'd actually be worth a feat except that you only get to use it once a turn.

For 2d8 the damage increase is ~1.8 on the first attack. Pretty meh.

If you upcast to make it 4d8, the bonus is ~2.6 on the first attack. That's more in an absolute sense, but less in a proportional sense. The proportional sense is what you're trading against here because a simple stat boost would gives you a +1 to attack, which might represent anywhere between +10% increase to a +30% increase in your damage on each attack! That's in addition to just giving you a straight +1 to damage!

so, assuming two characters, one with SA and one with +2 to dex/str/cha...

Savage Attacker character: 2*[2d8+3] * 40% + 1.8 = 11.4
Boring 'ol ASI character: 2*[2d8+4] * 45% = 11.7

And that +2 dex/str/cha is doing more than just giving damage. If you have more than 2 attacks, the comparison gets even worse. There is a niche for Savage attacker, which is that you might put it on a very high-level barbarian or fighter on some circumstances. But that is a niche.

[Man_Over_Game]

No.

Savage Attacker actually gets less good as you add more die. Basically it lets you choose the best of 2 rolls, and if you have a lot of dice in each roll they're more likely to be similar to each other. It's best with a small number of large dice. For example, on a single d12, Savage attack adds ~2 damage. That'd actually be worth a feat except that you only get to use it once a turn.

For 2d8 the damage increase is ~1.8 on the first attack. Pretty meh.

If you upcast to make it 4d8, the bonus is ~2.6 on the first attack. That's more in an absolute sense, but less in a proportional sense. The proportional sense is what you're trading against here because a simple stat boost would gives you a +1 to attack, which might represent anywhere between +10% increase to a +30% increase in your damage on each attack! That's in addition to just giving you a straight +1 to damage!

so, assuming two characters, one with SA and one with +2 to dex/str/cha...

Savage Attacker character: 2*[2d8+3] * 40% + 1.8 = 11.4
Boring 'ol ASI character: 2*[2d8+4] * 45% = 11.7

And that +2 dex/str/cha is doing more than just giving damage. If you have more than 2 attacks, the comparison gets even worse. There is a niche for Savage attacker, which is that you might put it on a very high-level barbarian or fighter on some circumstances. But that is a niche.

Good numbers, Bloke!

There was a good thread a while back about fixing Savage Attacker. Although I think the easiest fix is just to make it a half feat (increasing Strength or Dexterity).

[PeteNutButter]

No.

Savage Attacker actually gets less good as you add more die. Basically it lets you choose the best of 2 rolls, and if you have a lot of dice in each roll they're more likely to be similar to each other. It's best with a small number of large dice. For example, on a single d12, Savage attack adds ~2 damage. That'd actually be worth a feat except that you only get to use it once a turn.

For 2d8 the damage increase is ~1.8 on the first attack. Pretty meh.

If you upcast to make it 4d8, the bonus is ~2.6 on the first attack. That's more in an absolute sense, but less in a proportional sense. The proportional sense is what you're trading against here because a simple stat boost would gives you a +1 to attack, which might represent anywhere between +10% increase to a +30% increase in your damage on each attack! That's in addition to just giving you a straight +1 to damage!

so, assuming two characters, one with SA and one with +2 to dex/str/cha...

Savage Attacker character: 2*[2d8+3] * 40% + 1.8 = 11.4
Boring 'ol ASI character: 2*[2d8+4] * 45% = 11.7

And that +2 dex/str/cha is doing more than just giving damage. If you have more than 2 attacks, the comparison gets even worse. There is a niche for Savage attacker, which is that you might put it on a very high-level barbarian or fighter on some circumstances. But that is a niche.

I was trying to crunch the numbers on this for my current character, and I think we are missing out on the choice factor. If you have 3 or more attacks, if you wait until there is a larger discrepancy you can get a bigger pay off.

For reference my character is a Gloomstalker 5/champion 3/Sorcerer 3. When I do my big action surge on round one that gives me 6 attacks
with my 3rd level shadow blade (3d8). I was probably going to take sorcerer next level and pick up an ASI. Since my strength is already capped (hill giant belt), I was eyeballing savage attacker, even though its usually terrible, as a way to increase my damage.

Rather than do math that seemed a bit complicated, I simulated the 6 attack action with 3d8 and savage attacker. After 50 iterations of the 6 attacks I found that on average, taking the better of 3d8 on every attack gave me roughly 4.5 more damage. If instead I had perfect knowledge of future rolls, taking the better attack only at the optimal time, I'd get approximately 9 more damage. Naturally, you don't have perfect knowledge, but you could reasonably wait until you roll low on damage before you use savage attacker. It's that choice that makes it somewhat useful. There are too many variables, to make it simple, (such as hit chance, crit chance), but in practice I could potentially see the upside even higher if I use it on crits (champion with frequent advantage from SB).

TL;DR While you are right, that more dice push the odds of savage attacker towards not helping as much, the choice factor if you have several attacks allows you to potential make more use out of it as you can wait for a low roll and/or crit. Granted there is always a chance you miss entirely with your later attacks, making you get zero benefit from it.

Honestly though, I'll probably end up taking something more practical like resilient wisdom or some such.

[Snails]

It's best with a small number of large dice. For example, on a single d12, Savage attack adds ~2 damage. That'd actually be worth a feat except that you only get to use it once a turn.

Your points are all correct, of course.

I would suggest that this feat is okayish assuming (1) you are rolling a big die (e.g. d12), (2) your attack stat is already 20 (as per your math), (3) you do not get a large number of attacks anyway (e.g. barbarian).

This feat protects you from annoyingly bad damage rolls, and that matters to some players. You do not reroll that 5 on a d12, unless you really believe a high roll might drop the target. You especially do not bother to reroll that 5 if you have a second attack coming.

Is this an optimal use of an ASI? Surely not. But, as I said, it does matter to some players to not have to live with a 2 rolled on a d12. I think there is room in this game to support that kind of player. It falls short of being attractive to an Optimizer, but it is not so weak to be a trap feat or wasted feat.

[Man_Over_Game]

Your points are all correct, of course.

I would suggest that this feat is okayish assuming (1) you are rolling a big die (e.g. d12), (2) your attack stat is already 20 (as per your math), (3) you do not get a large number of attacks anyway (e.g. barbarian).

This feat protects you from annoyingly bad damage rolls, and that matters to some players. You do not reroll that 5 on a d12, unless you really believe a high roll might drop the target. You especially do not bother to reroll that 5 if you have a second attack coming.

Is this an optimal use of an ASI? Surely not. But, as I said, it does matter to some players to not have to live with a 2 rolled on a d12. I think there is room in this game to support that kind of player. It falls short of being attractive to an Optimizer, but it is not so weak to be a trap feat or wasted feat.

The problem comes in that every class that uses weapons either:

Can use large die weapons (1d10 or 1d12), but uses Extra Attack (when +1 to hit and damage on x2 attacks is so much better)
Can use large die weapons and doesn't have Extra Attack, but often requires more stats or attacks less often (Paladin, Cleric)
Uses smaller die weapons and won't get much damage gain (Rogue, Monk).

The niche example where I'd see it get the most use is from a War Cleric that doesn't cast spells and has relatively low Wisdom, or maybe a Barbarian.

[Snails]

The niche example where I'd see it get the most use is from a War Cleric that doesn't cast spells and has relatively low Wisdom, or maybe a Barbarian.

I think your assessment is on the money.

I see it as more about the player personality so happening to be someone who hates to roll a 2 on a d10/d12, than it is about it being a solid positive boost like a lot of feats really are. It is a so-so means of mitigating bad luck. Boosting a saving throw or getting more HP or the Luck feat is probably a better means of mitigating bad luck, but I would not call this feat so terrible as to be a trap.

[Man_Over_Game]

I think your assessment is on the money.

I see it as more about the player personality so happening to be someone who hates to roll a 2 on a d10/d12, than it is about it being a solid positive boost like a lot of feats really are. It is a so-so means of mitigating bad luck. Boosting a saving throw or getting more HP or the Luck feat is probably a better means of mitigating bad luck, but I would not call this feat so terrible as to be a trap.

I disagree on this. A feat generally is designed to open up what a player is able to do. It gives them options that they didn't have before, or it makes them powerful in a very specific way (such as the Resilient feats).

However, Savage Attacker neither is beneficial in the hard numbers, and it's not beneficial in the way that it opens up more opportunities for the player to interact with the game. When you grab it, you're grabbing it under the assumption that it's increasing a specific combat stat (like Resilient), and it's not.

Nobody really has a major problem with Linguist just because it's not very combat effective. Duh, you're getting exactly what you're getting, and only people know think that's valuable will ever take it. However, Savage Attacker APPEARS to be better than what it actually is. In reality, the default choice (of not having a feat) is strictly better, which says a lot when you consider that the ASI actually opens up more opportunities (considering it increases saves, DCs, and skills).

I think Savage Attacker is the worst feat, if only because it's usually grabbed because people aren't familiar with statistics. It's success is reliant on player ignorance.

[strangebloke]

Good numbers, Bloke!

There was a good thread a while back about fixing Savage Attacker. Although I think the easiest fix is just to make it a half feat (increasing Strength or Dexterity).
Yeah, that would make it absolutely work taking, but still not overpowered to the point of being compulsory.

I was trying to crunch the numbers on this for my current character, and I think we are missing out on the choice factor. If you have 3 or more attacks, if you wait until there is a larger discrepancy you can get a bigger pay off.

For reference my character is a Gloomstalker 5/champion 3/Sorcerer 3. When I do my big action surge on round one that gives me 6 attacks
with my 3rd level shadow blade (3d8). I was probably going to take sorcerer next level and pick up an ASI. Since my strength is already capped (hill giant belt), I was eyeballing savage attacker, even though its usually terrible, as a way to increase my damage.

Rather than do math that seemed a bit complicated, I simulated the 6 attack action with 3d8 and savage attacker. After 50 iterations of the 6 attacks I found that on average, taking the better of 3d8 on every attack gave me roughly 4.5 more damage. If instead I had perfect knowledge of future rolls, taking the better attack only at the optimal time, I'd get approximately 9 more damage. Naturally, you don't have perfect knowledge, but you could reasonably wait until you roll low on damage before you use savage attacker. It's that choice that makes it somewhat useful. There are too many variables, to make it simple, (such as hit chance, crit chance), but in practice I could potentially see the upside even higher if I use it on crits (champion with frequent advantage from SB).

TL;DR While you are right, that more dice push the odds of savage attacker towards not helping as much, the choice factor if you have several attacks allows you to potential make more use out of it as you can wait for a low roll and/or crit. Granted there is always a chance you miss entirely with your later attacks, making you get zero benefit from it.

Honestly though, I'll probably end up taking something more practical like resilient wisdom or some such.
Right right.

The choice is important and isn't reflected in my analysis. Well, one part is. The ability to choose makes it comparable to other /turn damage sources in that you're much more likely to be able to use it. +2 damage to one attack that hits is better than +1 damage to two attacks that may or may not hit.

The other factor you're talking about here is that you can choose to use the reroll only on bad damage rolls. Sometimes you'll use it and turn a 1 into a twelve. That feels great!

I went back to my anydice code. You're pretty much correct. As the number of hits increases you get more chances to roll terribly, and more chances to make Savage Attacker look good.

So, to show my work, let's look at a simple 1d12:
1 attack: 2.0
2 attack: 3.0
3 attack: 3.6
4 attack: 4.0
5 attack: 4.2
6 attack: 4.5
7 attack: 4.6
8 attack: 4.7
9 attack: 4.8
10 attack: 4.9

I'm not confident that my math is right here. I think I have an off by one error. But the general trend holds.

Personally i think that this jives with my earlier statement. Don't take the feat unless you've already maxxed strength and taken PAM and GWM.

But I guess I'm now thinking that this would make for a good feat for a high level champion.


I think Savage Attacker is the worst feat, if only because it's usually grabbed because people aren't familiar with statistics. It's success is reliant on player ignorance.

Grappler is worse, as it actively makes you worse at the thing you're specializing in, which isn't worth specializing in to begin with.

This will still feel good sometimes, when you get to reroll your crit that came up with double 1s.

[Mith]

A thought on Savage Attacker: Have it raise the floor on dice rolled? Something like "Treat the minimum damage rolled of X dice as Y."

I know that's vague, but if X = half ability mod round up (likely max = 4 for Barb +7) and Y as what ever seems as a reasonable floor, this means that for a Barb, they have a reasonable floor on their brutal critical, Rogues and Paladins get a boost to SA and Smites. If you allow one to modify X dice per turn, you can use it to improve damage out over a turn if you are not getting extra damage dice.

Probably too fiddly, but I am always a bigger fan of raising the floor rather than dice re rolls. I know too many people who say they suck at rolling dice.

[stoutstien]

A thought on Savage Attacker: Have it raise the floor on dice rolled? Something like "Treat the minimum damage rolled of X dice as Y."

I know that's vague, but if X = half ability mod round up (likely max = 4 for Barb +7) and Y as what ever seems as a reasonable floor, this means that for a Barb, they have a reasonable floor on their brutal critical, Rogues and Paladins get a boost to SA and Smites. If you allow one to modify X dice per turn, you can use it to improve damage out over a turn if you are not getting extra damage dice.

Probably too fiddly, but I am always a bigger fan of raising the floor rather than dice re rolls. I know too many people who say they suck at rolling dice.

I tried something similar for great weapon fighting style. Treating all 1s as 2s. Ok mathematically but it is a little wonky with wording.

Savage attacker I just made it so they can maximize one die roll per round. Still probably be a half feat

[Mith]

I tried something similar for great weapon fighting style. Treating all 1s as 2s. Ok mathematically but it is a little wonky with wording.

Savage attacker I just made it so they can maximize one die roll per round. Still probably be a half feat

Yeah, I'll try that next time I get around to a game. Although, I do wonder if it is better to say floor = 3 since the original feat has you reroll 1s and 2s.

Granted that does mean that Great -Sword always hits for at least average -1 damage, which is a bit much.

[Waazraath]

Very interesting, thnx for the replies & the math!

[Segev]

The interesting question with Savage Attacker is what your threshold for using it on not-your-last-attack in a round is. Obviously, you use it on your last attack in a round if you haven't used it yet; there's no reason not to. But what's your threshold of "this attack is too low" before you use it on an earlier one?

Obviously, it's worth more the better your prediction/the lower the die you use it on, if only because the odds it will roll significantly higher are better. Do you use it if you're below half? Below a third? Below a quarter?

Comparing it to an ASI, which improves your damage on every roll by 1 (not counting the improvement to chance-to-hit, which isn't insignificant), you need it to do on average more than N more damage on the die roll it's used on, where N is the number of attacks you have in a round. This is because you need it to be doing enough extra damage on this one attack to make up for losing one effective damage from EVERY attack. And, given the increased accuracy imparted by an ASI, it actually is more than N extra damage you need to get on average on the die you choose.

So the problem is choosing a minimum acceptable roll, or a maximum roll over which you will not use it, on each attack, without knowing which will be the lowest damage attack in the round.

Given that the lower your threshold, the rarer it will be that you use it (except on the last roll, which is only the standard shift of any single attack where you can roll twice and take the better), but the more average boost you get when you do it, there's a trade-off to be measured.

[Bloodcloud]

My fix to savage attacker would be:
-Auto-apply on crit
-Can choose to re-roll a number of dice, instead of taking new result than reroll all keep best.

That way, it's always there for your crits and should have a use every turn or so.

[Phoenix042]

So real quick I'll just point a few things out:

Savage attacker applies only to weapon damage dice, but there are a few things that can increase the number of weapon damage dice you roll during an attack.

1) Orcish Fury (the feat from Xanathar's) lets you choose to add an extra weapon damage die once.
2) Scoring a crit doubles any weapon damage dice (including, possibly, your orcish fury die)
3) Being a half orc also grants you a bonus weapon damage die when you crit.
4) Barbarians get an extra weapon damage die on a crit at 9th, 13th, and 17th level.

It's not common or easy, but a half orc barbarian with Orcish Fury deciding what to pick up with his 19th or 16th level feat could decide to grab Savage Attacker, figuring a +2 to damage every round isn't GREAT, but he when he crits, he gets to roll 8d12 dice TWICE and pick his favorite roll.

And the difference then becomes about 6. That is, if he uses savage attacker every time he scores a crit, he'll gain about 6 average damage on his crits. That's a little less than he gains from the savage attacks feature from half orc, or each of the brutal critical dice themselves.

That's obviously not factoring in intelligent player choice.

[Chronos]

Segev, that analysis would also have to take into account how hard the enemy is to hit. Against a high AC enemy, if I make my first of three or four attacks and hit, that might still be my only chance to use Savage Attacker that round, because the others might miss.

You might also have different damage dice on your different attacks, as, for instance, from Polearm Master, two-weapon fighting, or (at low levels) the monk's Martial Arts.

[PeteNutButter]

The interesting question with Savage Attacker is what your threshold for using it on not-your-last-attack in a round is. Obviously, you use it on your last attack in a round if you haven't used it yet; there's no reason not to. But what's your threshold of "this attack is too low" before you use it on an earlier one?

Obviously, it's worth more the better your prediction/the lower the die you use it on, if only because the odds it will roll significantly higher are better. Do you use it if you're below half? Below a third? Below a quarter?

Comparing it to an ASI, which improves your damage on every roll by 1 (not counting the improvement to chance-to-hit, which isn't insignificant), you need it to do on average more than N more damage on the die roll it's used on, where N is the number of attacks you have in a round. This is because you need it to be doing enough extra damage on this one attack to make up for losing one effective damage from EVERY attack. And, given the increased accuracy imparted by an ASI, it actually is more than N extra damage you need to get on average on the die you choose.

So the problem is choosing a minimum acceptable roll, or a maximum roll over which you will not use it, on each attack, without knowing which will be the lowest damage attack in the round.

Given that the lower your threshold, the rarer it will be that you use it (except on the last roll, which is only the standard shift of any single attack where you can roll twice and take the better), but the more average boost you get when you do it, there's a trade-off to be measured.

That's exactly what I was getting at. Theoretically, there would be a mathematical solution for optimal use of the feat, but when statistics get this complicated, I usually resort to simulation instead of equations.

Given my 3d8, 6 attack, high hit chance scenario, I suspect it'd something like:
1st Attack & 2nd Attack: Use savage attacker if the 3d8 yields 7 or less (6.5 below average).
3rd Attack: Use SA if 3d8 yields 8 or less
4th Attack: Use SA if 3d8 yields 9 or less
5th Attack: Use SA if 3d8 yields 11 or less
6th Attack: Use SA if you haven't yet.

But the problem is the math on all these change depending on the hit chance against any given monster, which can often change when going from monster to monster. Frankly the math is too complex to know exacts for each scenario. (Though if someone can come up with an equation, I'd be impressed.) Ultimately, if I do take the feat, I'll just have to go with my gut. I roll like crap often enough, I'm sure I'll get some use out of it.

[Theodoxus]

I've only ever seen SA used on a level 1 Vhuman fighter, in a game that was a one shot, so extra attack was never a consideration. Using SA on every roll was actually kind of a pain (player was using a great sword, and had GWF, so 1s and 2s were rerolled first, and then SA was rolled, and I don't think it's RAW, but the DM allowed any 1 and 2 rolled on the SA be rerolled as well. So, a few times, it was upwards of 8 rolls, and a lot of finger counting (sigh) on. each. attack.

Does slowing down combat count as detrimental?

[Segev]

An interestingly metagame way of running it with perfect knowledge would be to resolve all your attacks, keeping track of each of their damages, and then choosing one to re-roll.

This works more-or-less fine as long as it's all just attacks on your turn; the only difference it might make is if you were switching targets if and when you dropped something, and the choice of which attack to reroll damage on changes when you dropped it. If you include reaction-based attacks, however, this gets problematic because things can happen between them and the last attack of your turn.

[LudicSavant]

I am of the opinion that Savage Attacker is not worth saving, because it offers nothing of worth in terms of flavor or gameplay experience. Best case scenario you adjust the numbers so that it's a bland weapon damage boost roughly on par with other bland weapon damage boosts. Not only that, but it slows down gameplay a bit too (because the player has to stop after each weapon damage roll and go "hmmm, do I want to reroll this one, or might I roll worse on my next attack?" Even if it only takes a fraction of a second that sort of thing adds up to slow down play for no good reason).

[Man_Over_Game]

I am of the opinion that Savage Attacker is not worth saving, because it offers nothing of worth in terms of flavor or gameplay experience. Best case scenario you adjust the numbers so that it's a bland weapon damage boost roughly on par with other bland weapon damage boosts. Not only that, but it slows down gameplay a bit too (because the player has to stop after each weapon damage roll and go "hmmm, do I want to reroll this one, or might I roll worse on my next attack?" Even if it only takes a fraction of a second that sort of thing adds up to slow down play for no good reason).

I think the name, and what it implies, is worth saving. The specifics of the feat...not so much.

Maybe something like: When you kill a creature with an attack, hostile creatures adjacent to you or the target must succeed on a Wisdom Saving Throw or be Frightened of you until the start of your next turn. The DC for this Saving Throw is equal to 8 + your Proficiency.