thelee Posted October 18, 2018 Share Posted October 18, 2018 (edited) UPDATE NOTE Some people were getting extremely confused/hung up by some of the technical details, so I've done a lot of rewording and hidden some of the details inside spoiler tags; any time you see spoiler tags, you can expand it to read some nitty technical details--if you're not interested you can just skip it. This is just a note because if you're reading through the immediate replies it might seem like people are quoting random things that don't exist, and this is why. BACKSTORY I was planning on writing up another big combo character build and mechanics guide like I did with my Umezawa write-up, and while doing that I found it difficult to reason about certain interactions purely with simple pen-and-paper math. I ended up writing a script that I could plug in virtually any combination of stats, modifiers, etc. and let it run hundreds of simulations against varying Armor/Deflection situations and do experimental comparisons. I came up with some very interesting outcomes, and rather than wait to put it inside a character build, I'm sharing my findings here. KEY PHRASE - added in update "net damage" <= essentially, the multiplicative increase in sustained damage compared to alternative situations. KEY ASSUMPTIONS/DISCLAIMERS - added in update Statistics is statistics. "Average" doesn't mean "absolute truth." You will absolutely 100% be able to come up with counterexamples, because an average is just an average outcome. Where the variance that underpins that average can have significant impact, I try to make note of them. But just because you can come up with a counterexample or two doesn't mean the "average" case isn't "generally" true. Moreover, a key assumption underpinning everything in these simulations is PotD difficulty, and the fact that in most combat the amount of martial abilities martial classes get to use will be dwarfed by the length of combat. So while they may spike their DPS for short periods of time (especially in extreme cases, e.g. an empowered Inner Death from a monk), on average the net effect of their active offensive abilities will be smoothed out and dwarfed by persistent effects (sneak attack, carnage, confident aim, item enchants, etc.). WEAPON STYLES This is purely going to be evaluating the melee offensive styles. In other words, because weapon and shield style accomplishes a very different goal and ranged weapons are so different, I'm not going to talk about it. Most people intuitively think that two weapon style is the best style in Deadfire. And in general, that insight is right. I suspect it is actually due to a buggy or mistaken implementation of two weapon style, which reduces recovery time by 30% and then 15% with the talent. This is because, in general, two handed weapons do about 30% more damage than equivalent single-handed weapon classes, and then the two handed style adds an additional 15%. Seems like an interesting coincidence, huh? And I think whatever designers came up with these numbers intended this coincidence to balance out 2w vs 2h. Except that because of how recovery time bonuses work, it is actually much stronger than the numbers suggest (-30% recovery time actually translates into +42.8% action speed or +42.8% dps through the recovery; -15% recovery time translates into +18% action speed or +18% dps through the recovery, read my Umezawa build guide for more details). Added note in update: this is to say that a recovery time reduction is much stronger than what people might intuitively think, especially if one is not paying close attention to how the game talks about "action speed" in some cases and "recovery time" in other cases, and which are actually radically different concepts (and also the continuing source of much ongoing confusion about deadfire mechanics). This is not to say that 2w will give you a +42.8% net damage increase over other styles. So in a vacuum, it already dwarfed two handed style. Add to the fact that dual-wielders are much more responsive in combat, then it's no contest, unless you were specifically singling out using reach weapons, which lets you do melee-quality damage safely. That is, until Obsidian added a change in a patch that added +1 PEN to two handed weapons. This means that they all essentially have a +1 PEN advantage over "equivalent-class" single-handed weapons. Intuitively this means two-handed weapons are better in low-pen or just-on-the-edge-of-overpen situations. Running hundreds of simulations shows that on average, two-handed weapons do about 10% more net damage than dual-wielding. However, because this is a PEN-based interaction, the variance is important: in situations where PEN doesn't matter, two-handed weapons do worse, but the situations where that +1 PEN moves you up a PEN level or gives you OVERPEN heavily, and on PotD there are enough of these situations (mostly early on) that that +1 PEN moves two handed weapons from a crappy almost-trap choice into a viable alternative. You still give up some responsiveness in combat, but seems like a price some people might want to pay for extra damage. (This also means an optimal outcome would be to get both 2w and 2h and switch between the two based on PEN situations. But it does mean you can just wield 2h weapons all day long and still get viable outcomes instead of being stuck with a trap choice.) Enter single-weapon style. It is objectively a worse choice than either two handed and dual-wielding for straight-up damage. I think most people probably intuitively realized this. But mathematically, you have a greater than -10% net damage penalty compared to dual-wielding in most cases. (It varies wildly based on your base accuracy compared to the enemy). Single-weapon style does help you get more crits, so if you are a specifically crit-focused build (i'm thinking of monk, barbarian, or maybe chanter, who can do all sorts of powerful, non-damage-related effects on a crit) or are narrowly focused on a specific weapon that has a crit-based proc single-weapon style can be worth it. But in my mind it's extremely niche and is close to a trap choice for most players. STATS Here's where things get real interesting, because the simulations helped me realize things that weren’t obvious to me through simple calculations. First, let’s start off with some rules of thumb. If you’re trying to evaluate the relative quality of two different stats, how would you quickly reason about them? Through some simulations, I can tell you that +5 Perception is ~10% net damage, or about 2%/Perception (or per accuracy for that matter). This is not that too big of a surprise for people who follow the forums, since people have been computing this based on some basic assumptions about accuracy and enemy deflection since the backer beta. The real surprise is might. What’s the rule of thumb here? You might intuitively think 3% per might as it states in game but… you’d be wrong. In fact, each +5 might is also very nearly ~10% net damage, so similarly 2%/Might. Seems weird, right? There’s several factors at play here. The first is the way the game handles negative modifiers from graze: due to the double-inversion you recover from penalties following a polynomial curve, which means bonuses contribute very little and slowly ramp up until a negative modifier is fully cancelled out. In the case of a graze, therefore, might actually does not contribute 3% increased damage. In fact, assuming no other damage bonuses, might on a graze doesn’t actually contribute even 2% increased damage until about 27 might. Expand the next spoiler for a graph illustrating this. And on the top-end of rolls, when you crit, might is additive with the other damage bonus you get from crit and possible overpenetration, so the impact of might gets muted. So the end result is that you get approx 2%/might on average in most cases. Figure 1. Graph showing your graze modifier for Might from 10 to 35. Notice the curve. Except there’s a further tickle here. While the rule of thumb for both perception and might are the same, in practice perception is actually very slightly better than might (it’s more like 11% net damage per +5 Perception). That’s because while in many cases perception performs on par with might, in cases where you are underpenetrating, that extra perception might increase your chance to crit, and critting gives you a +50% boost to your penetration, which means your crit damage might do many times more than your normal hit damage. Dexterity, on the other hand, is pretty straightforward and very nearly 3% net damage per dex. It varies a bit based on what armor you’re wearing and what other action speed bonuses and penalties you have, but this is why this is a rule of thumb and not an absolute truth. So, in terms of dealing damage, the stat order is Dex > Perception > Might. For casters who don’t have great autoattacks and rely more on one-shot spells, the order might be more like Perception > Might > Dex. If you had asked me a week ago, I would have weighted Might higher than Perception and it’s solely because of these simulations that I came away with another answer. For barbarians, perception gets greater weight if you try to make use of carnage at all. Again, here, if you had asked me a week ago I would have significantly weighted Might as best damage stat, due to the fact that carnage only scales with Might and Power Level. In fact, each point of Perception/accuracy gives you basically polynomial returns, because carnage only activates on a hit and carnage itself has to make an attack roll to hit. That means a +1 to accuracy helps your initial hit, and then helps again to land your carnage hit. Might, by contrast, essentially is a linear, one-time multiplier for your total damage. It’s a subtle difference, but to illustrate in an extreme case, if you only had a 1% chance to hit against an enemy (who has a friend next to them), and you would normally do 100 damage on a hit (33 carnage), then one additional point of Might would take your expected total damage from 1.0033 to 1.0334, where expected carnage damage per attack attempt from .0033 to .00334. By contrast, one additional point of Perception/Accuracy would take your expected damage from 1.0033 to 2.01, where your expected carnage damage per attack attempt goes from .0033 to a whopping percentage increase to .0132 (effectively giving you +300% carnage damage, which makes sense of square(2) - square(1)). In fact, you need a very high base chance to hit (as shown by in-game targeting) of ~high 60% (virtually 70% or so) before an additional point of might begins to outweigh an additional point of perception in terms of total barbarian melee damage. On Path of the Damned, this is extremely rare for much of the game (and possible even for the mid-late game depending on how 3.1.0 rebalances the mid-late PotD game) so I feel comfortable advising all carnage-loving barbarians to invest in perception first. On lower difficulties, it might be easier to get to higher base chances to hit at which point might becomes closer to comparable. FAST WEAPONS VS SLOW WEAPONS People love them some sabres, and it's true they have a high inherent damage. Running simulations, though, you may be surprised that a sabre actually gets dwarfed by daggers, clubs, or rapiers. First, comparing equivalent-class fast weapon (.5 attack/3s recovery) and a slow weapon (.7 attack/4s recovery), ignoring weapon-specific bonuses, they are extremely close and comparable; the lower damage of fast weapons is largely cancelled out by the reduced total time spent doing an attack and recovery. (And unlike in PoE1, lower damage weapons aren't "more vulnerable" to armor than higher-damage ones.) The difference basically boils down to that rule of thumb about might and perception further up in the post. Daggers, rapiers, and clubs all get +5 accuracy as their weapon-specific bonus. Sabres get +10% damage. Converting that into stats, +5 accuracy => ~5 perception, or ~10% net damage. Sabres +10% damage => ~3 might or ~6% net damage. And sure enough, in simulations, sabres on average do about ~5% less damage than a dagger, rapier, or club. In fact, the +5 accuracy is good enough that for maximizing damage output the main reason why you'd use any slow weapon is essentially because their weapon modals lets you situationally get +2 PEN (or are a mace, which has inherently high PEN), whereas the stiletto is the only fast weapon that has a weapon modal for bonus PEN (and the stiletto has worse base damage than other fast weapons and in fact performs generally the worst of all weapons except in cases of extreme PEN).Now, to be fair, there are plenty of unique sabers, and one could argue that there are plenty of unique slow weapons with far more powerful effects than unique daggers, rapiers, and clubs. And this is true, and in the end the unique weapons will likely dwarf these differences between vanilla weapons, except in specific niches (e.g. pukestabber, marux amanth's special abilities, extra survival of the rapier with escape). But it does mean that if you've only been using sabres or axes for your characters early on, you should consider also giving the fast +5 acc weapons a shot. Edited October 19, 2018 by thelee 10 Link to comment Share on other sites More sharing options...
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!Register a new account
Already have an account? Sign in here.Sign In Now