Attack speed 2.0

[Kaylon]

Thanks - I also wanted to know that. What if you add DEX 20 (+30% speed) into the mix? 1.5*1.2*1.3 = 2.34 => base/2.34? Or how is the bonus from DEX calculated?

You're correct. Everything is divided by (1+DexBonus)

[Boeroer]

Thanks. That means ~43% of reloading time for a ranger with Swift Aim, Gunner and DEX 20. Hm, quite a decrease (counted in frames) if you use an arquebus... Too sad there are no enchantments that buff reloading speed. Attack speed buffs for weapons with reloading animation are just a waste (I'm looking at you, Hold-Wall!).

Edited February 2, 2016 by Boeroer

[Dr. Hieronymous Alloy]

A lot of things that impact attack speed only should also hit reload speed.

[MaxQuest]

 

Thanks. That means ~43% of reloading time for a ranger with Swift Aim, Gunner and DEX 20. Hm, quite a decrease (counted in frames) if you use an arquebus... Too sad there are no enchantments that buff reloading speed. Attack speed buffs for weapons with reloading animation are just a waste (I'm looking at you, Hold-Wall!).

 

Checked that setup in-game:

Arbalest with (20 dex, Swift Aim, Gunner) resulted in: 43 attack frames, 47 recovery, 74 reload (164 total frames)

Arbalest with (10 dex and no buffs) resulted in: 52 attack frames, 75 recovery, 167 reload (294 total frames)

So yeah, reloading speed was reduced to 44%.

and

Hold Wall with (20 dex, Swift Aim, Gunner) resulted in: 43 attack frames, 32 recovery, 74 reload (149 total frames)

 

-------------------------------

 

@AndreaColombo

So. I have paid close attention to the example you have provided, made around 20 extra tests, and now I can assert the following:

 

- First of all you are right in saying that stacking in plain additive manner will not bring that setup to zero recovery.

new_recovery_time = base_recovery_time * (1- 0.5 - 0.15 - 0.2 - 0.15 + 0.5 - 0.15) = base_recovery_time * 0.35

 

This formula hints us that recovery phase will pass around 3 times faster, but yes it won't be zero.

 

- Second: stacking bonuses in plain multiplicative manner, also doesn't help:

-- approach 1:

new_recovery_time = base_recovery_time / 1.5 / 1.15 / 1.2 / 1.15 / 0.5 / 1.15

new_recovery_time = base_recovery_time * 0.73

 

-- approach 2:

recovery time and recovery speed are inverse proportional, so:

new_recovery_time * new_recovery_speed = base_recovery_time * base_recovery_speed

new_recovery_time = base_recovery_time / (base_recovery_speed / new_recovery_speed)

 

where

new_recovery_speed = base_recovery_speed * 1.5 * 1.15 * 1.2 * 1.15 * 0.5 * 1.15

new_recovery_speed = base_recovery_speed * 2.38 * 0.5 * 1.15

new_recovery_speed = base_recovery_speed * 1.36

 

thus:

new_recovery_time = base_recovery_time / 1.36

new_recovery_time = base_recovery_time * 0.73

 

It's like: you are in a good mood, and taking a trip by car, from city A to city B. You know that it will take you 1 hour if you move at your usual (base) speed of 1km/minute. Now imagine that your impatient friend is asking you to speedup by 50%. Then by 15%. By 20%. And once again by 15%. If you would check the speedometer, you could see that current speed probably is 2.38 km/minute. The thing is you will not get from city A in city B in zero time, unless you speed-up x infinity.

 

- Third: the calculations you provided indeed do work. Yes, the formula seem to lack the aforementioned correlation with physics/reality (see 2nd spoiler). But it doesn't matter. Devs implemented it that way, and the tests confirm such behavior.

 

It would make sense to rename all "increases Action Speed" to "reduces Recovery Time".

Edited February 3, 2016 by MaxQuest

[AndreaColombo]

^ yeah, I was taken aback by the nomenclature too at the beginning, but then I thought the devs must've figured "attack speed" would be easier to understand for casual gamers.

[AndreaColombo]

-------------------------------

 

@AndreaColombo

-- approach 2:

recovery time and recovery speed are inverse proportional, so:

new_recovery_time * new_recovery_speed = base_recovery_time * base_recovery_speed

new_recovery_time = base_recovery_time / (base_recovery_speed / new_recovery_speed)

 

where

new_recovery_speed = base_recovery_speed * 1.5 * 1.15 * 1.2 * 1.15 * 0.5 * 1.15

new_recovery_speed = base_recovery_speed * 2.38 * 0.5 * 1.15

new_recovery_speed = base_recovery_speed * 1.36

 

thus:

new_recovery_time = base_recovery_time / 1.36

new_recovery_time = base_recovery_time * 0.73

 

It's like: you are in a good mood, and taking a trip by car, from city A to city B. You know that it will take you 1 hour if you move at your usual (base) speed of 1km/minute. Now imagine that your impatient friend is asking you to speedup by 50%. Then by 15%. By 20%. And once again by 15%. If you would check the speedometer, you could see that current speed probably is 2.38 km/minute. The thing is you will not get from city A in city B in zero time, unless you speed-up x infinity.

 

- Third: the calculations you provided indeed do work. Yes, the formula seem to lack the aforementioned correlation with physics/reality (see 2nd spoiler). But it doesn't matter. Devs implemented it that way, and the tests confirm such behavior.

 

 

Re: your example with the car getting from A to B.

 

If you increase your speed 100%, you'll get from A to B 50% faster (e.g. if A to B is 100km and you travel at 100km/h, it takes you 1 hour; if you increase your speed 100% to 200km/h, it takes you half an hour.)

 

However, I believe this is not a suitable comparison to applying speed bonuses to recovery. What we're doing here is not, in fact, increasing the speed with which we attack; rather, it is decreasing the interval between attacks.

 

A more apt analogy would be to decrease the distance between A and B. Of course it is physically impossible to decrease the distance between two geographical locations by any amount in the real world, but conceptually if we were to decrease the distance between A (first attack) and B (second attack) by 100%, we could then get from A to B in no time (i.e. with no recovery.)

 

 

On a mildly related note, I do not consider the undefeatable 4-frame delay that remains even after decreasing recovery by 100% to be part of either your attack animation or your recovery. I believe it stands on its own as something the developers added to prevent visual glitches in attack animations in case speed bonuses were stacked like we're doing here, up to 100%. After all, in the grand scheme of things 4 frames really are a negligible delay: 4/30 = 0.13 seconds, unlikely to impact anyone's DPS in any significant manner.

Edited February 3, 2016 by AndreaColombo

[MaxQuest]

However, I believe this is not a suitable comparison to applying speed bonuses to recovery. What we're doing here is not, in fact, increasing the speed with which we attack; rather, it is decreasing the interval between attacks.

Exactly. That's why I proposed to rename "speed increase" to "time decrease".

Although this might sound like the same thing, it actually allows the player to make the right estimation of expected benefit (e.g. decreasing time by 50% is not the same as increasing speed by 50%). Not to mention that the bonuses are applied only to recovery part, and not the attack. Which can be confusing for casual/newcomer player. For example I had the feeling that something is not right.

 

On a mildly related note, I do not consider the undefeatable 4-frame delay that remains even after decreasing recovery by 100% to be part of either your attack animation or your recovery. I believe it stands on its own as something the developers added to prevent visual glitches in attack animations in case speed bonuses were stacked like we're doing here, up to 100%. After all, in the grand scheme of things 4 frames really are a negligible delay: 4/30 = 0.13 seconds, unlikely to impact anyone's DPS in any significant manner.

- Yes, from that point of view, I agree. That is a delay between actions. So it is not part of action.

- On the other hand, as a player I am only interested in how many times my dps will increase if I will take those talents/buffs/enchants/etc. And that delay is part of what happens between two consecutive damage appliances. Part of what we could call "Full Action Cycle". You name it. As for negligence, it is negligent, unless someone wants increased precision and probably dual-wields fast weapons.

Edited February 3, 2016 by MaxQuest

[knownastherat]

^ Josh said they know how painful it is for players to calculate their DPS in this game and it is something they'd fix with high priority for a sequel.

 

For PoE, however, this is as good as it gets.

 

Sounds good. 

[Furism]

Since most of the theorycrafters hang around this thread I thought I'd ask here : is there an up to date chart showing the benefits (for damage output) of each attribute over the levels? If I remember correctly there was one in the past showing that Might was strong at the early levels, then fell flat, whereas Dexterity was the opposite and Perception was fairly linear.

[Kaylon]

Since most of the theorycrafters hang around this thread I thought I'd ask here : is there an up to date chart showing the benefits (for damage output) of each attribute over the levels? If I remember correctly there was one in the past showing that Might was strong at the early levels, then fell flat, whereas Dexterity was the opposite and Perception was fairly linear.

Yes, you can try to make a DPS chart, however it wouldn't tell the entire truth. A dagger and high dexterity could give you the best DPS against xaurips, but against a dragon your DPS could benefit more from might if you're using sabres...

[MaxQuest]

If I remember correctly there was one in the past showing that Might was strong at the early levels, then fell flat, whereas Dexterity was the opposite and Perception was fairly linear.

If you swap Dex and Perception there, you get it right. (I mean benefit from dex is a somewhat linear increase. While perception/acc is an increase with few down-bumps (on 15,35,50), and then a flat wall once you achieve acc - enemy_def>= 100.)

 

But tbh there are too many variables/scenarios to take into account. For example:

- on grazes. might is most important

- on crits, might is less important

- enemy has high DR? might again becomes more important

- for cipher and rogue physical attacks, might will provide less benefit in comparison to other classes. (whips/sneak attack)

- but if you miss non-stop, might and dex are irellevant. Get perception

- got too much acc such that you crit 100%? getting more per will not increase your damage any further

- dex provides the most flat benefit of those three. Although just stacking it, is also subject to relative diminishing returns

- might (like dex) is also subject to relative diminishing returns, but enemies with high DR somewhat nivelate that

and so on.

Edited February 27, 2016 by MaxQuest

[hilfazer]

MIG and PER are subjects to overkill while DEX is not.

[AndreaColombo]

MIG and PER are subjects to overkill while DEX is not.

 

DEX is subject to diminishing returns and DPS loss from DR, though.

[MaxQuest]

Uhm guys... did you know that you can attack without recovery when DW, with just [Two-Weapon Style talent], [Gloves of Quickness] and [swift Aim]?

And even wear 10% recovery armor. Their bonuses seem to be applied twice.

 

Have checked, and re-checked, and re-checked it again.

 

Tests sheet (test v7)

Edited May 8, 2016 by MaxQuest

[wiarygodny]

Uhm guys... did you know that you can attack without recovery when DW, with just [Two-Weapon Style talent], [Gloves of Quickness] and [swift Aim]?

And even wear 10% recovery armor. Their bonuses seem to be applied twice.

 

Have checked, and re-checked, and re-checked it again.

 

Tests sheet (test v7)

 

Something weird might going on, bonuses to defenses from anti-affliction talents and lopped rope were also doubled last time I checked, but I didn't played since 3.0.

[Boeroer]

Also: Thrust of Tattered Veils procs twice when you score a crit with Badgradr's Barricade. Savage Attack gets applied twice (hover over your ACC on your char-sheet when Savage Attack is active). Something's really broken inside PoE atm...

[Boeroer]

I just took a glance to the support forum: none of those issues was reported yet I think. Maybe we should do that? It seems to be important...

 

edit: ok, I reported the things I said.

Edited March 27, 2016 by Boeroer

[AndreaColombo]

When I load a save from my last play through and activate Savage Attack, I don't get a double penalty in 3.02 Is this only affecting new games started after patching?

[Boeroer]

Nope. My game is from 3.0 originally and my Edér gets -5 AND -5 from Savage Attack. Don't know about the damage. Have to check that. Would be ok if that also gets applied twice.

[AndreaColombo]

Dunno; I've just loaded up a save again and I'm only getting -5 from Savage Attack. I tried on my main, Edér, and Zahua.

[MaxQuest]

Something weird might going on

Something's really broken inside PoE atm...

Hmm, I am not sure if that (double bonus from "attack speed") is a bug or is related to recent changes. Because:

- it seems that it was the same in 2.0. (I have an un-updated 2.0 on an old laptop which I haven't touched for quite awhile; I have staged and frapsed a similar situation, with only Gloves, 2WpnTln and SwiftAim, and got zero-recovery as well) (added as test v11 to the spreadsheet)

- it could be intended. Do you remember the old diagrams, were it was told that fast weapons do have 20 frames attacking phase and 20 for recovery? In order to decrease the action time (att+rec) by 20%, devs could apply 20% to recovery twice as a workaround, while leaving attack_duration untouched. And as we know atm, attack_duration is indeed not being affected by all those attack_speed bonuses. Why could they have done it this way? I don't know, although I could assume that it is related to animations. Standing in place (recovery) is easier to shorten or extend than the animation of swinging weapon without producing any tearing or abrupt transitions. 

 

Maybe actually the bug is that 1H gets the bonus only once. We'll never know, unless Obsidian actually reveals base attack/recovery/reload frame durations and the formulas as well.

I bet there is something really messed/unconsistent there.

 

P.S. Interesting, what was the reason for going from 20/20 to 4+(3+20)/15... 

Edited March 28, 2016 by MaxQuest

[Kaylon]

Ok, I understood what happens when you have two weapons equiped. When a single weapon is equiped you have 100% recovery. While using two weapons the game removes 50% recovery and you're left with 50% recovery for each hand. The speed on the weapon is affecting ONLY the recovery for that weapon - that means it's possible to have 0% recovery for one hand and some recovery for the other hand. Of course, the speed bonuses coming from other sources than weapons are affecting both hands.

 

Also the two weapon style bonus stacks multiplicatively with the other speed bonuses. For example Zahua, once he gets Anitlei ability, is able to attack without recovery: 1.2 (two weapons style) x 1.15 (gloves of swit action) x 1.1 (Anitlei) =1.52.

[AndreaColombo]

So the 50% cut to recovery from when you dual-wield stacks additively with every other bonus?

[Kaylon]

So the 50% cut to recovery from when you dual-wield stacks additively with every other bonus?

Yes

[Boeroer]

What a mess with this whole attack speed thing...