Potential Calculator Inaccuracy for Frenzy Attack Speed der=0>

[onderduiker]

UPDATE 21 DECEMBER 2009 Hammerman has posted in the Calculator Inaccuracy for Frenzy Attack Speed topic in the Amazon Basin's Technical Discussion, explaining how Frenzy's attack speed is calculated when the first swing hits and isn't blocked: the break-points derived from the process he describes match those I've found experimentally. Hammerman's post is quoted in its entirety in this post.


There's recently been a debate in the Beginner Frenzy Barb! thread about whether TitanSeal's Attack Speed Calculator was accurate for Frenzy: although the initial argument about whether the calculator and Kyrene's formula produce the same break-points has been resolved, tests reported from this post onwards have indicated that the calculated break-points don't match actual break-points.

Although poison length isn't the most reliable means of measuring attack length, subsequent tests with the mana cost method have confirmed all calculated break-points for Double Swing with various combinations of Phase Blade, Colossus Blade and Colossus Sword (including those with weapon Increased Attack Speed and the known bug when the primary weapon is equipped in the right weapon slot), as well as level 20 Frenzy break-points when equipping two weapons with the same weapon speed modifier and no weapon IAS.

However, when two weapons with different WSMs were equipped or one of the weapons had IAS, the actual Frenzy break-points were different to those calculated (see this post and this one).

It looks as though left weapon slot WSM-right weapon slot WSM is always used to calculate Frenzy attack speed in some way, regardless of whether the primary weapon is equipped first in the left or right weapon slot: it also looks as though weapon IAS on the secondary weapon is taken into account in some way. However, I haven't sat down and tried to work out the actual attack speed formula yet, and more tests may be necessary to gather more data.

I just wanted to bring this to your attention, in the hope that others will test this themselves, either to refute my results and point out any errors, or help to work out how Frenzy's attack speed is actually calculated.

[SnickerSnack]

I just got a frenzy barb to level 45. I have a wide range of items to test with (several grey/white of nearly every 1-hand weapon).

The mana-cost-method you cite appears to require modding (ias charms). Is there any way that doesn't require modding?

I think I'd try something like stacking mana gear to get at least 500 mana, then swinging very low damage weapons at a target with enough iDR to negate the damage (or just high level, high defense with life regen), and seeing how long the barb can keep that up. The longer the run time for each test is, the less significant human error will be (keeping time).

[onderduiker]

You can't do precise tests without at least editing Increased Attack Speed items: IAS on items is normally a multiple of five, and the maximum you can equip without weapons is 165 IAS... but the difference between EIAS break-points doesn't become 5 IAS or greater until 142+ IAS.

Aside from mana cost, the Next Delay of missiles can be used to measure attack speed: however, this requires modifying the game files instead of just items. Poison and cold length are other possibilities I've considered, although I haven't found them 100% reliable.

However, anybody can perform a general test to confirm or refute what I've reported: for example, with no IAS whatsoever, equip a Falchion (WSM 20) first in the left weapon slot (above gloves), then a Cutlass (WSM -30) in the right weapon slot (above boots), and attack with Frenzy (the lower level the better, since Frenzy's EIAS isn't applied all at once and can make visible differences in attack speed harder to spot); repeat after equipping the Cutlass first in the left weapon slot, then the Falchion in the right weapon slot.

If the calculator is accurate, attack speed should be the same regardless of which weapon is equipped first in the left weapon slot, whereas if my tests are accurate, attack speed should be much faster when the Falchion is equipped first in the left weapon slot.

Without modifying, you could test my results so far to see if they're at least in the right ball park, and you can do likewise with any further results. It might take a while for me to post further results, though, since finding unknown break-points is more time consuming than confirming accurate ones.

[SnickerSnack]

I had a barb with level 1 frenzy swing an Axe [10] and a scimitar [-20] at a high level paladin with HS and Defiance on. The attacks were faster when scimitar was equipped first. The barb no ias.

[onderduiker]

Just to confirm, did you equip the Scimitar in the left weapon slot (above gloves)? If it were equipped in the right weapon slot (above boots) then based on my own tests it would indeed be faster (it appears as though it doesn't actually matter which is equipped first, just where it's equipped).

Before I continue testing Frenzy, I'll test Double Throw and the Barbarian's other attacks when equipping two weapons, and possibly do likewise for the Assassin: hopefully that'll confirm that any issue (if there is one) is limited to Frenzy.

I'd forgotten another possible method for measuring attack speed: recording a video of the attack and then playing it back to count the frames.

[onderduiker]

Okay, further tests indicate that the calculator should be accurate for Double Swing, Double Throw, Normal Attack, Throw and skills like Concentrate when equipping two weapons: it only seems to be inaccurate for Frenzy (I haven't checked Zeal, Feral Rage or Left Hand Throw).

Unless SnickerSnack reports back to confirm where the weapons were equipped, or others perform the test and report likewise, it seems unlikely that I'd be able to confirm the calculator's accuracy for so many skills and only make mistakes whenever I tested Frenzy: it also seems unlikely that this only affects single-player, but you never know...

Having conducted those tests I think I have an idea what might be going on with Frenzy: I've noticed that contrary to DS, the length of each Frenzy swing is noticeably different when equipping weapons with very different weapon speed modifiers. This, coupled with what I observed when testing Normal Attack break-points (based on the animation itself and the calculator's break-points), leads me to believe that the length of each Frenzy swing is calculated separately.

[SnickerSnack]

I didn't take note of which item slot held each item (only which was equipped first), and I deleted the character since I thought I was done. I'll redo the tests sometime this week.

Can you give a list of what should be tested? I can shop for plain items and I could use a frenzy level higher than 1. I could modify some ias charms.

leads me to believe that the length of each Frenzy swing is calculated separately.

That could be.

[onderduiker]

TitanSeal has responded to the Potential Calculator Inaccuracy for Frenzy Attack Speed topic in the Amazon Basin's Technical Discussion. I'm not sure I could summarise it properly yet, so take a look if you're interested: it seems I'm not crying wolf, which is a relief.

There are four possible combinations when equipping two weapons with different weapon speed modifiers, like a Falchion (WSM 20) and Scimitar (WSM -20), both of which can be bought in Normal:

LEFT WEAPON SLOT   RIGHT WEAPON SLOT
____________________________________
1. Falchion        2. Scimitar
1. Scimitar        2. Falchion
2. Falchion        1. Scimitar
2. Scimitar        1. Falchion

The primary weapon (1) is whichever is equipped first, and is always displayed in the Barbarian's right hand. If you switch weapons or retrieve a corpse, the game will equip the weapon in the left weapon slot (above gloves) first, making it the primary weapon: the only exception is when the weapon in the left weapon slot cannot be equipped until the weapon in the right weapon slot (above boots) is equipped (for example, if it cannot be equipped without a Breath of the Dying 'VexHelElEldZodEth' weapon's +30 to All Attributes).

If the calculator were currently correct, then the two combinations with the primary weapon in the left weapon slot would result in the same speed with no Increased Attack Speed and level 1 Frenzy: however, my tests indicated that the combinations with the same weapon in the left weapon slot resulted in the same speed, regardless of whether or not it was the primary weapon. Furthermore, attack speed was fastest when the weapon with the slowest base speed (the Falchion) was equipped in the left weapon slot.

If your test results agree with those of TitanSeal and myself, then you could start finding break-points by adding IAS and mana. After testing other skills while equipping two weapons I'm taking a break, but when I resume I'll go into more detail.

[onderduiker]

A quick update... according to TitanSeal and corroborated by testing, the calculator is accurate whenever the first attack in the Frenzy sequence misses or gets blocked: since most monsters don't block and almost all weapon attackers strive to maximise their chance-to-hit, this means that the calculator will still be inaccurate more often than not against monsters... but even with maximum 95% chance-to-hit, it will be accurate more often than not against targets with 48+% block.

SnickerSnack, this might explain why your results differed from mine, even if you had equipped the primary weapon in the left weapon slot (above gloves): my Barbarian had 95% chance-to-hit against a target with no block, while yours was attacking a Paladin with plenty of +% Enhanced Defence from Holy Shield and Defiance, and presumably with 75% block... if I had 95% chance-to-hit and you had 5% chance-to-hit against 75% block, then according to me the calculator would be inaccurate 95% of the time, while according to you it would be accurate 98.75% of the time.

[SnickerSnack]

SnickerSnack, this might explain why your results differed from mine, even if you had equipped the primary weapon in the left weapon slot (above gloves): my Barbarian had 95% chance-to-hit against a target with no block, while yours was attacking a Paladin with plenty of +% Enhanced Defence from Holy Shield and Defiance, and presumably with 75% block... if I had 95% chance-to-hit and you had 5% chance-to-hit against 75% block, then according to me the calculator would be inaccurate 95% of the time, while according to you it would be accurate 98.75% of the time.

So, I should test against a target with no block and little defense, but lots of life then?

Here are some tests done with ~5% to hit and no blocking on the target:

So it looks like it's fastest to put the scimitar in the right slot first, if the first hit misses a lot (not really a useful test). These tests were done without frenzy "active".

LEFT WEAPON SLOT   RIGHT WEAPON SLOT
____________________________________
2. Falchion        1. Scimitar
                   >
1. Scimitar        2. Falchion
                   =
1. Falchion        2. Scimitar
                   >
2. Scimitar        1. Falchion

I'll have to see what I can do to get a target with lots of life and low defense/level.



Edit:

Okay, so frenzy_test (lvl 24) has a scimitar, a falchion, ~2k mana, and 7 ias charms (64,32,16,8,4,2,1). frenzy_target (lvl 9) has ~10k life and ~130 life regen. I can swing frenzy at nearly any speed and for as long as I want. What would you recommend for finding bps? Do I need to record the tests? Or is there an easier way?