Guild Wars 2

Dojo.1867:

SB autoattack: 0.76 per second
LB autoattack: 0.72/0.52/0.4 per second (Note that even on long range it is inferior not factoring in bleeds if you flank)
Using LB Rapid Fire: 0.75 per second (3.75 over 5 seconds)

Don’t forget that longbows have higher base weapon damage than short bows too.

Aegis.9724:

SB was always bugged and never got that 100% increase to quickness. Now, after the nerf, it feels more like 30% instead of 50, but never ran timed tests to prove it.

I have. Your “feel” is pretty close. Quickness increases short bow attack speed by 31.3% (used to be 62.6%).

This is a great analysis! Good job.

However, one thing that is not explicit in our weapon skills is that quickness does not actually increase our attack rate by 50% (and pre-patch it didn’t increase it by 100%, either). For shortbow, it’s closer to 31% now (which I tested here). If you’re using a shortbow and maximizing use of quickness, that means this patch hit you for about 5.1% of your DPS over extended fights.

ItIsFinished.9462:

Toma.6874:

Is it me or its the same ? It feels slower somehow.

You are correct. The time between each shot has been increased slightly, but the traveling speed of the arrows I believe have been increased to compensate. I have no hard data or proof.

I do. (See my post above.) Shortbow attack speed has not changed.

Bri.8354:

That’s interesting, receiving exactly half of what it used to despite this update. I always thought it was a bug but this makes me wonder if they intentionally cut the benefit crossfire receives from quickness by half.

Well they cut the benefit everything receives from quickness by half in this patch. It’s just that before we were only getting a 62% increased shortbow attack rate under quickness, and now we’re getting 31% (by my numbers).

Dojo.1867:

Can confirm this. Updated my thread aswell.

My calculated coefficients are 1.16 (old) and 1.57 (new).

I’m still curious as to why we’re getting different numbers. At least we’re getting consistent results, if offset a little bit.

From the patch notes:

Ranger
Maul skill:
Increased base damage by 26%.

Maul (Greatsword #2 skill) previously had a coefficient of 1.120, and with the patch today, the damage coefficient is now 1.500 (test data here). That’s an increase of 34%!

Rangers OP now? :P

I recorded quickness frame-by-frame and observed 8 attacks in ~3.17 seconds. (I wanted to be sure all 8 shots were fully within the quickness window.) That’s ~2.52 attacks per second, which, compared to 1.92 attacks per second without quickness, is a gain of 0.6 attacks per second.

Before today’s patch, quickness increased the shortbow attacks by 1.2 attacks per second, so while it’s true we’re not getting the advertised 50% increase to attack rate during quickness, we are getting half of what we were getting before.

No, shortbow attack speed is the same as it was before today’s patch.
Still 1 attack every 0.52 seconds.

Test data here.

OGDeadHead.8326:

Nice number crunching, but a bit misleading imho. Quickness is for burst damage (or for quicker use of ressing, stomping, whatever) – it’s what it did during the 4 seconds it was running that that mattered – in short double damage (or should’ve been) during those 4 seconds from direct damage (not really in the case of the shortbow due to other things that’s already been discussed). You can’t really compare damage during the course of a minute, it’s really not of any interest when talking about abilities of this kind.
Not only have the direct damage part been nerfed, but also the ability to stack bleeds if we’re talking shortbow, the ability to stack might if talking about sword… the list affected by this goes on and on.

That’s a fair point about burst damage, but remember that anything that procs on hit with no internal cooldown ends up scaling proportionally with attack speed. That is to say that the bleeds from shortbow skill #1 are not affected any more than the direct damage (again, with respect to sustained DPS).

From today’s patch notes:

Reduced quickness from 100% attack speed to 50%.

How does this affect rangers’ shortbow damage?

• Shortbow’s base attack rate is 1 per 0.520 seconds, which is 1.923 APS (attacks per second). (Test data here.)
• Before the patch, quickness increased the attack rate to 1 per 0.320 seconds, or 3.125 APS. (Test data also available at link above.)
• With quickness’s effect now halved, we should expect (theoretically) that the new APS with quickness is halfway between the old quickness APS and the base APS, which comes out to (1.923+3.125)/2 = 2.524 APS during quickness. This is a gain of precisely 0.600 APS over the base attack speed. (When I have time, I will seek to confirm this through testing. Edit: Confirmed. See note at bottom.)
• Quickening Zephyr (utility skill) now lasts 5s and therefore grants a total of 3.0 extra attacks over that time.
• Zephyr’s Speed (Beastmastery minor trait) lasts 2s and therefore grants a total of 1.2 extra attacks over that time.

So let’s say our shortbow attacks deal 520 damage on average. (Let’s ignore condition damage and pet damage for now.)

• With no quickness effects, that’s {520 damage per attack × 1.924 APS} = 1000 DPS.
• Using QZ (60s cooldown) raises total DPS to {(520 damage × 2.524 APS × 5s uptime + 520 damage × 1.923 APS × 55s downtime)/60s cooldown} = 1026 DPS, which is a gain of 2.6% (pre-patch gain would have been: 4.2%). If it’s traited (48s cooldown), it’s 1033 DPS instead, which is a gain of 3.3% (pre-patch: 5.2%).
• Pet-swapping and gaining ZS every 20s increases the base 1000 DPS to 1031 DPS, which is a gain of 3.1% (pre-patch: 6.2%). If pet swapping is traited (16s cooldown), that’s 1039 DPS instead, or 3.9% (pre-patch: 7.8%).
• Maximum total gain by traiting for and using QZ and ZS is {3.3% + 3.9%} = 7.2% over using no quickness effects at all (pre-patch total gain: 13.0%).

In your opinion, does this change make Quickening Zephyr and/or Zephyr’s Speed non-essential? (Or if they were already at that point, does this change make QZ and/or ZS weak?)

Edit: I’ve confirmed through testing that the quickness change is indeed working as I assumed above. Shortbow #1 skill attacks at the rate of 1 attack per 0.396 seconds, or 2.525 APS (within measurement resolution error of ±0.003 APS).

Yeah, my testing from a few months ago showed that Swoop’s tooltip does not line up with its actual performance. It deals damage at a 1.0 damage coefficient, but the tooltip shows damage numbers as if the coefficient were 1.4.

phys.7689:

…what many of you are asking for (an unstoppable 40% ranger damage bot) is completely unreasonable.

That’s an inaccurate generalization of what most people are asking for.

Most rangers in this thread (and with whom I’ve talked to outside of this thread) don’t want an unstoppable AI beast that does 40% of our total damage. I know what I personally want is to be able to control as close to 100% of my damage as possible. If the AI beast following me around can’t be adequately controlled (in other words, if whether my pet lives or dies is not up to me), then I want most of that 40% of damage put back into my hands.

If my ranger accounted for, say, 90% of my damage (with my pet as the other 10%), then I would be much less irritated when 90% of my damage dodges out of the AoE but 10% does not, or when 90% of my damage can be dealt from on top of a wall while 10% cannot.

Solandri.9640:

The GS autoattacks with the steady weapons in this minimal config (916 pow) were 27, 27, 32.

I’ll have to see if I can replicate those results.

Solandri.9640:

So with truncation the maximum error in the calculated coefficients should be 1/27 and 1/32 = 3.7% or 3.1%. If it’s being rounded off, the maximum error should be half that.

I agree with this, but keep in mind that the truncation error is only one-sided, whereas the error for rounding is on both sides. This means the size of the uncertainty range for both cases is exactly the same. For example, using your data, Power Stab’s “true” damage value is in the range 32.00 to 33.00 if truncated, or 31.50 to 32.50 if rounded.

From my results, the coefficient uncertainty ranges for each skill at varying amounts of power did not overlap unless the numbers were assumed to be truncated rather than rounded.

Solandri.9640:

Tibbel.3450:

I tested each of the ranger weapon skills back in October.

I came to different numbers than you did, although I used slightly different assumptions as well.

I’ve been way too busy to play or test the past few weeks. But the ranger GS skill coefficients Dojo got match the ones I calculated when I tested in February. His are all within 0.01 of mine.

I hope you don’t mind if I ask about your methodology, then.

Specifically, what was the target armor that you assumed for your tests? I’m also curious whether you used runes, sigils, or traits, or accounted for rounding-vs-truncation. It seems like the numbers you and Dojo got are right around 5% higher than mine, which seems like it could be due to one of these factors.

Haldunius.7864:

Ok, what if I try this way….please help this lonely soul fulfill its purpose….

How’s this?

Dojo.1867:

No. Might have been a stealth nerf but I assume it must have happened at some point comparing our data. It is nearly as if they just swapped Maul and Swoop damage around, haha.

Your saying this made me double-check the image I linked, and I saw that the image actually had the Maul and Swoop coefficients reversed. (D:) Maul is actually 1.120 and Swoop 1.000, at least as far as I have tested. I’ve fixed it now.

Dojo.1867:

Also for my modifiers scaled per second with cooldowns and animation times. Here is the formula I used:

(Chain1 modifier + Chain2 modifier + Chain 3 modifier)/(Chain1 animation time + Chain2 animation time + Chain3 animation time)

So this would be the skill modifier per second for just autoattacking and we call it X. For the other skills we can now calculate the skill modifier per second for using them whenever cooldown allows it and autoattack otherwise:

((Cooldown – Animation time)*X + Skill modifier)/Cooldown

Can you check if this makes sense or did I make a mistake?

That formula seems right, although the attack animation times shown on the in-game tooltip are not an accurate indication of time between attacks. For example, the longbow #1 skill tooltip shows an animation time of 0.75s; however, we have shown that the actual time between attacks is more like 1.24s. I have not done any attack speed testing on ranger skills other than shortbow #1, longbow #1, and longbow #2; and I haven’t done any attack speed testing for warriors at all.

Yeah, I didn’t mention it in my previous post, but I did all the testing with no traits, runes, or sigils, and I filtered out crits and condition damage, so the only variable was the power from jewelry (which was accounted for in the data sheet).

I didn’t remember that Swoop was nerfed. Do you remember when that happened (or have a link to the patch notes, by chance)?

I tested each of the ranger weapon skills back in October.

I came to different numbers than you did, although I used slightly different assumptions as well. (This was back before they fixed the armor values of the medium PvP target golem and before they buffed the direct damage on GS#2 Maul.)
Here’s my full testing data, with a summary below:

https://docs.google.com/spreadsheet/ccc?key=0AjK4QtqFi-4ddHBodk4yYmszeDR4cWhRQmRFVEk1ZEE&usp=sharing

Summary of Results:
http://i.imgur.com/RliRtfs.png

Assumptions:

• The damage coefficients are probably round numbers, probably to the nearest 0.025. (i.e. 0.403 is probably actually 0.400)
• The target golem armor values are not necessarily round numbers. My starting assumptions are listed below (however they were refined as I tested multiple weapon skills against them):
  - Light: 2200
  - Medium: 2400 (off by quite a bit)
  - Heavy: 2600
• Note about the medium golem: Initially, its armor was below that of the light golem. (This was the case at the time of this testing, which found the armor value to be in the range 2016-2017.) However, it was changed in a more recent patch to an armor value in between that of the light and heavy golems. Subsequent testing results in an armor range of 2318-2322.
• Note about GS#2 skill, Maul: The direct damage of Maul was buffed since this testing was performed. These results showed a coefficient of 0.800 for Maul (pre-buff), and the patch notes for the change stated a 40% buff. Informal testing (performed using the same methods here, only with data not recorded here) has confirmed that the buff was indeed the stated 40% (up to a new coefficient of 1.120).

Method:

• Use the PvP steady weapon to perform each of the attacks against each of the golems, and record the non-crit damage value as reported in the combat log.
• Since the reported damage values are all integers, the “true values” must have been either rounded or truncated. If rounded, the true value would lie within -0.5 to +0.5 of the reported value, or if truncated, between +0.0 and +1.0. In order to be acceptable, all test results must fall within this expected range for each of the skills tested.
• Using the above technique, three ranges of possible true values are generated (one from each golem), each of which is very close to a round number, e.g. 0.501- 0.507. In this case, the round number, e.g. 0.500, is assumed to be the true value. In the datasheet linked above, the limits of these ranges are shown as “Coef. min.” and “Coef. max.”.
• From the round coefficient, the original armor assumption is adjusted so that the coefficient ranges from each golem and each skill enclose their respective round numbers. This adjustment includes a range of possible assumed armor values so that, for example, the damage range from above is in the range of 0.494-0.500 up to 0.500-0.506. There is a “makes sense?” column which tests whether this is possible for the given armor assumption. This range of possible armor values is shown in the bottom right of each tab of the datasheet.
• Since the GS#4 skill Counterattack requires a successful block, a different method was required. Once the coefficients for GS#1 skill Slash was found, it was used against each of the PvP NPC opponents (the NPCs that have class names back in the rear area), and the damage difference was used to find their armor values. Then GS#4 was used and a successful counterattack was triggered against the NPC, and the NPC’s armor value was used to narrow in on the skill’s coefficient.

Other Findings:

• Damage numbers are truncated, not rounded. If an attack “should” hit for 99.8 damage, it actually hits for 99, not 100.
• Axe skill #4, Whirling Defense, applies its stack of vulnerability before its damage applies (so its first tick of damage benefits from the first stack of vulnerability, its second tick benefits from the second stack, etc).
• I also performed testing on the attack speed of bow auto-attacks. The longbow #1 skill, Long Range Shot, fires 1 shot per 1.24s. The shortbow #1 skill, Crossfire, fires 1 shot per 0.52s
• Quickening Zephyr increases the attack rate of the shortbow #1 skill, Crossfire, from 1/0.52s to 1/0.32s, or an increase of 62.5%.

In case you’re interested, I tested warrior longbow coefficients in the same way:

1. Dual Shot (single hit): 0.300
2. Fan of Fire (single hit): 0.400
3. Arcing Arrow (single hit): 2.350 (yes, really)
4. Smoldering Arrow: 0.200
5. Pin Down: 0.400
   Burst skill, Combustive Shot (single hit): 0.450

I will not use spirits, pretty much no matter what they do to their mechanics. I play a ranger to deal ranged physical damage, not summon magical forest ghosts. (This is, of course, my completely subjective opinion.)

I guess the only chance would be if they were completely untargetable and immune to all damage, but I would still rather that the “unparalleled archers who are capable of bringing down foes from a distance with their bows” have more utility options that actually help their bow skills.

My opinion is that the pet’s share of overall damage should be greatly reduced, and the ranger’s increased. If the current split is 70% ranger damage to 30% pet damage, that ratio should mover closer to 85/15.

Reducing rangers’ reliance on pet damage will soften the effects on balance that are caused by their pets’ lack of survivability, attack uptime (such as from keep walls), and stat scaling.

Zaith.9132:

Tibbel, you’re talking progressively and theoretically. If I roll a 100 sided die 10 times, approximately 1/3 will end up 35 or lower.

Right, and when you rolled it 10 times, you’d have a 98.7% chance of at least one proc.
I’m not sure what you mean by progressively. If you mean cumulatively, then you’re correct — and cumulative probability is important when considering a proc with a cooldown.

Zaith.9132:

Every time it did, that effect would trigger.

This is what is disputed by the claim that there’s an internal cooldown.

Zaith.9132:

• Improved the effect chance on ranger spirits to 35% base and 50% when traited.

The spirits have an internal cooldown of ten seconds. The effect can only be triggered once every ten seconds. If you attack 3 times, untraited, you are statistically promised the effect will trigger. The only way this buff helps is if you attack twice or less every ten seconds.

That’s not how probability works.
Here’s the chance you will get a proc given the number of times you attack (assuming a 35% proc chance each hit):

1. 35.0%
2. 57.8%
3. 72.5%
4. 82.1%
5. 88.4%
6. 92.5%
7. 95.1%
8. 96.8%
9. 97.9%
10. 98.7%

RustyEyeballs.8927:

This was done with 2130 power with steady weapons on the heavy target dummy in the mists.

…

Axe
Damage: 44,50,53
Time Between Attacks: 1.071
DPS (bounce back to 1st target): 90
DPS (single tar): 44

I could not replicate these results.

I tested using 2139 power and the PvP Steady Axe (127-127 weapon strength). I used the Berserker’s Amulet, Berserker’s Jewel, and 30 points in Marksmanship, but no major traits or runes. Before recording results, I allowed all conditions to wear off the targets, which were the tight group of heavy target golems (armor=2597) in the Heart of the Mists.

Each attack from the axe #1 skill Ricochet was dealing precisely 52 damage (non-crit) to each target it hit.

This corroborates my previous testing, which found that direct weapon damage uses the formula:
D = P×W×C/A
Where D is the damage done, P is power, W is weapon strength, C is a skill-specific coefficient, and A is the target’s armor.

Using P=2139, W=127, C=0.500, and A=2597 [heavy target golem]:
D = 2139×127×0.500 / 2597 = 52.3, which is truncated to 52.

Dardamaniac.1295:

Actually ive tested this way myself.I was thinking.When attacks means any of the 5 weapon skills,that includes the bounces that axe make.So 3 bounces mean 35%35%35%=105%.

That’s not the way probability works.

When we have 3 chances for a 35% proc chance, we calculate the chance that all 3 attacks don’t proc:
Probability of no procs = (100%-35%)(100%-35%)(100-35%) = 27.5% chance of no procs

And then take the inverse of that:
Probability of at least 1 proc = 100% – 27.5% = 72.5% chance of at least 1 proc

Can we maybe start by establishing once and for all, with proof, whether the patch did what it said or not? As I understand it, it’s contended by GW2 staff on here that it is what it says – 40 ms.
I’ve seen people claim otherwise. Surely someone can pull a video together with a timer and give us irrefutable proof one way or the other?

I’ve posted my testing results from last week around in a few threads, trying to get the facts out there. Here they are again for good measure.

Summary:

• Crossfire went from .480 s/shot to .520 s/shot, so precisely a 40 ms increase.
• Quickness is reducing Crossfire to .320 s/shot instead of the expected .260 s/shot.

That means either quickness is not giving a full 100% increase to attack speed, or that quickness is only affecting part of the .520 s between shots. More testing is needed to figure out which it is.

No offense, but if your starting data is “16 shots over 4 seconds”, then that doesn’t provide enough precision to then distinguish between a difference of 10ms in your final results.

If you use the methodology of rounding down to the nearest whole number of shots, then the base attack period of Crossfire pre-patch could have been anywhere from 471ms to 500ms and you would have arrived at the same results.

NamelessOne.6815:

Meanwhile in sPvP. I know, 2300 armor is not that much, but 10k damage in one hit?
I have only 1 question: Dishonored or Torchlight 2? Or maybe both?
<Attachment showing 9724 damage taken from 1 hit of eviscerate>

For fun… math!
Here’s how much power you’d need as a ranger in order to crit someone for 9724 using a bow.

Assuming:
— 2300 target armor
— hardest hitting bow skill is longbow #1 skill at 1000+ range: coefficient = 0.900
— exotic level 80 longbow: 1080 max damage
— 50% added crit damage
— 5% longbow damage trait
— 5% damage sigil on longbow
— 25 stacks of might
— 25 stacks of vulnerability on the target

Damage = {(Power + Might) * Weapon * Coef. / Armor} * CritDmg * Bonuses * Vulnerability
9724 = {(Power + 875) * 1080 * 0.900 / 2300} * 2.0 * 1.10 * 1.25
Power = 7492

This is going off testing and calculations I recorded here.

I agree that the shortbow nerf was unwarranted, but this here is terrible logic at play.
The effect of 7% less attacks results in 7% less bleeds.
The amount of stacks you keep up has no bearing on any calculation, the moment you apply the bleed, the damage is guaranteed so as long as your foe doesn’t cleanse it off or you didn’t hit the damage cap.
The loss of a stack is inconsequential, so to speak.

Very much this.

Each bleed stack is its own effect, and does the same damage independent of any other bleed stacks on the target. Attacking 7.7% slower means we do the same damage (direct damage hits as well as condition applications), only 7.7% slower.

When someone says, “I used to be able to get to 8 stacks, but now I can only get to 7; therefore it’s a 12.5% nerf!” what they’re effectively doing is artificially rounding to the nearest 1/8th.

The complaints about quickness may be a legitimate grievance, but I have no pre-patch quickness testing results to offer toward that discussion.

I performed testing on Crossfire this evening (data and methodology available here), and found that its attack period did indeed increase by precisely 40ms, from 480ms to 520ms.

I also found that the attack period is reduced to 320ms under the effect of quickness, which is not quite a 50% reduction. (50% would be 260ms.)

If we assume that there must be an inherent gap between the shots which is constant regardless of attack speed, and that quickness only affects the hidden “cast time” of the attack, then a base cast time of 400ms and gap of 120ms would give the above results.

Just thought I’d help get the facts out there.

Psychrome.9281:

actually i just got done examining the link she posted and it is an awesome way to test it.

Thanks! I definitely appreciate the peer review.

The reason the stopwatch-and-counter method works for the auto-attacks is because they have no cooldown and can be strung together consecutively for as long as needed.

Some comments have stated that it would be virtually impossible to detect a difference of 40ms (0.04s) on one shot, and with that I agree. However, by timing a large number of shots performed consecutively, a potential 0.04s difference is multiplied by the number of shots in the trial. In this case, if the attack period was actually 0.52s instead of 0.48s, then the full 125-shot trial would take 65s instead of 60s.

Similarly, we can work backwards: If we assume my proficiency with the stopwatch is only accurate to within 0.500s (a very conservative estimate), then there’s a resulting uncertainty of 0.004s per shot over the 125-shot trial. In my opinion, this is sufficient for these tests.

Psychrome.9281:

when the guys from mmomechanics actually post there findings about this current change regarding damage done now compared to damage done before then i will be satisfied. same tester with a history of accuracy and a repeatable testing method. that says win.

This is the reason I put all the details about the test method in that post; if someone wants to run the same test, they should be able to do so from the information provided, and they should expect the same results.