I gathered some data on Barrage, attack speed and multiple projectiles. It behaves inconsistently.

TL;DR:

1. Barrage does not scale in a linear fashion with attack speed.
2. Barrage's attack time is not dependent on the number of projectiles fired, in contradiction to apparent developer intent.
3. That said, LMP or GMP each seem randomly capable of inducing a 4-frame delay in the attack time, which remains until you restart the game. I assume this is just client-side, and this is just my best interpretation of what's going on with some confusing data that I don't want to put hours more into figuring out.




In the aftermath of a discussion on reddit where it became clear that no one really understood the mechanics of Barrage, despite the relevance of a detailed understanding of these mechanics to Cast on Crit builds, I decided to do some testing. As far as I am aware, dev insight on Barrage is currently the following, taken from this thread:

"

Mark_GGG wrote:

Adding extra projectiles to barrage will make it take longer than your attack time, as each one takes time to fire.

"

Mark_GGG wrote:

The reason that can't happen is that Barrage (like other attack skills) uses the attack time of the attack to know how long to take - the initial "windup" phase takes 40% (from memory, might be slightly wrong) of the attack time, and each projectile takes a set percentage of the attack time for itself.
Since Barrage's actual total time taken is necessarily a function of the attack time value, it's not possible to have it match in the lots of projectiles case.
Long term, we'd ideally be able to introduce a separate stat for barrage to show how long the full barrage will take, but for now the actual stat we have is the attack time stat of the skill, which the skill uses to time the animations.

Multiple players have performed testing intended to determine the most desirable attack speed for CoC builds in order to avoid the 50 ms cooldown, and all appear to agree on an attack speed of about 2.25. However, as far as I know no one has examined how Barrage's attack speed is actually calculated in the general case.

As such, I recorded myself attacking with Barrage using various tooltip attack speeds, and determined the real attack speed by counting the frames that elapsed in between each use of the skill (detailed methodology in the spoiler below). The major original intent of this was to figure out what percentage of the tooltip attack time was taken up by the windup animation and what percentage was used to fire each projectile, but as you will see, this quickly turned out to be tricky. Full data is available from the link below, but here's a summary (apologies for the crappy formatting, I've sunk enough time into this already):





Basically, the relationship between Barrage, base attack speed, extra projectiles and real attack time seems to be inconsistent. At some attack speeds, LMP is slower while GMP is just as fast as unsupported Barrage. At some, it's the reverse. This is in direct contradiction to the GGG responses quoted above, so may well be a bug. Furthermore, attack speed gains are unpredictable.

I am aware that this is not (yet) an enormously comprehensive data set, but given the accuracy of my controls, I'm pretty sure there's a real effect here. Anyone have any clue what's going on? I can't figure it out.

EDIT: I ran a lot more tests this evening. It's too late for me to want to present them all as anything more than the TL;DR above.


Raw data (.xlsx)

Source videos are available on request, but the folder is 18.1 GB, so be warned.

1800 view and no replies, clearly a very interesting
topic to the community.. but a challenging one.

look forward to what comes outta this.

"

AgnosiousD wrote:

1800 view and no replies, clearly a very interesting
topic to the community.. but a challenging one.

look forward to what comes outta this.

It's certainly interesting, title was a bit click-baity but OP delivered. However, I think the real reason no one is responding is that no one cares. Not a single person is going to seriously consider using LMP or GMP with barrage. Even if Mark is dead wrong, and Multiple Projectiles strictly increases firing speed so that the total attack time is unchanged, GMP only breaks even on total damage (projectiles get doubled, damage gets halved), and LMP dps barely goes up (arrows multiplied by 1.5, damage by .7, total is 1.05x damage).

There are obviously *way* better support gems for Barrage, no matter how LMP and GMP interact with it.

But hey, it's always cool to prove a dev wrong, good luck!

"

However, I think the real reason no one is responding is that no one cares. Not a single person is going to seriously consider using LMP or GMP with barrage. Even if Mark is dead wrong, and Multiple Projectiles strictly increases firing speed so that the total attack time is unchanged, GMP only breaks even on total damage (projectiles get doubled, damage gets halved), and LMP dps barely goes up (arrows multiplied by 1.5, damage by .7, total is 1.05x damage).

There are obviously *way* better support gems for Barrage, no matter how LMP and GMP interact with it.

Cast on crit with barrage (arguably the best way to use cast on crit) cares.

"

Kaezin wrote:

Cast on crit with barrage (arguably the best way to use cast on crit) cares.

Oh, you're right, I'm an idiot who failed to read the first paragraph :)

To be fair, most of the views will be from redditors, not directly from this forum.

Yeah, I apologise for the click-baity title but I was writing this at 8am half-asleep and couldn't think of what to write that would fit in the 100-character limit. I'll fix it.

Also, I'm going to be doing some replicates of the conditions I've already tried when I get home in order to confirm a few things. I'll post up the results later.

Could this have anything to do with why Barrage causes so much desync? It doesn't appear to happen for everyone but for me and plenty of others (judging by reddit) it's constant.

It's like the server and client are calculating how long the skill and/or its animation should last differently. As a result you start moving again on your client at a different time than on the server. Desync. Same thing goes for Lightning Tendrils.

Did you consider that the game might only run at 60fps (common) or 30fps (far less) common despite your displayed fps? Also if your frames drop below that you cant cast/attack faster than your own fps.

"

eps1lon wrote:

Did you consider that the game might only run at 60fps (common) or 30fps (far less) common despite your displayed fps? Also if your frames drop below that you cant cast/attack faster than your own fps.

As I understand it, Fraps with frame lock enabled forces both your game and your recording to run at the intended speed (if your hardware is capable). In testing of this feature, the in-game FPS (according to the F1 display) is not particularly stable, but it never drops that much, so this shouldn't matter too much aside from possibly causing some things to show up late.

I have been doing some testing at 120 FPS and found it to actually be slightly less accurate, which I can't explain. It may be something technical which I have no understanding of, or simply something I've overlooked. However, the most important point as far as I am concerned is that all controls so far have given results that are at most 0.01 of a second off the predicted attack time, which suggests to me that this is a satisfactorily accurate method.

Thanks for your effort i am looking forward to a Dev response.

I am tempted to test barrage against a moving character in a PVP duel.
With zealoth and EB we could gain some new insights rofl.

Or battle rouse/Ondar's Flight with Shavronne's Revelation.