[UPDATED] Barrage/CoC Mechanics: There Is No Optimal Attack Speed (but high attack speed is okay)

UPDATED TL; DR: If these data are accurate, there is no perfect attack speed for CoC Barrage. Average proc number per skill use decreases smoothly with attack speed, rather than dropping off suddenly. Proc rate per second does increase with attack speed, but not significantly until a high attack speed is reached. The current goal of ~2 APS seems to be just about the worst possible attack speed range.



UPDATE

Following on from all the below, I've run some more tests at higher attack speeds to start looking at optimisation as opposed to just mechanics. Here you go:





So this is interesting. While again these numbers are not perfect (sample sizes of about 130 crits, ideally you'd want something like ten times that), it seems that 2.0 is close to the worst possible attack speed. After that, your proc rate actually starts to scale better with attack speed (from 2 to 3 APS is a ~20% proc rate increase, from 3 to 4 is about 25%). My assumption is that this is due to you essentially having lots of "free attempts" with a high attack speed - you still get just as many chances to proc as a lower attack speed build, just not in a row, so the numbers you get are skewed to the high side of the theoretical maximum.

So the question now is, do you bother with attack speed? It does not seem to be an easy question to answer. 3 APS gets about 20% more procs than 1.5 APS, but obviously requires a lot of effort to reach, and this could alternatively be channeled into survivability or spell damage. Thoughts?





ORIGINAL POST

The Story So Far

Barrage is a strange beast. Its unusual mechanics and importance for CoC builds have made it the subject of much consternation, especially since the increase in CoC's cooldown to 50 ms. My previous thread has some detail on this and GGG information on it (including a post on client-side issues with how Barrage is displayed here which rendered my methods more or less moot), but in short, there's long been the assumption that there is an optimal attack speed for Barrage, that being the maximum attack speed that does not cause individual projectiles to land within 50 ms of each other. However, there has been significant confusion on exactly what this value is.

Current Testing

I attempted another round of testing similar to that which I've linked above with the release of 2.0, hoping that lockstep might help. It didn't appear to, so I gave up again for a while. However, following continuing discussion on reddit I decided to try a different approach, measuring proc rate directly rather than trying to investigate attack times. This should bypass any client-side display issues.

In order to do this, I tested a few methods. The one I hit upon is detailed at length in the spoiler below, but to put it briefly, I set up a character tanky enough to ignore my Barrage damage, wailed on him in PvP, and used CoC SRS to count procs. I have not yet tested the full range of conditions (currently I have only tested GMP-linked Barage, and not at a full range of attack speeds), and my dataset is not really large enough to have much confidence in the quantitative data (I haven't tested for significance and really cannot be bothered to do so). However, I think this is enough to draw some preliminary conclusions. Here are the main data summaries of interest, in my opinion:





Again, don't read too much into the exact numbers just yet. Here are the main points to draw from this.

• There is no threshold at which you drop from fully proccing Barrage to proccing it half as often. Proc rate per crit drops off smoothly. This is directly contrary to established wisdom on Barrage CoC.
• In order to achieve maximum proc rate (theoretical 5.28 procs per crit with the level 17 CoC used, which eerily enough exactly matches my number), or to have a single instance of one use of Barrage proccing CoC 8 times, the literal minimum possible attack speed had to be attacined. As far as I am aware, 1.08 is as slow as you can possibly attack with Barrage, requiring a 1.2 APS weapon (no wand or bow is slower as of 2.0), The Anvil, and no sources of increased attack speed.
• While attack speed does have a net positive effect on proc rate per second, it is very small. Increasing attack speed by 70% increased proc rate per second by ~10%. Again, don't take these numbers to be exact, but simply an indication of a trend. Note also that the maximum attack speed in this dataset is still very slow. I will hopefully have a chance to test more realistic speeds soon, but I imagine the usefulness of attack speed will drop off even further.

I'd love it if other people could take a look at the data and see whether I've done something stupid.

Also, I'm wondering how this occurs. As far as I can think, the only way this could happen is due to some sort of rounding or RNG meaning that times between projectiles are inconsistent.

My thanks to SnoopythegoD for his services as a pincushion.

Raw data (.xlsx) (I've recorded all my experimentation, but the folder is 161 GB so you'd better give me a good reason to compress and upload it somewhere if you want it!)

I don't understand how to read you chart.

1.82 APS = each projectile lands 68ms apart? If that's the case, you haven't passed the 50ms threshold yet.

Your procs/second line goes up and down, I don't think you can make your third conclusion from it.

I am at about 4 aps with barrage and feel a damn lot more procs than with 3.. How does that come then?

I believe the attack speeds you used in your tests are too low to hit a plateau. You need to reach >20 projectiles per second during the burst phase of barrage to hit a plateau. Yes, higher attack speed will still yield more procs, but the function should look something like the following: (sorry for the graph that looks like ass)

Also note, that the proc rate will never surpass 20/s. The red line only slowly approaches it.

I believe you didn't hit the first breakpoint with your attack speed. Could you test with higher APS?

"

1.82 APS = each projectile lands 68ms apart

Incorrect, because you are not actually accounting for the windup time.

The 8 projectiles are not evenly spread out through the entire attack time.

He's already hit the point where a breakpoint *should* occur.

From earlier testing in the shatterchuck build thread from right after the nerf I remember the first break point being at ~2.25aps. At that point a 50% srs proc rate was observed.

Thanks for this.

I didn't do an organized experiment like you did but I have experimented with various attack speeds and I can also see the smooth transition of the top possible casts per barrage. This is a bit strange, I am not sure why it happens! Could it be related to chance to cast?

In any case though, I have read in the thread in which Mark has given the information about barrage that the attack time of each barrage may increase if you add greater multiple projectiles etc.

However I have tested this and I don't see this being this way: I have tried this with a slow setup to be able to count the barrages easier. I have then shoot it for a minute and counted how many time a barrage has happened. Whether having greater or not (so either having 4 or 8 projectiles) I got exactly the same occurrences and when I did the math it matched what I saw in the tooltip.

Can anyone else confirm this?

I am wondering if barrage/cast on critical was changed after the awakening or something? Could it be that there was a 2+ cap before, but because it was not fair they have changed how it worked?

In any case I think that based on the findings of the testing viperesque did, there is no cap, or rather there is a cap but you need to have so slow attack speed to overcome it, it's simply not worth it. So in the end stacking attack speed may be a good strategy after all.

interesting data, cheers.


i feel like id still use an engraved wand purely because of the animation lock time and what that means for kiting/stutter stepping etc. Only time I ever died on my main coc barrage build was against turbo palace dominus, he ran up to me and I fired 1 barrage at him, with the plan being to fire once at point blank then run out of slam range... he managed to run up to me and touch slam me before I could get out of that 1 barrage animation and run away even though i started it before he was even in slam range. That was with a 1.5 aps wand.

I think it procs much more if u got high aspd. Im currently using this gears on my coc. Every time i pop up vaal haste i do crit and proc alot more. I do have around 2.5 apsd

Okay, I have done some more testing with attack speeds between 2 and 4 and updated the first post. Seems like the people above who are saying they get more procs at very high attack speeds aren't wrong! This is even more complicated than it first seemed in terms of implications for build optimisation, but I'm fairly sure this does mean that everyone currently aiming for roughly 2 APS (I believe the "breakpoint" was considered to be 2.05?) is basically doing the worst thing possible.

Again, I'd love for people to take a look at my methods and data and tell me whether it looks sound or not. I can't rule out making some error of logic here. Even better, do your own tests! :D