Alacrity breaks assault rotation problem?!

[KamikazeKommando]

Hey folks, I want to reach out to see if anyone else is experiencing this problem with alacrity interferring with commando assault spec [ionic Accelerator] procs. As a refresher, IA has a 7.5S ICD, when it's up [Charged Bolts] or [Full Auto] finish the cooldown on [Mag Bolt] and make it free. It's a core function in an assault spec rotation where we base our rotations to maximize it's use.

 

Here is a very basic, simple 15s rotation:

 

00.0 GCD01: AP

01.5 GCD02: SB

03.0 GCD03: IR

04.5 GCD04: FA (+IA)

06.0 GCD05:

07.5 GCD06: MB (-IA;+BC)

09.0 GCD07: HS

10.5 GCD08: HS

12.0 GCD09: CB (-BC;+IA)

13.5 GCD10: MB (-IA)

 

Note that IA procced twice at 7.5s intervals. FA procs IA at the top of GCD4, CB procs IA at the top of GCD9 by virtue of it being instantly cast through Blazing Celerity. With 0 alacrity I can execute this rotation endlessly without any problems. With alacrity pulled into the picture, I'm able to cast FA the 2nd time through the rotation before IA's ICD is done, so it won't proc it and the rotation breaks. I can reproduce this at will, on different machines, different ISP networks, etc.

 

I have a video of this posted here:

https://youtu.be/a0Jjpjd-aQ0

 

Here's the short combat log of the problem you see in the video when I equip the alacrity gear and try the rotation:

www.kamikazetank.com/misc/log_alacrity_misalignment.txt

 

It shows that Full Auto at the 4th GCD on the 2nd pass through the rotation did not proc IA, exactly as the video I posted was showing. Line 76 was the last proc of IA triggered by the 1st pass 9th GCD CB. Line 108 is that 2nd Full Auto which should have procced it but didn't. You'll notice how 6.949s elapsed between these two lines and that wasn't enough to allow IA's ICD to refresh.

 

This issue has been occuring for a very long time now, it's nothing that has been newly introduced. Yet, some are reporting that they cannot reproduce this problem. It's annoying because in present form it means alacrity will mess up the rotation. I'd love to know if anyone else can replicate this problem.

Edited October 29, 2015 by KamikazeKommando

[Gyronamics]

Alacrity screwing with assault was demonstrated before 3.0 and is the reason why it launched with a 10% surge buff and not a 3% alacrity buff.

 

But that wasn't a fix, that was just changing the one ability that had the highest chance of screwing up the rotation in an uncontrollable way.

 

It didn't actually change the way that alacrity could mess with internal cooldowns.

 

That's why experienced commandos know that if you pop alacrity buffs like Recharge Cells + Utility or Supercharged Celerity at the wrong time it derails the rotation if you try to keep going.

 

Now there's the incentive to stack very large amounts of alacrity we can also see that high values of passive alacrity can also break the rotation.

 

Somewhere over 650 alacrity will break my rotation so I stop approximately there.

 

There was one very simple suggestion put forward, again before 3.0 launched, which would have copper bottomed the proc rotation and prevented this.

 

That was to lower the proc internal cooldown from 7.5s to 7.0s, too short to get more attacks in but with 0.5s (without any alacrity) of buffer time which would be extremely difficult to break with any realistic value of alacrity.

 

The same fix would still work now.

[Goblin_Lackey]

I want to utilize this thread to try and figure out the why behind the alacrity issue for Assault and IO rather than tie up space on the larger thread.

 

I did a deep dive into your log there to try and figure out the root cause. And I think I have identified what is causing the issue: Blazing Cererity + Charged Bolts.

 

 

Here is the reason why I think that this one move is the issue:

 

I started by looking at the time between Ionic Accelerator Activations and found that the time period between what should have activated them was different:

(Note: I dropped the minutes and hours and focused just on the seconds)

Full Auto 1 @ 39.665

Charged Bolts @ 46.881 (7.216 after last activation)

Full Auto 2 @ 53.830 (6.949 after last activation)

 

Since the 6.949s is a good deal faster than the previous activation and this time duration is under the alacrity modified ICD length, the proc did not activate.

Alacrity % = 7.6733% = 0.30 * (1 - (1 - 0.01/0.3)^(([708]/65)/1.25)) (the 708 is from your video)

ICD Length = 7.5s / (1 + Alacrity %) = 7.50 / 1.076733 = 6.9655s

GCD Length = 1.5s / (1 + Alacrity %) = 1.3931s

 

This lead to the obvious question: why?

I further broke down the Ability activations:

• Full Auto @ 39.665
  
• Mag Bolt @ 42.566
  
• Hammer Shot @ 44.068
  
• Hammer Shot @ 45.472
  
• Charged Bolts @ 46.881
  
• Mag Bolt @ 48.158
  
• Assault Plastique @ 49.560
  
• Serrated Bolt @ 50.969
  
• Incendiary Round @ 52.403
  
• Full Auto @ 53.830
  

 

Here is the same list but now with the "duration" of each move (time for next move - time for current move)

• Full Auto @ 2.901
  
• Mag Bolt @ 1.502
  
• Hammer Shot @ 1.404
  
• Hammer Shot @ 1.409
  
• Charged Bolts @ 1.277
  
• Mag Bolt @ 1.402
  
• Assault Plastique @ 1.409
  
• Serrated Bolt @ 1.434
  
• Incendiary Round @ 1.427
  
• Full Auto
  

 

Most of the moves are activated at just over the GCD length (due to the normal input lag) except for 1 move that is over a tenth a second shorter than the reduced GCD.

 

 

I am forced to conclude that the reason why the rotation is prone to fall apart is because the instant activation of Charged Bolts under Blazing Cererity does not cause a full GCD.

 

I am actually curious if this is true under no alacrity as well or if somehow the code behind it is double counting alacrity in this case.

Edited October 29, 2015 by Goblin_Lackey

[KamikazeKommando]

I am forced to conclude that the reason why the rotation is prone to fall apart is because the instant activation of Charged Bolts under Blazing Cererity does not cause a full GCD.

 

I am actually curious if this is true under no alacrity as well or if somehow the code behind it is double counting alacrity in this case.

 

You may be on to something, that does appear to be what is causing the misalignment. With 0 alacrity this never occurs, here's a parse of 3 iterations of the rotation without any on the gear:

 

http://www.kamikazetank.com/misc/no_alacrity_parse.txt

 

I broke it down already here for you:

 

08.389 Assault Plastique @ 1.609

09.998 Serrated Bolt @ 1.522

11.520 Incendiary Round @ 1.611

13.131 Full Auto @ 3.104

16.235 Mag Bolt @ 1.586

17.821 Hammer Shot @ 1.499

19.320 Hammer Shot @ 1.614

20.934 Charged Bolts @ 1.494

22.428 Mag Bolt @ 1.49

 

23.918 Assault Plastique @ 1.597

25.515 Serrated Bolt @ 1.505

27.020 Incendiary Round @ 1.611

28.631 Full Auto @ 3.198

31.829 Mag Bolt @ 1.615

33.444 Hammer Shot @ 1.493

34.937 Hammer Shot @ 1.51

36.447 Charged Bolts @ 1.597

38.044 Mag Bolt @ 1.488

 

39.532 Assault Plastique @ 1.505

41.037 Serrated Bolt @ 1.624

42.661 Incendiary Round @ 1.595

44.256 Full Auto @ 3.096

47.352 Mag Bolt @ 1.606

48.958 Hammer Shot @ 1.605

50.563 Hammer Shot @ 1.514

52.077 Charged Bolts @ 1.544

53.621 Mag Bolt

 

Seems like a full GCD is utilized when BC procs CB without any alacrity in the picture. The fact that the CB cast is so fast with alacrity does seem to suggest there's a double dip or some bug. I'm surprised you didn't run into this when you tried the rotation with your 8% though.

Edited October 29, 2015 by KamikazeKommando

[Gyronamics]

I specifically get a rotation crash when I do:

 

Tech Override -> instant Charged Bolts

Reserve Powercell -> Full Auto

 

That's to guarantee the RP gets used on the FA as queuing it during the hard cast CB has a horrible chance to be consumed with no effect by the CB cast itself so the FA activates at full cost.

[KamikazeKommando]

Figured I'd post an update after digging into this some more. The fact of the matter is that [blazing Celerity] does indeed reduce the activation of the instant [Charged Bolts] or [serrated Bolt] below the GCD, but only when Alacrity is present. All in all I've experimented with the original 708 in the parses we've analyzed to date, but then went to my original non-alacrity set loaded up with 14 alacrity augments, and started playing with what happens when you go piece by piece 73 alacrity at a time. The ability does double dip into the alacrity, the higher the value of alacrity, the greater the extent of the reduction.

 

Here's the odd bit. The actual level of fixed Alacrity plays a key role in whether your overall alignments will crash or not. That said, if you gain a temporary alacrity buff while the ICD of [ionic Accelerator] is still refreshing, you're going to absolutely crash the rotation, so you will need to be quite careful with [supercharged Celerity] or the [Cell Capacitor] utility.

 

So going from 0 to 14 augments, I went through the simple rotation for a few minutes and observed whether a crash occurred and how hard it was to reproduce. Results were VERY interesting. Seems like there's a magic range where you can get the crashes to occur, it's in the roughly 650-750 area, above or below that I can't get the crash to occur, but in that 708 set it occurs 100% of the time every time. Going a bit above and below shows about a 50% crash rate, then it gets rarer until you're totally out of the window. I'm not quite sure exactly what is causing this magic though. This is very odd indeed. You see in my video the 708, that 2nd Full Auto which should proc IA fails EVERY time. Add or subtract 73 alacrity from that and it aligns EVERY time.

 

So, I'll say this for sure. Blazing Celerity does double dip into alacrity and reduce the ability below the GCD, that is for certain. Yet, whether that apparently plays a major role in getting your rotation to crash will depend greatly on the amount of alacrity that is in the picture. Above all else, dynamic temporary alacrity can and will crash up your rotation if used while the ICD is refreshing.

 

Anyway, hope this additional info helps, it's still a bug in my books, and a very mysterious one at that.

[Gyronamics]

Who feels like betting that a rounding issue happens right there.

[Volcan]

YES! this issue has been driving me nutz. I have to pull my alacrity all the way down to 650 to fix the issue, any higher and the rotation breaks.

[Volcan]

 

There was one very simple suggestion put forward, again before 3.0 launched, which would have copper bottomed the proc rotation and prevented this.

 

That was to lower the proc internal cooldown from 7.5s to 7.0s, too short to get more attacks in but with 0.5s (without any alacrity) of buffer time which would be extremely difficult to break with any realistic value of alacrity.

 

The same fix would still work now.

 

Pre-4.0 this issue was tolerable because we popped the right ammo/heat management abilities at the right times. Now in 4.0 this fix is desperately needed. We should not be forced under 650 alacrity simply because of poor design.

 

If there is an update on this issue please let me know.

[Elite-Force]

still a problem...

also affects power surge (the merc skill that make the next cast instant)

and convection (lightning sorc passive, that utilizes force speed for 2 instant lightning bolts)