JA Delay Testing (Results)Follow

I'd heard that any JA would add a 1 second (60 delay) pause to your next attack round. This seemed inconsistent with my own experience, so I intended to find the cause of this extra delay.

My theory was that the 1 second delay would simply be doubled when dual wielding, accounting for the longer-than-expected wait. So: I decided to test things.

I did this by first recording 100 normal attack rounds with dual wield, and parsing the timestamps between swings. (The first swing of each round only. The second was disregarded, as it actually occurs simultaneously with the first, and only the delay between rounds is relevant.)

Afterwards, the same test was done, only using one dagger instead of two. The first was done with an Avis/Darksteel Kukri+1, while the second used only an Avis. Ninja subjob both times, with no haste gear or dual wield gear, only standard attack speed with dual wield II.

----------------------------------------------------------------------

Anywho, dual wield results(Time in seconds):



This comes out 572~ seconds for 100 attack rounds, meaning:

5.72 seconds per round, or 343.3~ combined delay.
2.86 seconds per dagger, or 171.6~ regular delay.

Purely mathematically, (201+194) * .85 = 335 delay, or 168~ per dagger. The results are, therefore, accurate enough.



This comes out to 765 seconds for 100 attack rounds(including a step), meaning:

7.65 seconds per round/step, or 459 delay.
3.825 seconds per hand, or 229.5 delay.

Therefore:

Without a step: 343 delay / 171 per hand
With a step: 459 delay / 229 per hand.

120 delay total is added, or 60 delay per hand.

----------------------------------------------------------------

And of course, one handed results:



So: 100 attack rounds takes 337 seconds.

3.37 seconds/round, or delay: 202.2.

A delay 201 dagger was used, so this is consistant and expected.



So: 100 rounds with a step take 525 seconds.

5.25 seconds/round, or delay: 315.

So, +115~ delay, close enough to safely say that 120~ delay is added. (Within the margin of error)

---------------------------------------------------------------------

To conclude: The JA delay is not modified by whether or not you dual wield.

However, testing shows that the delay is in fact 120, or two seconds, rather than one.

Sambas, Waltz, and Flourishes appear to behave the same way (in the few brief checks I did, the attack rounds matched rounds with steps), but I did not test them thoroughly enough to confirm this.

This would imply that, using 4 steps / 2 flourishes per minute, and a samba every 2, one spends 13 seconds 'frozen' due to JA usage. This is roughly equivalent to 27%~ slow, in terms of attack speed. This is without taking waltz, weaponskills, or group one flourishes into account.

The effect would appear more dramatic in a single round when using faster weapons, but over time, it evens out regardless of your delay.

Two steps and reverse flourish come out to 6 seconds spent to gain 26 tp, or 4.33~ tp/second. With full merits, the gain becomes 41, and the tp/second rises to 6.833~ per second. One would have to compare with their own setup to determine at what point of haste, if any, auto-attacking would begin to outpace gaining tp through step/flourish spam.

Unsure: Whether or not haste gear effects this rate in any way. The lack of a change for dual wield suggests otherwise; the delay seems to be independant of all other delay. Still, worth looking into.

Uh, yeah, they do. I've had many a fight where I had to spam Waltz to stay alive, and your swings (and therefore rate of TP regain) will be heavily slowed.

This is true, yes.

Unfortunately DNC cannot spam Waltz quite that quickly, so whenever I need to do that I'm eating the full effects of the delay addition >_>

What are you talking about? Other JAs don't lock you from WS or performing another JA (at least, not any more than dances do). They just stop your autoattack timer from progressing during their animation, which this test supports.

This is an interesting topic since enmity research (Not as much, but quite interesting enough).

One thing I want to confirm:
How do you time those delay? stopwatch + eye? Or do you use fraps and analyze the video recorded like I used to do it with ranged delay.

Another thing is, although it implicitly written/some people can infer from the number, you need to state your SJ level and do you use Suppanomimi or not in this test. Just so people can get the situation more easily.

Interesting... simply interesting because before I quit, there are 2 things I'm not sure:
- Gaining TP from reverse flourish instead of normal hitting. Which one is really faster.
- SAM/DRG high haste with Jump JA and WS spam. I think it's not quite as efficient as people think.

Triple post, but oh well.. different content

I kinda forget if JA have this too. When you spam JA, is there any "You need to wait longer to perform that action" message like when you tried to spam ranged attack? This is a system delay that really messes up the actual JA delay if this happened and you didn't count it out.

Edited, Jan 22nd 2010 8:31pm by VZX

I'd be surprised, because...


So, when you're doing SATA you need /ja "Sneak attack" <me>, /wait 1, and /ja "Trick Attack" <me> to get them combined as soon as time permitted. But I never seen /ja "Chocobo jig" <me>, /wait 1, "/ja "Curing Waltz III" <me> working for me (of course, assuming you are at the level you have JA and have enough TP to perform Curing Waltz)

Edited, Jan 25th 2010 2:22am by VZX

General confirmation testing.

Thf/Dnc vs Savannah Dhalmels in Saruta_S. No Brutal, so only native Triple Attack possible procs.
Azoth (delay 210), single-wielded.
0% Haste

Captured fight with FRAPS, then analyzed in VirtualDub.

Frame chosen as "the" time that the hit occurred was the first frame where the HP changed at each hit. That also happens to be the same frame where TP changes in almost all cases.

Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
1           95            0         0 
2           203           108       216 
3           302 (triple)  99        198 
4           410           108       216 
5           523           113       226 
6           625           102       204 
7           732           107       214 
 >> TA 
8           904           172       344 
 >> SA 
9           1076          172       344 
10          1175          99        198

Average normal: 105.1 210.2 ~= 210 delay on weapon

That's sufficient to get an average reading on normal swings. Going to do another fight running through all the JAs I can.

Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
1           76 (triple)   0          
2           187           111       222 
 >> SA 
3           359           172       344 
4           466           107       214 
 >> TA 
5           628           162       324 
 >> Feint 
6           797           169       338 
7           906           109       218 
 >> Mug 
8           1080          174       348 
 >> Steal 
9           1249          169       338 
 >> Spectral Jig 
10          1420          171       342 
11          1529          109       218 
 >> Curing Waltz 
12          1688          159       318

Average with JA: 168.9 frames ~= 337.8 delay

Average additional delay: 127.6.


Any of 128, 125 and 120 delay could conceivably work, but most likely (and easiest to work with) just 120 ~= 2 second delay.


Since 1 second delay is sufficient to stack JAs together (EG: SA + TA), the melee swing delay is separate from the JA restriction delay.

Testing weaponskill.

Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
1           83            0 
2           180           97        194 
>> DE 
3           355           175       350

It would appear that weaponskills add the same 2 second delay.

And with Steps:

Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
2           327            
>> Quick 
3           502           175       350 
 
n           933 
>> Box 
n+1         1093          160       320 
 
q           1524 
>> Quick 
q+1         1692          168       336

Again, roughly 2 seconds.


Stacked JAs need more testing:

Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
n           205 
>> SA + TA 
n+1         389           184       368 
 
 
Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
n           168 
>> SA + TA + Feint 
n+1         400           232       464

Rough first estimate is that stacking two JAs doesn't add any additional time (or a very small amount), but adding a third JA adds another 2 second delay to the melee.


Edited, Jan 25th 2010 6:25am by Kinematics

Good data and test. Waiting the choco jig one. I feel that one JA is really special. I feel other JAs have those normal 120 delay.

Is the same delay for JA also added to spell casts, beyond thier normal casting time?

Was wondering this mainly for Blue mage spells, where it is listed as 0.5sec, but in practice would seem more like 2 sec.(maybe its 2.5?)

More JA stacking testing:

Hit         Frame         dFrame    "Delay" (@ 1 frame = 2 delay) 
 
~~~~ 2 JAs 
 
n           331 
>> SA + TA 
n+1         538           207       414 [+204 delay] 
 
n           195 
>> SA + TA 
n+1         405           210       420 [+210 delay] 
 
n           238 
>> SA + TA 
n+1         439           201       402 [+192 delay] 
 
~~~ w/WS 
 
n           174 
>> Weaponskill: SA + VB 
n+1         384           210       420 [+210 delay] 
 
~~~~ 3 JAs 
 
n           356 
>> SA + TA + Spectral Jig 
n+1         591           235       470 [+260 delay] 
 
n           158 
>> SA + TA + Spectral Jig 
n+1         384           226       452 [+242 delay] 
 
n           168 
>> SA + TA + Spectral Jig 
n+1         402           234       468 [+258 delay] 
 
~~~ w/WS 
 
n           194 
>> SA + TA + VB 
n+1         422           228       456 [+246 delay] 
 
~~~~ 4 JA 
 
n           153 
>> SA + TA + Spectral Jig + Curing Waltz 
n+1         459           306       612 [+402 delay] 
 
n           151 
>> SA + TA + Spectral Jig + Curing Waltz 
n+1         429           278       556 [+346 delay] 
 
n           270 
>> SA + TA + Spectral Jig + Curing Waltz 
n+1         575           305       610 [+400 delay]

Estimated summary:

        Delay     dDelay    probable dDelay 
Base    210       0 
1JA     340       +130      +120 +10 
2JA     410       +200      +180 +10 
3JA     460       +250      +240 +10 
4JA     550:610   +340:+400 +300:+360 +10 +30

So it looks like the preliminary guess of 2 JAs merging was incorrect. Instead it looks like there's a delay of 2 seconds for the first JA, then another +1 second for every additional JA, with an extra +10 delay added in in general (though that might just be an artifact of the capturing process/frame rates/ping delay/etc).

The 4 JA results are interesting, though. Going to run a few more 2 JA test to test a theory. The above tests for 2 JA used a single Sneak Attack; /wait 1; Trick Attack macro. The 4JA test had to be split across two macros. I'm going to split up the 2 JA macro and do them manually from two macros.

n           269 
>> SA + TA (fairly long delay between) 
n+1         502           233       466 [+256 delay] 
 
n           180 
>> SA + TA (fairly long delay between) 
n+1         414           234       468 [+258 delay]

So two JAs with a fairly long wait between them act as two separate +2 second delays, for a total of 4 seconds.

n           169 
>> SA + TA (short delay between) 
n+1         380           211       422 [+212 delay] 
 
n           298 
>> SA + TA (short delay between) 
n+1         498           200       400 [+190 delay]

And two JAs with minimal delay between them stack, and the second one only has a 1 second additional delay added to the melee time. Note that I had to hit the second JA almost immediately to get them to stack like this. If most of the SA animation was complete before I hit TA, I got the 4 second melee delay.

It's probable that in the 4 JA tests above, I just wasn't getting the timing of the last JA down fast enough to get it to stack, except for the middle test. Even so, it seems that the fourth JA adds another innate half-second (30 delay) delay on top of the other stuff.

There was a discussion about this jig thing a while back: the conclusion is that while normal ability position lock could be disabled with gear changes (COR soloing anyone) they especially locked the DNC movement on JA (probably to avoid using chocobo jig in a kite/holding tool)

TP gain with Reverse Flourish.


Given the above testing, as well as comments about this issue, I thought I'd try to work through the math.


First sample, my own dnc.

Level: 60
Last party: vs Defoliators & Puktraps in Aydeewa Subterrane.
Melee accuracy: 92.1%
Debuff accuracy:
- Box Step: 94.4%
- Quickstep: 90.6%

[Note: I always use Quickstep as the first step in a fight, to make later steps easier to land. That would be why Box Step has a higher success rate. I'll use 92% as the average step land rate.]

Weapons:
Avis (d23, delay 201)
Bone Knife +1 (d18, delay 190)

Dual Wield 2
Delay per round: (201+190)*.85 = 333
TP Per hit: 4.8

Store TP: +5 (Rajas)
Total TP per hit: 5.0

Haste: Spell + Samba + March = ~29%

Delay per round: 237

Average TP per round: 5.0 * 2 * 92% = 9.2

Average TP per second (melee only): 9.2 / (237/60) = 2.33


Time per step: 2 seconds

Assuming 1 Violent Flourish used per 3 steps so that you end up with 5 Finishing Moves per Reverse Flourish:

Average number of steps needed: 3 / 92% = 3.26

Time spent: 3.26 * 2 seconds (steps) + 2 seconds (RF) = 8.52 seconds.

Melee TP sacrificed (time): 8.52 * 2.33 = 19.85
Melee TP sacrificed (steps): 30

Total TP cost: 49.85

TP return: 60

Net gain: 10.15 TP over ~50 seconds.

TP gain rate: 0.20 TP/sec, or about 12 TP per minute.

~~

If using two steps only (not using FMs for any additional Flourishes):

Average number of steps needed: 2 / 92% = 2.17

Time spent: 2.17 * 2 seconds (steps) + 2 seconds (RF) = 6.34 seconds.

Melee TP sacrificed (time): 6.34 * 2.33 = 14.77
Melee TP sacrificed (steps): 20

Total TP cost: 34.77

TP return: 46

Net gain: 11.23 TP over ~35 seconds.

TP gain rate: 0.321 TP/sec, or about 20 TP per minute.


~~----------------------------------------~~

Second sample at level 75, only modifying haste as a primary factor.


5 HS merits
DW @80%
15% haste in gear
Azoth + BK+1
~assuming same accuracy

Delay per round: (210+201)*.8 = 329
TP Per hit: 4.8

Store TP: +5 (Rajas) +1 (Brutal)
Total TP per hit: 5.0

Haste: Spell + Samba + 2x March + gear = ~60%

Delay per round: 131

Average TP per round: 5.0 * 2.1 * 92% = 9.66

Average TP per second (melee only): 9.66 / (131/60) = 4.42


Time per step: 2 seconds

Assuming 1 Violent Flourish used per 3 steps so that you end up with 5 Finishing Moves per Reverse Flourish:

Average number of steps needed: 3 / 92% = 3.26

Time spent: 3.26 * 2 seconds (steps) + 2 seconds (RF) = 8.52 seconds.

Melee TP sacrificed (time): 8.52 * 4.42 = 37.66
Melee TP sacrificed (steps): 30

Total TP cost: 67.66

TP return: 60

Net gain: -7.66 TP per ~50 seconds.

TP gain rate: a loss of .153 TP per second, or 9 TP per minute.

~~

If using two steps only (not using FMs for any additional Flourishes):

Average number of steps needed: 2 / 92% = 2.17

Time spent: 2.17 * 2 seconds (steps) + 2 seconds (RF) = 6.34 seconds.

Melee TP sacrificed (time): 6.34 * 4.42 = 28.02
Melee TP sacrificed (steps): 20

Total TP cost: 48.02

TP return: 46

Net gain: -2.02 TP over ~35 seconds.

TP gain rate: a loss of 0.058 TP/sec, or about 3.5 TP per minute.


~~----------------------------------------~~

Second sample at level 75, adding in Reverse Flourish merits as well.


TP return with 5 FMs: 75
TP cost for 3 steps + flourish: 67.66

Net gain: .147 TP per second, or 9 TP per minute.


TP return with 4 FMs: 61
TP cost for 2 steps: 48.02

Net gain: .371 TP per second, or 22 TP per minute.



~~-----------------------

Without RF merits, and assuming 90% melee accuracy and 95% step accuracy, the breakeven point on haste is

2 steps:
26 TP per 6.2 seconds
at 5 * 2.1 * .9 = 9.45 TP per round
= 2.75 rounds in 6.2 seconds
= 2.25 seconds per round
= 135 delay
= 59% haste (assuming 330 delay after dual wield)

3 steps:
30 TP per 8.3 seconds
at 5 * 2.1 * .9 = 9.45 TP per round
= 3.17 rounds in 8.3 seconds
= 2.62 seconds per round
= 157 delay
= 52% haste (assuming 330 delay after dual wield)


So once you're in the 50%-60% haste range, Reverse Flourish is just barely keeping up with the TP lost from the melee delay. At lower haste values it comes out ahead, and once you add RF merits it pushes the haste threshold way up, and you generally come out ahead regardless.

A higher differential between melee accuracy and step accuracy (including the same spread but at a lower base accuracy) would also favor RF more.

I vaguely seem to recall that you can use a wait .5 instead of wait 1 in between macro lines, not sure if this is valid (possibly lower values too).