Dragoons: A Rotation Reborn

[Ayvar]

So I looked at the rotation in the OP in detail and it has 89% HT uptime (assuming HT does not buff itself, which is a safe-ish assumption based on the "snapshot" behavior for DOTs), 89% DE uptime, 89% CT uptime, and 80% PH uptime.


Another way to write your rotation is:

HT FR PH ID DE CT TT VT FT
HT FR PH TT VT FT ID DE CT
HT FR PH TT VT FT TT VT FT

The HT buff does not time neatly with the ID/DE/CT (herafter referred to as IDC combo or IDC), so you are using HT to anchor the rotation and shifting the IDC to accomodate the timing. The issue is that this full rotation does not overlap with itself. The third line of that rotation has no IDC sequence, because you used it at the "end" of the second line. The problem is that when you repeat this rotation, the first half of the "4th" line will not have the DE debuff or CT DOT, resulting in 89% uptime only.

Assuming we stick to strict-ish 2.5s GCDs (e.g. no/low skill speed), I drafted a rotation based on using IDC as the anchor (because it's the largest and most awkward part of the sequence):

IDC HT ttt PH ttt HT
IDC PH ttt HT ttt PH

This is a 2-line rotation that's anchored by IDC, and alternates HT-PH-HT with PH-HT-PH. This results in 100% uptime for DE, PH, HT, and CT. The "ttt" refers to the True/Vorpal/Full Thrust combo filler. The main difference in my rotation is that the staggered HT/PH and PH/HT allows the HT-PH 20-second timing to interleave with the IDC 30-second timing without conflict. The problem is that the rotation is susceptible to clipping the CT DOT (everything else doesn't matter much) if you have any skill speed because it is exactly 100% uptime (this is a bad thing). The solution to that is some combination of "use Jump before CT" or "use off-GCD buffs before CT" or simply stutter the CT by a half second (make sure CT ticks down to 1.xx before you press the button). The other issue is that with a lot of skill speed the whole rotation's borked because it's pretty tight.

I had considered this rotation before initially posting, but it runs into two main issues that made me decide against it. First let's compare how much extra damage you receive from complete uptime of HT/Disembowel vs. the slightly lower uptime of my own rotation, it's always best to consider potency and I'll use your notation to keep things simple.

Mine:

HFP IDC TTT
170-110/132-187/132-198-242-193/220-181-242-363

HFP TTT IDC
187-121/132-205/132-181-242-363-217-242-193/220

HFP TTT TTT
187-121/132-205/132-181-242-363-181-242-363

The Heavy Thrust buff falls off prior to its reapplication, as does the Disembowel prior to its own, which has been accounted for in the potency numbers. You also have to add on an extra two moves (so 29 moves total) for coming back to the beginning of the rotation as the initial Heavy Thrust and Fracture are now buffed by Disembowel, so +187-121/132 . Now the same treatment for yours:

Yours:

IDC H TTT P TTT H
180-220-176/200-187-181-242-363-205/132-181-242-363-205

IDC P TTT H TTT P
217-266-193/220-205/132-181-242-363-205-181-242-363-205/132

And then +217-266-193/220-205 for coming back to the start as these moves are now buffed by the end of the previous rotation (so 28 moves total). In total potency per second mine produces 109.1 potency/s (7594/69.6s) and yours produces 111.9 (7525/67.2).

So the potency is extremely close. This resolves our first issue, which is that you don't actually obtain any more damage by sacrificing Fracture for increased buff uptime. You could argue for increased autoattack damage in the very brief downtime of the buffs, but I doubt this would account for much extra damage.

The second issue is that the first rotation benefits from (1) boss downtime (such as moving out of AE) since you have more DoT damage ticking (MORE DOTS), and (2) it gets significantly better as we acquire more skill speed. As you remarked, the second rotation gets worse with added skill speed (which is inevitable given the future of the game), and since there are only two DoTs being applied instead of three, it starts falling behind as soon as you spend any time away from whatever you are hitting.

However, it deeply warms my heart to see people taking the dragoon moveset and trying to optimize it in ways I may not have thought of. I've been really pleased by how this thread has progressed (compared to say the "lol who is best dps" thread) and I've referred back to everyone's comments continually for useful information. So thank you all, let's keep doing the job that we love (no matter where we might fall on some unknown parse) and optimizing it in every way that we can!

[EasymodeX]

I see you fixed some of it. You should assume the looped buffs to make a simpler evaluation. The way you're compensating for it skews your buff uptime upwards inaccurately and also frontloads the FR buff, basically double counting / skewing it. Also, the last PH in the second line of my rotation is also wrong and should be /132, not /220.

In any case, your rotation is 106.8 potency per second. Mine is 109.5 pps (difference of 2.7, or 2.5%).

In addition, that is just for abilities, not autoattacks.

Adding in an assumed 35 PPS baseline for AAs, the two rotations are: 151.9 and 148.3 (diff of 3.6, or 2.4%). ::surprise:: I guess the autoattacks didn't favor full buff uptime as I had thought. Wonder why. Sidenote: they must favor the buff uptime less than special attacks do, and just up the baseline. Lulz.

That aside, we're not talking major differences.


The real analysis will be how sensitive the "IDC" rotation is compared to the "HT" rotation to skill speed buffs (there are ways to compensate for the CT clipping in the IDC rotation but it is an issue) next to the alternative cost of having Fracture selected.

There's also the general speculation of how complex it is to mentally maintain an "IDC is rotating" sequence versus a "PH and HT are alternating" sequence. I think the latter lends itself to a priority approach better:

1. If <5s is left on DE, use IDC.
2. If <3s is left on HT or PH, use that ability.
3. Finish TTT.

In fact it's probably easiest to perceive the sequence as "H -> IDC" alternating with "P -> IDC" and just refreshing and filling based on the 3 rules above.

That aside, the basic advantage to dropping FR is that you get another defensive cooldown to pop. That may or may not be worth much (and is not worth anything for DPS).

[Ayvar]

I see you fixed some of it. You should assume the looped buffs to make a simpler evaluation. The way you're compensating for it skews your buff uptime upwards inaccurately and also frontloads the FR buff, basically double counting / skewing it. Also, the last PH in the second line of my rotation is also wrong and should be /132, not /220.

Actually the reason I do not assume auto looped buffs is these rotations will primarily be formed with buffs down when they begin. Take Ifrit or Gaurda for example, you'll get through the initial rotation, then they teleport or you have to go kill adds and the buffs come off. Assuming buffs are already running doesn't take into account the reality of the rotations. In addition, my method of accounting for non-looped buff uptime works just fine. Essentially I am saying in a perfect rotation, how much potency will my rotation have created by the time you've executed 29 moves, and how much does your create by the time you've executed 28 moves, then converts that to potency per second. I also "frontloaded" the Chaos Thrust DoT in your rotation in the same way I did the Fracture DoT in my own, so that isn't an issue.

Factoring in the slight mistake on the Phlebotomize in your rotation that brings the comparison out to 109.1 for the first and 108.9 for the second.

EDIT: I took into account your possible meaning that I skewed the total % buff uptimes. With the way I have done it the Heavy Thrust buff is missing on 4/29 moves or has 87% uptime, and Disembowel is missing on 6/29 moves or 80% uptime, so it actually errs on the side of caution. With yours you are missing Heavy Thrust on 4/28 (86%) and Disembowel on 2/28 (93%)

[EasymodeX]

Ok so this is getting kind of annoying.

1. The Disembowel in the second line of the HT rotation should be 266, not 242.

2. All the Disembowels should be 266 even if you're truncating (266.2). You've allocated 264 in several places.

3. Rounding / truncation aside (sidenote: you truncated everything), the sum for the IDC rotation that you list is incorrect: based on the numbers you posted using your adjustment methodology, the total should be 7521, not 7316. You fat-fingered the sum somewhere in there. I'm not truncating / rounding in my calculations (except at the end), and the total I'm getting is 7534.

4. [Edit] Oh, you are using 2.4s GCDs. That changes the behavior of both rotations and improves your uptimes. I had actually done a second analysis and shifting from a 2.5s GCD to a 2.4s GCD (roughly 4.2% skill speed) changes the DPS loss for your rotation from 2.4% to ~1.1%.

5. Your adjustment method is arbitrary and assumes disengagements at random times. There may or may not be merit to it but that is a simulation consideration, not a sustained DPS consideration.


Overall I think it's more worth using the HT rotation because it has more stable scaling with skill speed, although the alternative skill slot is still a consideration and it's still ~1% lower in DPS.

As a sidenote: the next (1) part to assess is the performance of both when TP has run dry. Specifically, the HT rotation burns about 2.4% more TP than the IDC rotation (because Fracture is slightly inefficient).

In a TP starvation context, it would probably be mostly similar: both approaches would maintain HT and IDC and then do nothing except thrust x3 combos.

The other next (2) part to assess would be the DPS loss with replacing Fracture with another LNC ability if/when you want to load another defensive ability instead of FR.

[xyaie]

Hmm. Pps is an interesting concept globally. It seems like on paper each class's rotation could be converted to that (adjested for buffs) and you could probably give a good prediction on "who is the best dps"

[Ayvar]

Ok so this is getting kind of annoying.

1. The Disembowel in the second line of the HT rotation should be 266, not 242


2. All the Disembowels should be 266 even if you're truncating (266.2). You've allocated 264 in several places.

3. Rounding / truncation aside (sidenote: you truncated everything), the sum for the IDC rotation that you list is incorrect: based on the numbers you posted using your adjustment methodology, the total should be 7521, not 7316. You fat-fingered the sum somewhere in there. I'm not truncating / rounding in my calculations (except at the end), and the total I'm getting is 7534.

4. [Edit] Oh, you are using 2.4s GCDs. That changes the behavior of both rotations and improves your uptimes. I had actually done a second analysis and shifting from a 2.5s GCD to a 2.4s GCD (roughly 4.2% skill speed) changes the DPS loss for your rotation from 2.4% to ~1.1%.

5. Your adjustment method is arbitrary and assumes disengagements at random times. There may or may not be merit to it but that is a simulation consideration, not a sustained DPS consideration.


Overall I think it's more worth using the HT rotation because it has more stable scaling with skill speed, although the alternative skill slot is still a consideration and it's still ~1% lower in DPS.
As a sidenote: the next (1) part to assess is the performance of both when TP has run dry. Specifically, the HT rotation burns about 2.4% more TP than the IDC rotation (because Fracture is slightly inefficient).

In a TP starvation context, it would probably be mostly similar: both approaches would maintain HT and IDC and then do nothing except thrust x3 combos.

The other next (2) part to assess would be the DPS loss with replacing Fracture with another LNC ability if/when you want to load another defensive ability instead of FR.

1/2) fixed, but actually they were all 264, there was never a Disembowel that was at 242 that should've been 266 (so 7525 total)

2/3) yeah, must screwed it up somewhere, which is annoying because I checked it twice

3) I always round down because we have no idea how the game treats damage decimal places.

5) Simulation calculations will always be more applicable. If I'm going to use a rotation I want something that doesn't assume absolutely perfect conditions, we want to aim for perfection while being as realistic about using these rotations as possible.

Either way I completely agree with you. There isn't enough p/s difference between the rotations to justify the switch away from one that will scale better with time, but I have thoroughly enjoyed the discussion, I'm far from perfect and will definitely make mistakes here and there.

[Brynmor]

Some more data pertaining the discussion at hand;

The HT buff appears to activate ~1.75 seconds after initiating the attack. So when determining rotations/priorities it may be worthwhile to keep in mind that the buff does not apply immediately when the ability is used or at the end of the 2.5 (before Skill Speed) GCD.

Maybe this was obvious, but the HT buff does affect HT so reapplying while the buff still persists (if only barely) may be worthwhile.

As was previously mentioned, I was able to confirm that if you initiate an ability with the HT buff (or any buff) up but the animation/damage finishes after the buff has faded you will still receive the benefit of the buff.

[Ricauer]

Heavy Thrust - Phlebotomize - Impulse Drive - Disembowel - Chaos Thrust
x2 [True Strike - Vorpal Thrust - Full Thrust]
Heavy Thrust - Phlebotomize
[True Strike - Vorpal Thrust - Full Thrust]

Repeat. You'll need to add jumps and buffs between the combos but that's pretty much the perfect rotation. It's not as complicated like you make it seem. I wouldn't suggest taking Fracture because it's not really that good. Much better to run around with Internal Release, Bloodbath, Second Wind, Featherfoot and Mercy Stroke.

[Vies]

3) I always round down because we have no idea how the game treats damage decimal places.

A member of my FC was doing a little bit of number crunching and felt that the game generally rounds down. I think he was figuring out hit or crit %, but usually if a game's programmed to round down on one of those, it rounds down on all combat-related numbers. From what I've noticed, at least.

[Ayvar]

Heavy Thrust - Phlebotomize - Impulse Drive - Disembowel - Chaos Thrust
x2 [True Strike - Vorpal Thrust - Full Thrust]
Heavy Thrust - Phlebotomize
[True Strike - Vorpal Thrust - Full Thrust]

Repeat. You'll need to add jumps and buffs between the combos but that's pretty much the perfect rotation. It's not as complicated like you make it seem. I wouldn't suggest taking Fracture because it's not really that good. Much better to run around with Internal Release, Bloodbath, Second Wind, Featherfoot and Mercy Stroke.

Converting this to our usual notation comes out as:

H P IDC TTT TTT H P TTT

Looking at the pattern of the rotation you should already be able to see that your Heavy Thrust buff falls off by the second True Thrust, meaning Vorpal Thrust and Full Thrust don't get buffed, and you never want to cast an unbuffed Full Thrust considering the huge benefits it gets from combining both Heavy Thrust and Disembowel. Also, by the conclusion of the rotation, when you loop back to the start you overwrite the Heavy Thrust buff and Phlebotomize DoT from the end of the previous rotation before they had time to finish, and your Disembowel buff and Chaos Thrust DoT fall off 4 moves before their reapplication. Converting this to potency we get the following:

170-187/132-198-242-193/220-181-242-363-181-220-330-187-205/132-181-242-363

+187 for looping back around to the start since now the initial HT is buffed from the end of the previous rotation. In total potency this is 4356 over 17 moves (40.8s using a 2.4 GCD) producing 106.8 potency/s. Comparing this to the rotations used by myself or the one suggested by EasyModeX above you suffer a 3-6 potency per second drop, with no possibility of improvement with added spell speed, and less DoT damage when you're not beside the mob. The loss of 3-6p/s may not seem like a lot, but consider that is 900-1800 potency lost over a 5min fight with 100% uptime, and that gap gets worse when you have to move away due to AE or add spawns since there are less DoTs running. The 3p/s gap between EasyModeX's rotation and my own was more acceptable since this closes dramatically with more spell speed, and the added DoT damage helps in downtime, this is not the case with your moveset.

Assuming Dragoons are easy to play is something I see a lot, particularly when people are discussing Monks (note I am not commenting on which one is harder than the other because that is a pointless discussion). However, the assumption of simplicity arises from people's non-optimization of buff/debuff durations. Yes you can use a 15 or 10 or 5 ability rotation over a 24 or 27 length moveset, but to do this you sacrifice a large part of the dps Dragoons can produce.

A member of my FC was doing a little bit of number crunching and felt that the game generally rounds down. I think he was figuring out hit or crit %, but usually if a game's programmed to round down on one of those, it rounds down on all combat-related numbers. From what I've noticed, at least.

This is very useful to know, thank you for the info!