Forum Settings
       
Reply To Thread

"Silly" Cast Rotation ConceptFollow

#1 Sep 27 2013 at 10:45 AM Rating: Excellent
Scholar
44 posts
I'm at work - and obviously totally focused on my work - so I can't test this. I figured I would post to get my thoughts down and see if anyone could poke some holes in the theory.

Blizzard 3, Blizzard 3, Fire 3, Fire 3. Repeat.

Since AF/UI don't apply to you until the triggering spell hits your target, both casts of Blizzard (respective Fire) will be cast with the effects of Astral Fire (resp. Umbral Ice) active. The second cast of each spell, however, will impact the target after your buffs change to the opposite set. In other words, this sequence of events occurs (initially, assume AF3 is up):
  1. You start casting your first Blizzard 3. Since AF3 is up, it costs 80 mana and has a 1.75 second cast time.
  2. Cast finishes and the projectile leaves your hands. You start casting the second Blizzard 3. AF3 is still up so it also costs 80 mana and has a 1.75 second cast time.
  3. The first Blizzard 3 projectile impacts the target, switching your buff to UI3.
  4. Second Blizzard 3 cast completes and the projectile leaves your hands. You start casting the first Fire 3. UI3 is up, so it also costs 80 mana and has a 1.75 second cast time.
  5. Second Blizzard 3 impacts the target and refreshes UI3.
  6. First Fire 3 cast completes. You queue up the second Fire 3 cast, still under the effects of UI3.
  7. First Fire 3 impacts the target, switches your buff to AF3.
  8. Second Fire 3 cast completes, you begin casting the first Blizzard 3 of the next cycle.
  9. Second Fire 3 projectile impacts target, gets full damage buff from AF3.

The cycle is four casts long, each of them 1.75 seconds due to the cast speed increase from the opposite AF/UI buff. The Blizzard 3s do 220 potency each, the first Fire 3 220*70% = 154 potency (30% damage penalty from UI3), and the second Fire 3 220*180% = 396 potency (80% damage bonus from AF3). That's (2*220+154+396)/(4*1.75) = 141.43 potency per second. The complete cycle costs 4*80 = 320 MP, and you spend 3.5 seconds each cycle in UI3 so you'll get a mana tick. Effectively, mana will never dip below 90%.

Contrast with the typical rotation: B3,T2,F3,F1,F1,F1,F1,F1 with T3 on cloud procs and F3 on Firestarter. Ignoring procs, cycle length is 1.75+3+1.75+1.25+4*2.5 = 17.75 seconds with 220+(50+6*35)+220*70%+150*180%*5 = 1984 potency. An average cycle will have 6*5% = 0.3 cloud procs that add (60+8*35)= 340 potency and 2.5 seconds of cast time. We expect 5*40% = 2 Firestarter procs, each will add 220*180% = 396 potency and 2.5 seconds. Altogether, that makes for an expected 1984+0.3*340+2*396 = 2878 potency over 17.75+0.3*2.5+2*2.5 = 23.5 seconds, resulting in 122.47 potency per second.

What about B3,B3,F3,F3,F1,F1,...,F1, i.e, augmenting the rotation with some casts of Fire 1 w/ Firestarter during the AF phase? With AF3, a Fire 1 does 150*180% = 270 potency over 2.5 seconds, with a 40% chance to proc Firestarter for an additional 220*180% = 396 potency over 2.5 seconds: expected DPS is therefore (270+40%*396)/(2.5+40%*2.5) = 122.4 potency per second. That's less than the proposed rotation's expected DPS of 141.43 potency per second - casting more Fire 1s would *lower* the DPS.

Stuff to test:
  • Is it true that spellcasts get the mana cost / cast speed effects from the active AF/UI stack at beginning of cast, and the damage multiplier effects from the active AF/UI stack at time of projectile impact?
  • Does the cast speed reduction effect apply properly to the GCD as well?
  • Do damage parsers reflect the theoretical relationship between the two cycles, or is something missing from the theory?



Edited, Oct 2nd 2013 1:55pm by cartec
____________________________
Cross-Class Skill Sheet
Kawaii rate me down
#2 Sep 27 2013 at 2:34 PM Rating: Excellent
Guru
Avatar
**
362 posts
I'll have to look at this again later when I get home, but my first thought is that in your proposed rotation you're not accounting for the damage penalty of the first Blizzard III while in AF3.

What's funny is you get the damage penalty, but the damage per cast time is actually much smaller of a penalty because the cast time reduction is greater than the damage penalty. It seems to be worth looking into, and verifying the nature of the second spell cast in each state.
____________________________
Anakte Grey
BLM 50 SMN/SCH 50 WHM 40
MRD 26 DRG 30 MNK 30 GLA 25

http://art.penny-arcade.com/photos/416469141_Vt5aq-XL-2.jpg
#3 Sep 27 2013 at 2:52 PM Rating: Excellent
Scholar
44 posts
Anakte wrote:
...you're not accounting for the damage penalty of the first Blizzard III while in AF3.

That's certainly true, since I'm of the belief that AF does not reduce the damage of ice spells. If that is the case, for the "silly" rotation to be worse than the basic rotation the reduction "r" would have to be such that (220*(1-r)+220+154+396)/(4*1.75) < 122.47, or equivalently r > 1 - ((122.47*(4*1.75)-(220+154+396))/220) = 60.3%.

If it's the same 30% penalty as for fire spells under UI3, the silly rotation has expected DPS of (154+220+154+396)/(4*1.75) = 132 potency per second which is still around a 8% improvement over the typical rotation.

Edited, Oct 2nd 2013 1:57pm by cartec
____________________________
Cross-Class Skill Sheet
Kawaii rate me down
#4 Sep 30 2013 at 1:57 PM Rating: Excellent
Scholar
44 posts
Well, I'm happy to report that this damage rotation does not work as described herein: there is a hole in the theory. Some testing notes:
  • From interaction with AF/UI stacks, it's apparent that cast time is calculated at the beginning of a cast, and mana cost at the end.
  • The theory that AF/UI are applied on impact fails to predict the cast times of the B3B3F3F3 cycle: the second B3/F3 cast of each phase seems to take the full 3.5 seconds despite the opposite buff being active at 3 stacks.
  • Given that B3B3F3F3 doesn't benefit from the cast speed buff on every cast, I hypothesized that B3F3 would. Unfortunately, B3F3 wreaks all kinds of havoc with the spell queue: Sometimes the casts are 1.75 seconds, sometimes they complete in 1.75 seconds despite the client putting up a 3.5 second cast bar, sometimes they take the full 3.5 seconds. Sometimes the 1.75 second casts are followed by a short gap during which nothing can be cast. There's definitely inconsistency here that makes me think that disparity exists between the behavior on the server side and the predicted behavior in the client.


My current belief, given this evidence and the behavior in my Firestarter tests, is that the buffs are applied at cast completion but there is a substantial delay of 500-1250 ms before they become visible in the client UI.

In any case, the testing drives plenty of nails into the coffin of this thread's B3B3F3F3 rotation.

Edited, Sep 30th 2013 2:58pm by cartec
____________________________
Cross-Class Skill Sheet
Kawaii rate me down
#5 Oct 01 2013 at 7:37 AM Rating: Excellent
Guru
Avatar
**
362 posts
It annoys me that server/client stability changes what might theoretically work or not work. Anyway, thanks for the testing and the report!
____________________________
Anakte Grey
BLM 50 SMN/SCH 50 WHM 40
MRD 26 DRG 30 MNK 30 GLA 25

http://art.penny-arcade.com/photos/416469141_Vt5aq-XL-2.jpg
#6 Oct 01 2013 at 8:47 AM Rating: Good
***
3,178 posts
Given that the other thread was discussing Aways-Be-Casting and the lost Firestarter Procs. What if:

Fire III > Fire I > Fire III > Fire I > Fire III etc.

So that you don't lose fire starter procs at all. except that those Fire III's cost 1000+ MP.

#7 Oct 01 2013 at 11:53 AM Rating: Excellent
Scholar
44 posts
Gnu wrote:
Given that the other thread was discussing Aways-Be-Casting and the lost Firestarter Procs. What if:

Fire III > Fire I > Fire III > Fire I > Fire III etc.

So that you don't lose fire starter procs at all. except that those Fire III's cost 1000+ MP.

Given the data seem to suggest that buffs are applied at cast completion, it would be interesting to test if Fire III queued to cast during Fire I receives the effect of Firestarter triggered by that Fire I cast. I wouldn't be surprised if the client does some horrible hiccup after starting a cast animation only to stop 750 ms later when the server tells it that the Fire III actually went off instantly.

Fire III does 40% more damage than Fire I with a 40% longer cast time - so the DPS per spell is the same ignoring Firestarter. Mana cost is proportionally higher though: 512 / 319 - 1 = 60% more mana, so Fire III's DPM is ~60% that of Fire I. Interestingly, if you spam Fire I > Fire III then 40% of the Fire IIIs will theoretically be cast with Firestarter, so on average Fire III will take 60% * 3.5 + 40% * 2.5 = 3.1 seconds to cast and cost 60% * 200% * 512 = 614 MP.

Using the framework from my post in the Firestarter thread, that's 2.4 Fires of damage every 5.6 seconds or 2.4/5.6*2.5 = 15/14 burst DPS, with burst phases lasting approximately M gcds, resulting in SFDPS of 15/14 * M / (M + N). With the parameter values used for comparison in that post (M = 6, N = 2.5) that strategy results in an expected 15/14 * 6 / 8.5 = 0.756 Fires per GCD. That's right between wait-and-see and interrupt on proc in sustained DPS:
  1. ABC
  2. Interrupt
  3. 1-3-1-3 (this strategy)
  4. Wait-and-see
  5. Baseline

TLDR: It's not terrible, but not the best, either. If you had to focus 100% of your attention on some fight mechanic without ever looking at your buffs or hotbars, alternating Fire 1 and 3 would be an effective casting strategy handily beating pure Fire 1 spam.

Edited, Oct 1st 2013 12:54pm by cartec
____________________________
Cross-Class Skill Sheet
Kawaii rate me down
#8 Oct 01 2013 at 12:11 PM Rating: Good
***
3,178 posts
Actually under Astral III the later Fire III's would cost 1026 MP, which I think would put that rotation back further in overall DPS.
#9 Oct 01 2013 at 12:18 PM Rating: Excellent
Scholar
44 posts
Gnu wrote:
Actually under Astral III the later Fire III's would cost 1026 MP, which I think would put that rotation back further in overall DPS.

60% of them cost 1026 MP, 40% of them cost nothing because Firestarter. Hence the average cost of 614 MP which is very similar to Fire 1 under AF at 638 MP.

Edited, Oct 1st 2013 1:19pm by cartec
____________________________
Cross-Class Skill Sheet
Kawaii rate me down
#10 Oct 01 2013 at 12:43 PM Rating: Good
***
3,178 posts
Ah. I'm trying to follow along here. So then without casting Fire III during Astral III (just the one under Umbral III) are you saying the average MP cost for Fire 1 spam + Firestarter procs is below 500?

Edited, Oct 2nd 2013 12:33pm by Gnu
#11 Oct 02 2013 at 9:00 AM Rating: Decent
8 posts
I am almost certain that F3 queue'd while still casting F1 does not use a Firestarter. I spam F3 pre-emptively at the end of a mob's health bar regardless of mana (to try to proc firestarter as soon as possible to kill the mob) and if it doesn't proc, I'll start an F3 cast.

Edited, Oct 2nd 2013 4:28pm by tputs
____________________________
Mira Burwani
Black Mage of Exodus
#12 Oct 02 2013 at 1:02 PM Rating: Good
Scholar
44 posts
Gnu wrote:
So then without casting Fire III during Astral III (just the one under Umbral III) are you saying the average MP cost for Fire 1 spam + Firestarter procs is below 500?


If you cast M Fires at 638 MP each, you will have 0.4M Firestarter procs at 0 MP each. That comes out to an average mana per spellcast of 638 * M / (M + 0.4M) = 638 / 1.4 = 456 MP.

Edited, Oct 2nd 2013 2:13pm by cartec
____________________________
Cross-Class Skill Sheet
Kawaii rate me down
Reply To Thread

Colors Smileys Quote OriginalQuote Checked Help

 

Recent Visitors: 1 All times are in CST
Anonymous Guests (1)