castsequence lowers dps big time. 500dps loss

[maxel]

So this explains it. I was noticing something very similar on my lock and mage dps comparisons. The mages with just a /cast frostbolt macro were far outpacing my lock in the conversion of Gear to DPS while he was using a castsequence.

Perhaps the solution lies in more fancy macros with branching? What I mean is maybe make the macros simple and without castsequence altogether, but have the last line hit a new bar or move back to the start. Let's say we want a 3 sequence spell rotation Flame Shock >> Lava Burst >> Lightning Bolt >> LB >> LB. We could do the following:

Bar 1 - Button 1:
/cast [target=focus] Flame Shock
/changeactionbar 2

Bar 1 - Button 2:
/cast [target=focus] Lightning Bolt

Bar 5 - Button 1:
/cast [target=focus] Lava Burst
/changeactionbar 1

Assuming you've bound the keys for Button 1 and 2 to 1 and 2 your sequence would now be spam 1 > 1 > 2 > 2 > 2 > etc

I'm at work and unable to test and throwing out some brainstorming. Would this work?

FWIW I used the above code on my warlock and mages and it did in fact work and add about 500dps to both sets of characters. Here's what I have:

On each toon I have my primary bar and my switch bar (action bar 5) as well as 2 buttons we'll call them setup 1/2 and primary 1/2 for a total of 4 buttons:

Warlock:
Bar 1 Button 1: (Setup 1)
#showtooltip curse of the elements
/assist party1
/use 13
/use 14
/pet attack
/castsequence reset=300/combat/target Curse of the Elements, GarbageSpell
/changeactionbar 5


Bar 1 Button 2: (Primary 1)
#showtooltip Shadow Bolt
/assist party1
/use 13
/use 14
/pet attack
/cast Shadow Bolt

Bar 5 Button 1: (Setup 2)
#showtooltip Corruption
/assist party1
/use 13
/use 14
/pet attack
/cast Corruption

Bar 5 Button 2: (Primary 2)
#showtooltip
/cast Life Tap
/changeactionbar 1

My tapping sequence is now 1 > 1 > 2 > 2 > 2 > 2 (Repeat until corruption is falling off then start over

I also have 2 frost mages their sequence is as follows:
Bar 1 Button 1 (Setup 1)
#showtooltip Ice Lance
/assist Party1
/use 13
/use 14
/petattack
/cast Icy Veins
/cast Ice Lance
/changeactionbar 5

Bar 1 Button 2 (Primary 1)
#showtooltip FrostBolt
/assist Party1
/use 13
/use 14
/petattack
/cast Icy Veins
/cast FrostBolt

Bar 5 Button 1 (Setup 2)
#showtooltip Summon Water Elemental
/castsequence reset=144 Summon Water Elemental, Cold Snap

Bar 5 Button 2 (Primary 2)
#showtooltip FrostBolt
/assist Party1
/use 13
/use 14
/petattack
/cast Icy Veins
/cast FrostBolt
/changeactionbar 1

This uses the same castsequence as the warlock and bumped their dps from 1200ish to 1700 on a good streak. Keep in mind when I say bumped I really mean bumped. I have been performing the same benchmarks over the last year trying to maximize my dps and in 5 minutes of testing alone I can tell already that there is a ton of merit to the statement that castsequence does in fact lower dps.

[heffner]

A quick test with my L60 shaman on target dummy. I used 1) lightning bolt directly bound to a key, 2) /castsequence with five lightning bolts and 3) /castrandom with five lightning bolts.

Castsequence definitely waits until the cast ends. I do not get any error message with button spamming - "Another action is in progress". With Quartz, the "red bar" never really changed. Also, the next cast never went off before the Quartz cast progression bar hit the end of the bar (hope that makes sense).

Using Cast or Castrandom does not wait until the cast ends. I get an error message with button spamming and with Quartz the "red bar" changed often and the next cast went off before the moving cast progression bar hit the end of the bar.

I am not sure if /cast was better than /castrandom or not.

Not very exhaustive, but it is very obvious to me that the total number of casts over a specified time period would be less for the /castsequence macro. The difference between casts is visually noticeable.

[Sam DeathWalker]

In general, you cannot cast more than one spell with a single click of a macro. Most spells and some items trigger the global cooldown (GCD) which keeps you from taking too many actions at once. Even if a spell fails to cast, if it would trigger the GCD, it prevents subsequent /casts in the macro from running. This was not the case prior to patch 2.0 which is why you may still come across macros like the following:
/cast Overpower
/cast Execute
/cast Mortal Strike
/cast Sunder ArmorMacros like this do not work anymore. As soon as Overpower fails to cast, the game will block all the other spells from casting as well, even though the GCD is not actually triggered.

There is a bit of good news, though. Certain spells can actually be cast at the same time in a single macro. Any spell that is instant and does not trigger the GCD can be followed by another cast ("Next Melee" abilities like Heroic Strike fall under this category too). The spell's tooltip will tell you if it's instant, but you have to use the spell (or check a spell database site like thottbot.com) to know if it triggers the GCD. Example:
/use Talisman of Ephemeral Power
/cast Arcane Power
/cast Presence of Mind
/cast Pyroblast

Humm I don't know if using

/click
/click
/click

to get around the /cast limitation is umm.. At any rate thats a nerf waiting to happen and bad practice to get into.

Why would two /castrandom or /castsequence be usable in macros? Once there is ANY failure to cast a spell that triggers the GCD that failure will stop the WHOLE macro (including the 2nd /castsequence or /castrandom) correct?

And if your first one dosn't fail then why do you need to put in the 2nd anyway?

So if line 1 fails you dont get to line 2, if line one 1 works you dont get to line 2. Thus you never get to line 2.

Humm so the game checks for a failure with /cast but dosn't with /click, /castrandom or /castsequence? Meaning you can just use any of those three instead of /cast? ... well pffttt

/castrandom Flame Shock,
/castrandom Lightning Bolt,
/castrandom Fire Nova Totem,

Well ...

[heffner]

Ok, did a small amount of testing.

Scenario: L60 Draenai Shaman, Naked, No Talents, Had some 1% to hit buff the whole test time (not sure where it was from)
Cast Rank 10 Lightning Bolt at the L60 Target Dummy
I used Recount to calculate the Total Damage and DPS etc..

Macros:
1) /castrandom Lightning Bolt, Lightning Bolt, Lightning Bolt, Lightning Bolt, Lightning Bolt - I will call this RM
2) /castsequence Lightning Bolt, Lightning Bolt, Lightning Bolt, Lightning Bolt, Lightning Bolt - I will call this SM
3) just Lightning Bolt place on Hotbar - I will call this CN

Performed 10 trials for each macro. I spammed the key as consistent as I could, quite quickly (this is a variable however). I remained the same distance from the dummy each trial.

With my mana I was able to cast 13 LB's every time

Mean Total Damage (n=10)
RM - 5638.1
SM - 5800.2
CN - 5862.4

TTest (2-tailed) comparing mean Total Damage between RM and SM to CN and RM to SM (sorry I don't have ANOVA - or at least I don't know where it is in OpenOffice)

P-values
RM vs CN - 0.13
SM vs CN - 0.60
RM vs SM - 0.32

Total Damage was not significantly different. This is expected and indicates I can keep all data in the analysis (includes Crits and Misses).

Mean DAMAGE PER SECOND (DPS) (n=10)

RM - 166.12
SM - 154.92
CN - 172.49

TTest comparing mean DPS between RM and SM to CN and RM to SM (sorry I don't have ANOVA - or at least I don't know where it is in OpenOffice)

P-values
RM vs CN - 0.16
SM vs CN - 0.000022
RM vs SM - 0.026

There was a significant difference in mean DPS between the Castsequence (SM) Macro and the Castrandom (RM) Macro and Standard straight casting hotkey (CN). There was not a significant difference between RM and CN.

I will do some more testing later on and exclude all trials that have crits and misses. If I exclude them I lose half the data, but the p-values for the mean DPS become 0.000035 and 0.000038 for the SM vs CN and RM, respectively; while the mean total damage remains not different.

Out of 130 LB's:

RM - 1 Crit, 5 Misses
SM - 5 Crits, 3 Misses
CN - 5 Crits, 3 Misses

[heffner]

Well, for some reason I couldn't log in my other toons so I did 10 more trials for each macro type.

The same scenario as above.

Total casts = 260 LB's for each of the three "macro types".

Mean Total Damage:

RM - 5759 (8 crits, 10 misses of 260 total)
SM - 5765 (7 crits, 7 misses of 260 total)
CN - 5835 (10 crits, 7 misses of 260 total)

TTest comparison of means were not significantly different between all groups (p-value > 0.05).

Mean DPS:

RM - 170.1
SM - 153.5
CN - 173.4

TTest comparison of means:

RM vs. CN p-value = 0.274
SM vs. CN p-value <0.000001
RM vs. SM p-value = 0.000015

If I remove all trials that had at least 1 miss or 1 crit:

Mean Total Damage (n=# of successful trials of 13 LBs each):

RM - 5884 (n=7)
SM - 5816 (n=9)
CN - 5912 (n=9)

TTest comparison of means were not significantly different between all groups (p-value > 0.05).

Mean DPS:

RM - 173.9
SM - 154.1
CN - 174.0

TTest comparison of means:

RM vs. CN p-value = 0.90
SM vs. CN p-value <0.000001
RM vs. SM p-value <0.000001

So, there is a significant difference in mean DPS between the /castsequence macro (SM) and the /castrandom macro (RM) and simple hotkey (CN) for LB spam. The difference may not seem large but I don't know how it scales at higher levels and with gear/buffs.

[Kicksome]

This needs to use damage over time, not damage per lightning bolt(s). You're missing the time piece which is what we're saying is off.

We're saying there is a delay between casts with castsequence, that lowers your avg. damage per second because you're not doing anything between casts.

If you just take "100 lightning bolts" it doesn't take this into account.

The problem is with castsequence 100 lightning bolts may take 220 seconds to cast, vs /cast might take 200 seconds. But the total damage should be almost the same.

[heffner]

I used DPS. Damage per second. Recount provides this. I show this to be different between the macro types. Unless I am mistaken with what Recount provided.

Total damage was not different. The only reason I compared this was to determine if the crits/misses made a big enough difference in the data set. I expected this to be the same.

So, based on my observations with LB spam in this scenario: for a 10 second fight I would lose about 200 total damage, and for a 30 second fight I would lose about 600 total damage using /castsequence vs. a hotkeyed spell or /castrandom. This works out to be a ~11-13% loss of damage if you use /castsequence vs. /castrandom or the hotkey for LB spam. So, I would imagine that the faster the cast time of a spell the more DPS that will be lost and vice versa.

[Fuzzyboy]

/click flameshockbutton
/click LVBbutton
/click CLbutton
/click LBbutton


how about that?

That would cast FS every time its up, meaning once every six seconds. We really dont want that

Ideally, we would cast FS in the following scenarios:

- When the old FS expires (every 18 secs)
- When we change target, but only if LavaBurst has cooled down

Ideally we would cast lava burst in the following scenarios:

- Whenever its up, but only if FS has been applied

Ideally we would cast chain lightning (this is only true until they patch in the changes to lightning bolt)

- Whenever it's up, but only if one of the following conditions are true: a) lava burst is on cooldown b) it will hit multiple mobs

When lightning bolt changes are patched in, it becomes:

- Whenever it's up, but only if it will hit multiple mobs.

Ideally we cast lightning bolt:

- When lava burst is on cooldown AND chain lightning is on cooldown AND flame shock isn't about to expire

Unfortunately it's not really possible to macro those conditions.

A 2-button setup would work with /click, ie. one button for flame shock (you manually apply flame shock and make sure it's always on before casting lava burst) and one button for the macro /click lava burst /click chain lightning etc.

A 1-button setup only works with a castsequence or castrandom, both of which are inefficient. At the moment I'm using:

/castseq reset=combat/8 flameshock, lava burst, LB x 4-5

Which works ok (flame shock is applied more than necessary but I'm ok with the dps loss). However, add to that the problems with sequence and I might have to change some stuff or at least spend some time with the dummies

[Bigfish]

Ok, here's what I found out with cast-sequences last night. If you have differing spells, 2 castsequence will alternate. Example:

/castsequence Fireball, fireball, fireball
/castsequence frostbolt, frostbolt, frostbolt

will cast fireball, frostbolt, fireball, frostbolt, etc.

If you have identical spells in the same firing position in both castsequences, one cast will count as a cast for both sequences, with the exception of the first spell. Example:

/castsequence scorch, fireball, fireball, fireball, pyroblast
/castsequence scorch, fireball, fireball, fireball, pyroblast

will cast as scorch, scorch, fireball, fireball, fireball, pyroblast, restart. 1 or 2 scorches will go off at the restart of the sequence. Not sure what the contributing factor is.

Instant cast spells do NOT trigger the ~.2 second lag, but this is because their response from the server for a successful cast falls in the GCD anyway.

The classes the most affected by this are the ones with castsequences that use spells that aren't instant cast, ie shamans, firemages trying to juggle hotstreaks, etc. The more casts with a cast time you try and fit in to a period of time, the more it will affect your dps. Shamans in particualr, as LB with a 2 second cast time means you add a lot of dead space when spamming it.

Now, there ARE work arounds, depending on what sequence you want to get out. Personally, I'm going to just live with the dead space on my mage. I get more out of my fire rotation from instant cast pyroblasts than I lose from lag times. I may look over my other rotations to see how many extra seconds I'm adding in, but bottom line I'm not too worried about it at the moment.

For the issue at hand, the shaman macro. I'm not going to get that worked out for myself until I actually hit 75. Another potential fix that probably won't work:

/castsequence Flameshock, lava burst, lava burst
/cast Lightning Bolt

Should cast Flameshock, Lava burst, LB, LB, LB, LB (a 5th one if you have the spell haste), Lava burst, Lightning Bolt, restart at flameshock. By attaching some Lightning Bolts to the end of the Lava burst line, you should be able to delay the next application of flameshock until Lavaburst is up again. Once more, I can't test this myself; just theory craft on my part.

[Bigfish]

Ok, worked out my castsequence for my mage.

/castsequence reset=target scorch, Fire Blast, Living Bomb, Fire Blast, Pyroblast, Fire Blast, Living Bomb, scorch, Fire blast, Pyroblast, Fire Blast, Living Bomb, scorch, Fire blast, Pyroblast, Fire Blast, Living Bomb, scorch, Fire blast, Pyroblast //repeat
/castsequence reset=target scorch, null
/cast Fireball

which should run as:

0 Scorch
1.5 scorch
3 fire blast
4.5 living bomb
6 fireball
7.5 
9 fire blast
10.5 pyroblast
12 fireball
13.5 
15 fire blast
16.5 Living bomb
18 scorch
19.5 fireball
21 
22.5 Fire Blast
24 Pyroblast
25.5 fireball
27 
28.5 Living bomb
30 scorch
31.5 fireball
33
34.5 fire blast
36 Pyroblast
//repeat from 25.5

This assumes my crit rate is high enough to generate instant cast pyros 100% of the time (which its pretty darn close to). A Fireball gets inserted in to the sequence every time the castsequence lag should kick in after a scorch, as well as any point where the initial sequence is waiting on the fire blast cool down.

Edit: I have approximately a 90% chance to generate an instant cast pyroblast every sequence once the sequence gets running.