castsequence lowers dps big time. 500dps loss

[Kicksome]

Tested again. No totems and I'm not spec'd for lightning overload.

I kept casting until 1000 mana

1) /cast lightning bolt 262,142 damage, 1948.3 dps
2) /castsequence lightning bolt, lightning bolt,lightning bolt,lightning bolt,lightning bolt,lightning bolt,lightning bolt,lightning bolt,lightning bolt 291,531 damage, 1675.1 dps

I only had 125ms lag. When I tested it before I had 233ms lag (horde were in IF and I didn't know it). It seems like the better your latency, the less delay. So make sure to add that to your tests.

Naked test (only tabard, no totems):
1) /cast 18,238 damage, 459.3 dps
2) /castsequence 16,314 damage, 372.0dps

[Bigfish]

I'm not convinced this is a problem. I just did some test myself...

No totems, only Flametounge Weapon, lvl 70 target dummy. (i'm 75)

1) /cast LB only = 1400 DPS
2) /castsequence Lb, LB = ~1050 DPS
3) /cast [target=focustarget] LB = 1100 DPS
4) /cast LB only = 1100 DPS
5) /cast LB only = 1200 DPS

Thats where I stopped.

Are we sure thats its a macro problem and not just a randomness problem with Shaman? Remember shaman are all sorts of procs (like lightning overload) that makes them very random....

What about clear casting (randomly making the test longer but increasing your mana pool). And how about elemental Oath?

I'm going to test this over lunch. More later tonight as well.

[Kicksome]

Remember, I said I was casting until 1K mana. NO totems or buffs at all.

When you cast slowly, you use less mana - since you cast less. And you also benefit from casting MP5, since combat is longer. This means you run OOM slower.

When you cast more quickly, you use more mana.

I'm testing to confirm the delay in /castsequence. The debate on DPS vs. more total damage is something different.

[genocyde]

Or, for that matter, just cast LB a set # of times and time it. take procs/damage out of the test all together. If it takes you an extra 2.2 sec to get off 20 spells then you know you've got a 110 ms delay or something of the sort.

However, I beleive there has already been sufficient testing for me to agree this is the case and by default if your waiting for a server response; the worse your lag the worse it's effect on you will be.

[Kicksome]

naked test - 30 seconds (using wow stopwatch) (116 latency)

try 1
/cast 13654 damage, 434.8 dps
/castsequence 12667 damage, 396.8 dps

try 2
/cast 13610 damage, 434.5 dps
/castsequence 12835 damage, 405dps


try 3
/cast 14731 damage, 467.7 dps
/castsequence 13817 damage, 409.4 dps

try 4
/cast 15635 damage, 494.2 dps
/castsequence 13689 damage, 440.0 dps

[genocyde]

Sounds like it's time to start looking for another solution for anyone interested in maxing dps. Might be the only way is to multi task your own rotations. Anyone give Bigfish's two /castsequence idea a shot yet?

[Bigfish]

working on it right now. Will report soon.

[Fuzzyboy]

Your results will vary a lot because of crits. Even with only 10pct crit chance, you'd need something in the line of 10000 casts to get a near-reliable figure. If, however, we're talking about determining a variance in the 15 pct area, less might be enough to get an indicator, but keep variance in mind when testing :-)

[Bigfish]

Ok, here are the numbers I got, based on 20 fireball samples and their dps, crit rate, and time to complete:

Alternating castsequence macro:

/castsequence Fireball,,
/castsequence , Fireball,

DPS Crit Time
820 4/20 1:03
956 8/20 1:03
954 7/20 1:04

pure /cast

773 3/20 1:03
921 6/20 1:03
843 6/20 1:03


/castsequence fireball, fireball, etc

984 10/20 1:09
967 9/20 1:08
957 9/20 1:08
676 2/20 1:07



Castsequence came out with the generally higher dps, but I blame this on my stupid freaking crit streaks when I got to that point in testing. It was still taking an extra 4-6 seconds to get the same number of casts in. Take that as you will. Good news is the alternating castsequence macro works. And now I have to actually get back to work. Be abck later.

[genocyde]

Your results will vary a lot because of crits. Even with only 10pct crit chance, you'd need something in the line of 10000 casts to get a near-reliable figure. If, however, we're talking about determining a variance in the 15 pct area, less might be enough to get an indicator, but keep variance in mind when testing :-)

There might be a way around this inconsistency as well. Find yourself a friend (your other toon sitting next to you) and put on his full pvp gear and talent him for less % chance to be crit and duel yourself! If your dps'r is naked and your target is pvp geared or talented as such I doubt you'll ever see a crit. Just make sure you aren't specced in to any reactionary procs.

The other solution to this is to run a good dps meter like recount and run two toons on the same target dummy. On one toon run the /castsequence macro on the other run the /cast macro. You should be able to tell quite easily who is getting off more casts/min just be looking at the number of instances of your chosen spell on recount when the fight is over. doesn't matter if they hit / crit / miss it will know.
Also if your running cast bars on everyone you should be able to see a toon falling behind