This is a predictable result of the recent changes in /stopcasting behaviour. Simply put, there is now a short server-side queue for actions. The length of the queue is not known for certain, but is at least a few tenths of a second. For mages, healers and other caster classes, this means that you can cast one spell before the previous one has finished casting, and the server will automatically queue the second action and thus give perfect chain casting. For hunters, it means that if you hit your Steady Shot button during the 0.5s Autoshot cast time, it will queue the Steady Shot cast until the autoshot has fired, once again allowing perfect chain casting.
Yes, this is one of the possible outcome of the change. However, until we have test result, we don't know for sure if Auto is treated as a cast on the server even though it has a cast time and whether the client would clobber the steady instead of sending it to the server. Unlike mage/lock chain casting spell, this is overlapping cast.
In simplest possible terms: they made it impossible to screw up manual shot weaving, so macros should now be unnecessary, just as mages no longer require /stopcasting macros.
It is not that simple. Steady can still clip Auto, as shown in the 1:2 rotation log if you fire before the .5s window. Instead of manually time your steady after the Auto, now you time it to the beginning of the Auto but not before it. Or just use a weapon fast enough that there is very little room between the previous Steady and Auto so you won't fire 2 consecutive Steady. Incidentally, said fast weapon might also become the highest DPS weapon because it would provide a similar benefit of unlinked Auto, a 1.5s rotation of both Steady + Auto.
Last edited by Elfalora : 01/14/08 at 6:02 AM.
Reason: grammar
I tried this macro out today on Mag, and I was doing about 55% steady's with about 40% autos with some arcane shots mixed in. (extra dps during quakes while running to the cube) This macro seems to work very well. I am no theory expert. But I did notice, it works much better then the older macros. You do seem to use more mana, but the dps is nice.
Yes, this is one of the possible outcome of the change. However, until we have test result, we don't know for sure if Auto is treated as a cast on the server even though it has a cast time and whether the client would clobber the steady instead of sending it to the server. Unlike mage/lock chain casting spell, this is overlapping cast.
Yes, but my point is that following the recent changes, mages and locks are allowed to overlap their casts, and the server will automatically translate that into perfect, lag-free, latency-unaffected chain casting.
Originally Posted by Elfalora
It is not that simple. Steady can still clip Auto, as shown in the 1:2 rotation log if you fire before the .5s window.
I apologise. I should of course have said "practically impossible", in that you now have a whole extra half a second to time your steady shot for perfect chain-casted shot weaving. If your reaction and timing skills are sufficiently bad that you hit the button more than half a second too early, that really is your own problem! :-)
Of course, as others have pointed out, the 2:1 steady:auto shot rotation is in some circumstances higher DPS anyway, although not particularly sustainable in terms of mana usage.
With the 2:1 as BM spec anyway you end up trading some of your auto shots for steady shots. You end up shooting roughly the same number of total shots with less autos and more specials. The number of specials fired caps out pretty quickly at the rate of the GCD. This results in a slight improvement in dps if you are geared such that steady does more than auto on average. But if you look at the DPM this results in - 100 extra dmg for 100 mana - its a pretty terrible tradeoff.
For most people /castsequence !auto shot, steady shot will produce the same quantity of shots fired as /cast !auto shot ... /cast steady shot, close to the same total dps, and much better mana consumption. Manually timing the steadys for a 1:1 will achieve the same.
Yes, but my point is that following the recent changes, mages and locks are allowed to overlap their casts, and the server will automatically translate that into perfect, lag-free, latency-unaffected chain casting.
This is all academic but technically speaking, mages and locks aren't really overlapping their casts, according to the server. Their overlapping part is mainly due to latency. For hunters, it looks like Steady and Auto are actually overlapping their cast time.
I apologise. I should of course have said "practically impossible", in that you now have a whole extra half a second to time your steady shot for perfect chain-casted shot weaving. If your reaction and timing skills are sufficiently bad that you hit the button more than half a second too early, that really is your own problem! :-)
This is guess work, but I suspect that to achieve full overlapping, you would need to time your steady right at the beginning of the .5s window.
+----+ Auto
+--------------------+ Steady
instead of
+----+ Auto
+--------------------+ Steady
Feel free to debunk this therory.
Of course, as others have pointed out, the 2:1 steady:auto shot rotation is in some circumstances higher DPS anyway, although not particularly sustainable in terms of mana usage.
The 2 Steady rotation is higher than the /castsequence but not necessarily better than 1:1 with a fast weapon because it would clip the Auto. To not clip the Auto, you'd need weapon speed that reduces the clipping as much as possible.
1.5 x 2 x 1.15 = 3.45 (MM/SV) -- pre-BC only
1.5 x 2 x 1.35 = 4.05 (BM) -- not exist
All the reports with the 2 Steady rotation have the same characteristic: it works better with slower weapon. That is because you clip your Auto less. The DPS gain mainly comes from the Steady/Auto overlapping and the !Auto in the "new" macro is just a red herring.
With a 1:1 rotation on a fast weapon, you could avoid the clipping and get even higher DPS. Does this make sense?
I used Viper Aspect while shooting, so constant speed.
Weapon speed 2.7, speed bonus 1.38, final weapon speed 1.96.
Now, if you look at the log you will see steady shots hitting within under 1.2 seconds! This happens when steady shots follow each orther too.
Plus you can see an alternation of 1 auto, 2 steady to 1 auto, 1 steady. This is interesting.
My steady shot cast time should be 1.09 seconds, but I don't kow how 2 steady shots can land within ~1.2seconds every time I get 2 steadies after each other. (Should not be possible cause of global cooldown which is 1.5 seconds!)
Other stuff that can be recognized are that if the auto fires after 2 steadies it will land ~400ms later, if it fires after 1 steady it lands ~800ms later (which makes sence since auto timer is ~2.0 and 0.800+1.200=2.0)
My latency is 50ms (german servers rock!), so steady shot 1.2 seconds after the auto shot make sense (1.09 cast speed + 0.05ms latencty ~1.2), but what is irretating is that if one steady follows another they land within 1.2 seconds too.
Thus: It seems that the server allows the new steady to start when the other is finished. (But global cooldown should not be ready)
*edit*
I did the same test for /castsequence !auto shot, steady shot now too.
The steady shot after one auto takes ~200ms longer to shot ~1400ms on average.
Auto shots follow the steady ~450ms after the steady.
The distance between every auto is ~2.0 seconds on average which means that no auto shot gets delayed ever. (final bow speed should be 1.96 in theory and it matched quite well)
i think i read somewhere that haste resets global cooldown.
someone correct me if i'm wrong but if you're BM (and you are), your steady shots are affected by the passive haste from second serpent swiftness, and if the first statements about haste resetting global cooldown, then your steady shot (which i believe is 1.5 secs unhasted) should be ready to cast the same time the GCD is up, allowing you to chain cast it.
My steady shot cast time should be 1.09 seconds, but I don't kow how 2 steady shots can land within ~1.2seconds every time I get 2 steadies after each other. (Should not be possible cause of global cooldown which is 1.5 seconds!)
Those are perfectly logical results given (at least the way I understand them) the "new" latency mechanics. Before patch 2.3, if you hit Steady Shot anytime before your Auto Shot fired, you would start a global cooldown and clip your auto shot. After patch 2.3, if you hit Steady Shot .5 seconds or later before your next Auto Shot would fire, you trigger a global cooldown but delay that steady until after your Auto Shot fires.
For this reason I think it's probably better to use a 1:1.5 rotation for BM now. The most you'll ever clip an auto shot is about .5 seconds, and you'll be making up a ton of shots instead of having dead time in your rotation.
I'm really starting to wonder exactly how badly latency affects us at this point... Using a 3.0 speed weapon, no quiver, no talents, nothing but autoshot turned on, with around a 60-90 ms latency, I get:
1/14 23:05:30.295 Your Auto Shot hits Servant of Razelikh for 430.
1/14 23:05:33.436 Your Auto Shot hits Servant of Razelikh for 437.
1/14 23:05:37.170 Your Auto Shot hits Servant of Razelikh for 436.
1/14 23:05:41.061 Your Auto Shot hits Servant of Razelikh for 438.
1/14 23:05:44.233 Your Auto Shot hits Servant of Razelikh for 437.
1/14 23:05:47.905 Your Auto Shot crits Servant of Razelikh for 891.
1/14 23:05:51.139 Your Auto Shot hits Servant of Razelikh for 445.
1/14 23:05:54.217 Your Auto Shot hits Servant of Razelikh for 449.
1/14 23:05:57.467 Your Auto Shot hits Servant of Razelikh for 447.
1/14 23:06:00.780 Your Auto Shot hits Servant of Razelikh for 447.
1/14 23:06:03.842 Your Auto Shot crits Servant of Razelikh for 896.
1/14 23:06:07.233 Your Auto Shot hits Servant of Razelikh for 451.
Autoshot averaging around every 3.27 seconds, 8% slower than what it should be, an 8% damage loss from the character sheet from nowhere. Could the "0.5 cast time on autoshot" actually just be a client-server communication required between auto-shots, affected by latency?
For anyone interested, I tried the same test, only spamming a /cast !Auto Shot macro. It didn't have any impact on the results whatsoever. Furthermore, I tested Steady Shot macros under the same condition (3.0 speed, no haste). No matter whether it was "/cast !Auto Shot /cast Steady Shot", or just "/cast Steady Shot", I ended up chain casting Steady Shot, no auto-shots inbetween, except during the occasional lag spike (Though the steady shots parsed as nearly a perfect 1.5 seconds between them). I can't say whether it's a cast time or latency issue, but it's definitely still possible to completely clobber your autoshots, even if it's much harder than before. I also tried a /castsequence Steady Shot, Steady Shot, !Auto Shot. It worked for getting 2 steadys per auto, but the latency issues with /castsequence threw it off by about 0.3 sec between each step of the sequence, turning a 3 second rotation into a 4 second rotation.
I'm really starting to wonder exactly how badly latency affects us at this point... Using a 3.0 speed weapon, no quiver, no talents, nothing but autoshot turned on, with around a 60-90 ms latency, I get:
1/14 23:05:30.295 Your Auto Shot hits Servant of Razelikh for 430.
1/14 23:05:33.436 Your Auto Shot hits Servant of Razelikh for 437.
1/14 23:05:37.170 Your Auto Shot hits Servant of Razelikh for 436.
1/14 23:05:41.061 Your Auto Shot hits Servant of Razelikh for 438.
1/14 23:05:44.233 Your Auto Shot hits Servant of Razelikh for 437.
1/14 23:05:47.905 Your Auto Shot crits Servant of Razelikh for 891.
1/14 23:05:51.139 Your Auto Shot hits Servant of Razelikh for 445.
1/14 23:05:54.217 Your Auto Shot hits Servant of Razelikh for 449.
1/14 23:05:57.467 Your Auto Shot hits Servant of Razelikh for 447.
1/14 23:06:00.780 Your Auto Shot hits Servant of Razelikh for 447.
1/14 23:06:03.842 Your Auto Shot crits Servant of Razelikh for 896.
1/14 23:06:07.233 Your Auto Shot hits Servant of Razelikh for 451.
Autoshot averaging around every 3.27 seconds, 8% slower than what it should be.
On top of the delay, you've got some enormous variation in shot intervals (from ~3.1 to 3.9 seconds). I can't imagine how latency would be causing this if you're just letting Auto Shot run. You're doing nothing that might toggle it off and on?
Those are perfectly logical results given (at least the way I understand them) the "new" latency mechanics. Before patch 2.3, if you hit Steady Shot anytime before your Auto Shot fired, you would start a global cooldown and clip your auto shot. After patch 2.3, if you hit Steady Shot .5 seconds or later before your next Auto Shot would fire, you trigger a global cooldown but delay that steady until after your Auto Shot fires.
For this reason I think it's probably better to use a 1:1.5 rotation for BM now. The most you'll ever clip an auto shot is about .5 seconds, and you'll be making up a ton of shots instead of having dead time in your rotation.
You guys sometimes talk about 1:1.5 or 1:2 rotations. I have no clue what that means.
Another remarkable detail is that both logs I made have one thing in common.
At the end they both shot 1 arrow per second. For first log the result was 147xxx/147 and for the sequence log 256xxx/256, that was really funny when I saw that. The first log shots more steadies though and uses more mana.
Another thing that is important: With speed buffs, especially with rapid fire the first macro cannot be topped.
It does alot more damage then, I think this is because of the delay of castsequence.
We need some tests with really fast bows and many speed buffs, I want to see how this would turn out.
@lightpro you need to make a series of 100-1000 shots to get rid of the extreme values and get a more average one.
What works very well is to make a record and then use macros (automated stuff, you show the computer something and he repeats to do this) in your editor to get rid of every line instead of the shots you want. Then bring it into a csv-format (semikolons between every value). Now you can open it in Excel or sth like that.
Yeah, I'm doing alot more testing. It seems like with a lower latency and slow bow, the castsequence does just fine. But, obviously, some haste gear or a DST and a slower bow, the first macro Zork posted would be better. I'm still not sure how this really affects our DPS calculations, though, and what it means for haste rating in general. When you're not hasted, and doing 2 steadies per auto, you're clipping your auto shot by .5 seconds or so depending on weapon speed, so your haste gear is wasted. However, when IAotH procs, or quick shots / DST, that extra haste gear brings you closer to 1.5 seconds, which would be ideal.
So I'm not sure, I think the calculations stay the same using the "new" macro (which is ironically just like using no macro at all), but it's really complicated to prove, and I'm pretty tired, so hopefully someone will come along and do some math for us to prove this true.
I do not understand something about auto shot tests there.
I my tests plain auto shot did performed exactly as listed in character screen, like with 2.09 speed weapon got 58 shots in 2 minutes. To shoot it turned it on and stood afk for 2 minutes.
If I did weave 1 steady then auto got slowed by ~0.1-0.2 seconds(latency value?) even with very slow weapon.
For me latency effects my DPS alot. For some reason i can be sitting at 60-75 in Shat or other zones but i enter into MH or BT my latency jumps to 400's. I clobber the crap out of autoshots when i'm at 400 but smooth as butter at 75.
With respect to the first macro - has anyone tested at what complete cycle time it will stick to a 1:1 rotation?
I'm assuming from previous answers that it would be anything under 1.5s but would like to know how this might be affected by lag.
I'll do some testing when i get the chance but with work and raiding i'm not likely to get much of one:
Also, can someone clarify what 'exactly' is happening right now:
Is it:
(assuming a steady cast time of 1.1s
The first being a cheat of casting steady during the autoshot cast time, the second being an immediate cast of steady after the end of the autoshot cast(similar to new caster mechanics since 2.3)
Basically from what I can tell, if you attempt to fire a special shot (including steady) within the last .5 seconds of an auto-shot, it will start your global cooldown, but delay the casting of that shot until after your auto-shot fires.
Great thread and it has been awesome to get some more insight into shot rotation. I play a paladin as my main but am now looking to start working on my hunter in my spare time and had a few questions people with more experience may be able to provide me with more guidance as I work on his gear setup.
I will be aiming to do a fair bit of both PVP and PVE on him and whilst my Paladin is finished with BT and Hyjal we wont be taking alts through so it leaves me with Badge and ZA gear mainly to start developing him apart from PVP gear.
My question rotates around the shot rotation issues with haste, weapon speed and spec. I will probably look to remain MM so will stacking the haste gear be beneficial with my Sunfury (or ZA bow) and going a simple auto/steady shot rotation or am I better looking at the more traditional agi/AP stacking with crit and going the 1:1.5 rotation? Also is armor penetration proving to be worthwhile to itemise for?
Thanks for any help you can give, my hunters name is Weaksauce on Thaurissan realm if anyone would like to offer some gear upgrade advice as well.
To answer just one part of your question Pariah, stacking ArmorPen goes pretty much the same way as stacking any other stat - stacking for the sake of stacking isn't practically useful. While of all stats ArP probably benefits the most from being stacked (due to increasing absolute returns as target armour approaches zero), for practical gear choices it's better to just treat it as another stat contributing to the desirability of some piece of gear. If you look at the thread about finding the highest DPS gear combo for cheeky's spreadsheet, you'll see a few pieces with armour pen - this isn't just because they have ArP on them, it's because the ArP in combination with the other stats on those pieces make them the best for those slots - the two ZA daggers aren't amazing for having ArP, but the fact that they have it along with so much agi/AP make them desirable. The Gronnstalker pieces with ArP aren't picked because they help stack ArP, they just happen to have ArP as one of the stats that make them very good pieces.
You guys sometimes talk about 1:1.5 or 1:2 rotations. I have no clue what that means.
Here to help you out.
There are three "main" types of rotations 1:1; 1:1 Priority; and 1:1.5. People are theorycrafting about the viablilty of a 1:2 right now. Here is what all that means.
The first number is your auto shot number. The second number is the number of specials you do. So in a 1:1 your number of auto shots equals the number of specials. It's pretty hard to get this exactly perfect, but you can get it really close. For example in a 1:1, if you did 100 auto shots, you would do 100 Steady Shots as well.
A 1:1 priority is almost the same thing. However, you are replacing some steadies with arcanes and multis when thye are up. So the total number of Steady + Multi + Arcane should = Auto Shots.
A 1:1.5 is a lot like a priority, but your specials should add up to 1 and a half times your auto shots. Your still using Arcane, Multi and Steady. This one is mostly for MM and Surv Hunters.
Can anyone interpret this snippet from a parse of my combat log during a raid? I'm not sure I understand the intervals between some of the auto shots, most of which are surpringly varyingly fast -- 0.4 to 0.7 seconds apart, while Quartz is displaying a 0.9 attack speed. I'm using The /cast !Auto Shot /cast KC /cast Steady Shot macro. During this ~9-10 seconds I had bloodlust, DST, IAotH, and rapid fire all up with a 3.0 speed weapon as BM spec. Is this supposed to be how it works, or just an unintended side effect of new UI functionality like the double multi/aimed bug?
Yes, this combat log was taken after the multi-shot/aimed shot fix.
05:30'51.609 Talyx gains Lust for Battle
05:30'51.609 Talyx gains Rapid Fire
05:30'52.093 Talyx gains The Beast Within
05:30'52.718 Talyx's Auto Shot crits Teron Gorefiend for 2527
05:30'53.437 Talyx's Steady Shot hits Teron Gorefiend for 1089
05:30'53.562 Talyx's Auto Shot crits Teron Gorefiend for 2675
05:30'54.375 Talyx's Auto Shot hits Teron Gorefiend for 1099
05:30'54.890 Talyx's Steady Shot crits Teron Gorefiend for 2513
05:30'54.921 Talyx's Auto Shot hits Teron Gorefiend for 1115
05:30'55.406 Talyx's Auto Shot hits Teron Gorefiend for 1164
05:30'56.234 Talyx's Steady Shot crits Teron Gorefiend for 2707
05:30'56.406 Talyx's Auto Shot hits Teron Gorefiend for 1124
05:30'57.109 Talyx's Auto Shot hits Teron Gorefiend for 1147
05:30'57.734 Talyx's Steady Shot hits Teron Gorefiend for 1051
05:30'58.015 Talyx's Auto Shot hits Teron Gorefiend for 1139
05:30'58.734 Talyx's Auto Shot hits Teron Gorefiend for 1127
05:30'59.125 Talyx's Auto Shot hits Teron Gorefiend for 1097
05:30'59.500 Talyx's Steady Shot hits Teron Gorefiend for 1085
05:30'59.875 Talyx's Auto Shot hits Teron Gorefiend for 1179
05:31'00.562 Talyx's Auto Shot hits Teron Gorefiend for 1169
05:31'01.156 Talyx's Steady Shot crits Teron Gorefiend for 2627
05:31'01.312 Talyx's Auto Shot crits Teron Gorefiend for 2801
Ok, to be sure it wasn't a fluke... Almost 23 minutes against a Servant of Razlikh, nothing but /pet attack, /autoshot, /XBox360. Definitely not as bad as the first one, but in 1347.484 seconds, I get 433 auto shots, instead of the 449 expected, with an average of 3.11 seconds between shots instead of the 3.0 on the character page. Here's the edited log of just the Auto Shots, if anyone wants to parse it for any more info. Keep in mind, this log is taken at peak time, with a 100+ queue to log onto the server, so I plan on doing another run at a non-peak time to see what the impact is. (My worry is how accurate any other shot rotation testing is going to be, if even the autoshot timer isn't reliable.)
Edit: Adding in a short melee log, as a counter-ballance. 150 swings with a 3.6 speed weapon, expected 540 seconds, actual 540.75 seconds. Pretty dang accurate, especially given the large lag spike that starts a 22:42:47 that's quickly made up for.
I'm feeling a little nervous about posting here, as I've been trying to wrap my head around the changes that've been happening over the week, while this thread continues to post consistant advances on whats going on. I'd just had this forum pointed out to me by a guildmate a couple of days ago, who believed (correctly) that the mechanics in play here would be easier to understand if I looked at what others were doing. I feel like I do have a little something to contribute though.
I took our evening's attempt on magtheridon as an opportunity to test different weapons and macros in an effort to understand what exactly was going on with the new macro and with the old macro's as well.
New macro being: /cast !autoshot /cast steadyshot
and old macros being traditional ( "!" free) castsequence rotations.
There really wasn't anything new to contribute numerically from the evening, diligent hunters have already layed out far more thorough numbers earlier in the thread. And after the raid I didn't really feel like I had any clearer a picture in my head of why that macro was working so well. Additionally a hunter in our guild had mentioned something about lost arrows, where the number of shots in his log was less than the number of arrows missing.
So I wandered out to dr. boom, grabbed my wristwatch, and timed minute long bursts using the "new" macro, with wolfslayer and consortium blaster. (I am BM with SS, and the 15% quiver so the attack speeds are 1.96 and 1.74)
Some simple math told me that a perfect 1:1 rotation would fire off somewhere around 61 bullets with wolfslayer
(60 / 1.96) * 2 = 61.22 something
and 69 bullets with the consortium blaster
(60 / 1.74) * 2 = 68.96 something
+/- 2 because of the clean start cooldown wise and that I might get one last one is as I stopped targetting.
I fired 63 bullets with wolfslayer, and 71 with the blaster. which made my brain hurt even more (though at least I wasn't worrying about mysterious bullet loss). I was mystified as to the effeciency of it. (efficiency in a dps time sense, not in a mana sense). maybe I graph my rotations wrong but when I graphed it out on paper it looked waay uglier than the results I was seeing in game.
I didn't think to actually take down the number of steadys and autos for each, but I'm positive I had more steadies than autos, as observed by others in this thread, so I was certain it wasn't doing just a straight 1:1. And thats when it really clicked for me that if your GCD starts early due to you hitting the SS in the autoshots .5s "warmup" and if steadyshot is smaller than the GCD, that it actually "pulls" the gcd back toward the end of the actual steady shot.
I know none of this is news. But I wanted to post a thank you for all the hard data, observations, and insight thats condensed in this thread; and at the same time explain how I finally visualized it in my head in the hopes that other people who're still baffled at this stuff will maybe start to see it too.
For me I think of steady shots as little chunks of cast time and the GCD as a gooey bag they're stuck in. that macro strings them together as able, but puts Autoshots at a higher priority. So in my head, I've got a row of dividers that are my autoshots, pushing the chunky SSs one way or the other in that bag of GCD. because an autoshot will pull a steadyshots gcd up to .5 seconds before it, and because at my current speeds I have almost .5 seconds to play with, they're pretty fluid. Unless an autoshot hits ready at exactly the wrong time, you keep great autoshot timing while the gcd's flow into a pretty efficient special chain as well. its enough that for me, with my 250 latency and manual macro button pushing, I get better results letting the game glob them into place than trying a fixed 3:4 or 1:2 or 1:1 shot rotation.
So yea, talk about dumbing it down, but sometimes you need to understand something on that stupid level to start to be able to look at it on a more complex level. Sorry for the sparse numbers too, I was tired when I tried to do all this.
Someone correct me if I am wrong, but for BM hunters it seems a /castsequence !auto shot, steady shot macro should be superior to the /cast !auto shot ... /cast steady shot spam macro because for BM hunters steadys and autos hit for very close to the same values. With both macros you end up firing roughly the same number of shots. With the spam macro you just trade several auto shots for steady shots resulting in close to the same dps for a lot more mana consumption.
01/14/08, 6:01 AM
Similar Threads
