05/12/08, 12:40 PM   #1396
Happy October 19th!

Night Elf Druid

Dragonblight
 Originally Posted by Melador I know that 1392 is the theoretical point where gemming for haste is better than gemming for damage, but do we know exactly how much better it is?
It depends on how much damage and how much haste you have. Hence the spreadsheet. On the other hand, Drums and Heroism are straight buffs, not a trade-off, so they're pretty much always gonna increase your DPS.

05/12/08, 12:54 PM   #1397
thedopefishlives
Don Flamenco

Baelgun
 Originally Posted by Adoriele Oh oh oh. I think I see what you're saying, and I'm going to try and work through it once I get to working with Cast Rotations. Nature's Grace isn't a huuuuge issue, effectively I'll be treating cast rotations as a statistical weight of discrete cast times based on the number of expected procs, i.e. $T_0= x\% *(\text{total cast time with no procs}) + y\% *(\text{total cast time with one proc}) + ...$ limiting it to cast rotations with total cast time on either side of the DoT duration, allowing you to choose whether to clip the DoT ticks or not. Haste procs, though... that might be a little more squiggly. Haven't come up with the way I wanna worry about those yet.
The way I was viewing haste procs, which is how I came to the assumption that they are convergent, is as follows: Begin with some arbitrary amount of static haste. Each cast in a given cast cycle will thus have a fixed cast time, therefore, the rotation will last $X$ seconds and have $C_0$ casts during that time. Adding in a proc effect will cause an average reduction of $Y_1$ seconds, giving us $C_0+C_1$ casts during $X$. This, obviously, will give us some additional chance to proc, which introduces an additional $Y_2$ reduction, and so on to infinity. If I can factor in all the variables and derive the limit of this equation for even one proc-based haste trinket, it should be possible to derive an exact static haste value for such an effect without resorting to the clumsy iterative model I have in Rawr.Moonkin.

Last edited by thedopefishlives : 05/12/08 at 12:55 PM. Reason: Fixed my equation

05/12/08, 2:55 PM   #1399
Mercurial Rapper

Troll Druid

Mal'Ganis
 Originally Posted by Melador Yeah, I noticed I wasn't seeing the returns I was expecting out of haste, but obviously it's not based on Adorele's calculations at all, it's just simulating things as I believe them to be. So either something's impacting the value of haste, or I have a bug somewhere. I'll probably just release the source code after I clean it up a bit so people can "check my math".
So I played with this a bit at lunch, and here's what I'm seeing:

Haste isn't THAT much better than straight damage after the tipping point. Take someone with 1500 damage and 10 open sockets: either they can go 1620dmg and 0 haste, or 1500dmg and 100 haste. That translates to 2796 average starfires every 3 seconds or 2638 average starfires every 2.82 seconds. Or 932dps vs 936dps.

Obviously that assumes no crits and nature's grace procs, but I can't think of a reason why that would affect the results -- haste affect is proportional to the spell cast time. Same with resists -- over time their impact should be the same in either case. So yes haste is better, but it's not a TON better, from what I can tell.

One downside to haste in my simulator is that haste makes you cast more, so you hit the latency more often. Which means haste doesn't help quite as much as it would under a theoretical zero-latency situation.

 Trinkets. I know ur getting sick of people asking for an activatable trinket fix but I have to add my voice simply because the spreadsheet is so good otherwise
You might want to try my dps simulator for seeing the difference between trinkets. I have all the common ones simulated, and one of the advantages of running a simulation is that trinkets are trivial to simulate, whereas they can be tricky to theorycraft into a single formula for use in a spreadsheet.

Also, it's "you're", not "ur".

 05/12/08, 10:26 PM #1400 spi Von Kaiser   Spi Night Elf Druid   Akama Is that really how it is even with the spell queue system? Higher latency makes haste worse? I'm still skeptical about a lot of the latency=worse dps theories. according to the simulator i'd gain 100+ dps from improving my ping from 100ms to 45ms.
 05/12/08, 11:11 PM #1401 Melador Mercurial Rapper     Melador Troll Druid   Mal'Ganis Well, there's always going to be a gap between casts due to latency. The spell queuing stuff definitely helped -- we're not eating a full latency cycle between casts any more -- but there's still a gap which is what the Latency field represents. For example, if I spam wrath at a target I average 1.45s per cast, but my tooltip lists it at 1.38s, and my in-game listed latency is 200ms. There IS a full latency cycle when I first start casting, but the current "automatic stopcasting" system brings my chaincasting effective latency down to ~70ms. So yeah, I wouldn't just plug in your listed latency, but there's definitely still some time between "last spell is done" and "next spell is casting". The only reason latency negatively impacts haste is because haste causes you to cast more spells, and every time you cast a spell you eat a latency delay.
05/13/08, 3:56 AM   #1403
Efejel
Piston Honda

Tauren Druid

Dragonblight
 Originally Posted by spi Is that really how it is even with the spell queue system? Higher latency makes haste worse? I'm still skeptical about a lot of the latency=worse dps theories. according to the simulator i'd gain 100+ dps from improving my ping from 100ms to 45ms.
That sounds a little unrealistic since average human reaction time ( Reaction time - Wikipedia, the free encyclopedia ) may be substantially higher, though there isn't a category for a periodic reaction to a single-variable stimuli with at least 1.5 sec of advance notice of the specific variability... If I recall correctly, someone had done pretty thorough analysis of WWS parses and felt that ~50 ms was reasonable reaction time, but a bit on the slow side? I'm afraid I don't recall where I read that, it very well could have been somewhere in this thread.

*D'oh* Apparently responses to this effect spilled onto the next page, and I missed seeing that I was repeating something already mentioned. Lemme see if I can find the WWS parse comment to add some value to this post...

"Electronic communities build nothing. You wind up with nothing. We are dancing animals. How beautiful it is to get up and go out and do something." - Kurt Vonnegut

 05/13/08, 7:32 AM #1404 spi Von Kaiser   Spi Night Elf Druid   Akama I'm curious to know why people keep mentioning the human "reaction time" when you now just spam buttons for the queue system. Am I missing something here? Reaction time shouldn't really be considered when you're spamming a button. It would have made sense to me to consider the human reaction time when one uses stop casting macros, but I don't understand where "reaction time" comes into play for the new queue system where you spam one button. You aren't reacting to anything.
 05/13/08, 7:46 AM #1405 ♦ Lorewanderer Moof.     Lorewanderer Tauren Druid   No WoW Account While there may be an element of it, there are certainly things one can do to get around latency. The most straightforward way of getting around it thaqt I've heard of is to bind your primary nuke to the scroll wheel, and when the previous cast time is coming to an end, begin spinning it. I use a more complex system with a custom driver where if I hold down a button it hits it once, then spams it 15x a second after a short delay. Given my extremely jittery lag, I will occasionally see it nearly "skip" an entire NG-wrath cast castbar in Quartz. That's something I know I couldn't accommodate with any level of reflex, but actually works quite well with some intelligent driver tweaks.
 05/13/08, 10:43 AM #1406 Öwlcapwn Banned   Owlcapwn Tauren Druid   Korialstrasz I have a simple question, I hvae been raiding bt / hyjal for 4+ months now as balance, I have 4/5 t5 and 4/5 t6 atm, no Sunwell pieces yet. So I can choose to wear a high damage 4/5 t5 set or a high haste 4/5 t6 set. I was wondering what kind of concensus there was to which option would be best for a fight like brutalis that requires maximizing dps. in my haste gear (under ideal raid buffs / group which happens 80% of the time) i can roll with 175-200 haste and sit at about 1200~ damage unbuffed. Does 4 set t5 bonus outweight t6 until I get more haste or can wear both 4/5 t5 and 4/5 t6 sunwell? I've ran some of the numbers thru the spreadsheet but I'm having trouble finding real noticable results. The spreadsheet doesnt seem to be updated completely with 2.4 mechanics. EDIT: I would also wonder about Treants added to the rotation. They seem to do a tremendous amount of damage done to my boss fights, and some fights require them to be thrown out at well timed moments. Are they a dependable source of DPS? meaning should I always try and use them or not. I've had our main tank question the overal effectiveness as the chance that the boss parries my treant and cause more spike damage to the tank. Last edited by Öwlcapwn : 05/13/08 at 10:52 AM.
 05/13/08, 10:52 AM #1407 Melador Mercurial Rapper     Melador Troll Druid   Mal'Ganis Lorewanderer/spi: yeah, I don't think reaction time plays much of a role when you're chaincasting, I'm typically spamming my button pretty good when it's getting close to my current cast being done. It doesn't get around latency entirely (see my above example with wrath spam), but it helps. Owlcapwn: From what I can see, 4pc t5 is a ~100 dps bonus. If you're not getting that with your additional t6/sunwell gear, I'd stick with the t5.
 05/13/08, 11:29 AM #1408 ♦ Lorewanderer Moof.     Lorewanderer Tauren Druid   No WoW Account Treant safety would depend a good deal on how your tank transitions happen, and how you work around them. If Brut is being spun at all by taunts, the parry chances go up quite a bit. If there isn't much movement going on, dropping them in an intelligent spot behind the boss won't really risk additional parries. As for t5 v. t6, are you being asked to keep IS up (if no, and if tank death is a serious concern, why not)? Since you'll have the dots going regardless, I'd stick with t5 to take advantage of that. If tank death is less of a concern, and debuff slots are an issue, I'd go t6 to avoid making that issue worse and ease up the timing issues--keeping a tight rotation on moonfire without clipping/letting it drop off can be difficult.
 05/13/08, 11:37 AM #1409 Öwlcapwn Banned   Owlcapwn Tauren Druid   Korialstrasz I appreciate the feedback so far. I just noticed today that the spreadsheet has been updated for 2.4 content. With my current t6 set up it says I should pull more DPS then my t5 set up. I'll crunch some more numbers thru the day and hope I can clarify it for myself a little more. and the max dps rotation it suggests is MF x 2 SF x 13, i can only assume they mean MF, SF x 6-7, MF SF 6-7.
 05/13/08, 1:11 PM #1410 Celdhyrean Von Kaiser   Celdhyrean Night Elf Druid   Archimonde (EU) Treants and other pets will now automatically try to go behind the boss. Between the spot where you cast them and this new behavior, there shouldn't be any parry risk unless the mob spins around at times (to cast whatever on a raid member for example), and even then that risk wouldn't be higher than that created by other melee dps. You can use combat logs to verify whether there's any pary on their melee swings. They are really usefull if they can do their full 30s of dps and still OK even if they only last 50% (same +dmg, they did 3k for 15s against Kazrogal, for a GCD). If lower uptime, then don't bother.