Llilyth

Just to clarify, I'm not talking about spikes, which I don't mind at all as long as the numbers return to correct ranges after a while. I'm talking about sustained numbers that are way too high. I don't know for sure how long it takes to kill Fathom-Guard Tidalvess, but it's certainly not a few seconds like a trash mob. My TPS being over 2000 for the majority of Tidalvess's life is incorrect methinks.

Long live Omen!

 

Marroc

Just an out of the blue question for you Llilyth... Are you sure that your TPS number isn't accidently overlapping part of your threat% number?

That is to say if you have omen rather small (width wise) you can end up with something like this

@ = TPS
# = Threat %

Overlap: @@@###% (causing you accidentally miss read it as something like @@@@ ##%)
Normal:@@@ ###%

 

Apate

I was hoping for "pseudo-stuck" actually :P

Moreso, I wanted to provide information because I don't know what is helpful and what isn't. I know with Discord Action Bars (pre-2.0) I could use the mouseover setting to get something permanently stuck to my cursor and it was incredibly annoying (and easily repeatable by following the logical steps to make a mouseover minibar).

See you, auntie.

 

Marroc

Well the good news is Antiarc seems to have an idea on how to fix this pseudo-sticking :P

 

Llilyth

Nope, I turned off the Threat % column.

Omen is most definitely telling me I'm a super uber fantastic tank, which is doing wonders for my e-peen. ^_^

 

Antiarc

afk, renaming Omen to eViagra.

 

Marroc

While You're at it, can you add in the little 'you' and 'agro holder' arrows back in like I asked about a few pages back :P

No rush of course, just don't want you to forget ^__^

 

Antiarc

Heh, sure will. Sorry - I had forgotten about that.

(Bad developer!)

 

Marroc

An option to choose what side of the bars they show up on would be awesome as well (didn't have that when they were there initially, and I have omen shoved up against the edge of my screen).

 

Roana

I've done some experimenting with the TPS calculation by extracting it and running it with some artificially constructed sequences of numbers, and the results don't look terribly useful to me. For example, when doing 400 threat every four seconds, TPS bounced between 120 and 80, which on average is the expected value of 100, but not exactly useful information (plus/minus 20%, depending on when you're looking at the screen). Simulating a Shield Slam/Revenge/Sunder/Sunder rotation + auto-attack also didn't produce terribly accurate numbers. Also, for the beginning of the fight, TPS is consistently underestimated (and it's pretty obvious why, looking at how it's calculcated -- it needs quite a few data points to build up a reliable estimate).

 

Antiarc

Yeah. I'm in the process of reworking it now, having produced similar results after pulling it out. It tends to either be too jumpy, or decay too fast (producing low numbers).

Any algorithm suggestions? I would, ideally, like to do it while storing as little history as possible, which is why exponential smoothing was originally attractive. I may just end up with a moving average, though.

 

yoshi

I was wondering what our guild should get now, KTM or Omen? It sounds like Omen still has some bugs that need to be worked out, so should we just wait until it is in it's final stage of development?

 

Rinced

Maybe make the pullout bar resizable/scaleable - or if that option is available make it reachable through the tooltip.

 

Marroc

You can change it's size... Not sure if it's just click and drag like I'm thinking it is, or if you need to do it through the rclick menu.

 

CD

My biggest problem with Omen is in multi- mob situations where healing aggro seems to be enormous. If there are several healers erroneously at the top, then a tank, then dps below it's much easier for dps not to notice what has happened and overaggro.

I notice it most in Vashj and Kael'thas phase 2- it seems to be a serious problem and I'm not sure if other people have similar experiences - I'm using the default settings.

 

Llilyth

Threat r43671
Now the TPS calculation is giving numbers that are way too small. It's giving me TPS numbers like 50 on a boss.

Threat r43667 seemed to give good numbers though, around 1000.

 

Goreshot

I wouldn't bother trying to do anything fancier than a simple moving average for TPS calculations. Unlike DPS, a tank doesn't care how much threat she produces per second over a long fight. Current fight design is such that IF she even cares about the TPS value then it would be over the short term, i.e. on fights where there are aggro mechanics such that she only refers to TPS as an indication of how well she is "keeping up" with threat over a relatively short period of time.


If a tank really cared about how much threat she produces over the long term (like, say, during Phases 2 and 3 of Magtheridon) she would simply take the total threat number and divide by the duration of that fight. Any long-term threat value of shorter than 1 minute, especially when looking at current fight mechanics, is very likely statistically meaningless. Unless you're doing real work like averaging out the threat value of a certain phase in a certain fight over multiple attempts, you aren't going to find anything (statistically) significant in the data. And really, no one bothers doing this type of work for WoW tanking, because it's not necessary (any idiot can look at a threat meter and adjust his playstyle accordingly, regardless of who is tanking).


Weighted/exponential/etc. moving averages are best for when you are tracking at least one predictable quantity of some kind, so you can set the variables accordingly. (Yes, I know, moving averages by definition are used in noisy or unpredictable systems - what I am referring to is the predictability that comes with knowing WHEN the data is coming or HOW LARGE the data will tend to be; if you know one of those values then you can weight the algorithm accordingly. Otherwise you have no REASON to weight the data, which means you SHOULDN'T be weighting it.) Fights are variable enough that a weighting scheme for one fight will be almost useless for another, especially when aggro wipes/swaps/etc. are taken into account.

Unless you intend on analyzing the periodic schema of every raid fight and coding separate algorithms for each period (which is basically what a threat meter that takes into account aggro mechanics in raid fights DOES), you're better off just ignoring the whole thing altogether and going with the last 10-20 seconds. Really, a tank just cares about whether or not she is keeping up a certain pace required to allow DPS to do their job during certain phases. All this requires is a simple running average, since anything shorter than that is easily discerned via the basic threat list and anything longer than that is likely to be modified by some sort of aggro mechanic anyway (i.e. aggro wipe), so there's no need to weight "old" threat. All that does is make the TPS value too "jumpy."

"User is a tremendous douchenozzel"
-Actual EJ Forums feedback concerning Goreshot.

 

Antiarc

I was dividing once too many times. It's fixed now. And it's a moving average of the last 15 threat events

 

Rallick

The tank for the Hydross weirdness had omen #42302 and threat #42277.

 

Antiarc

That threat version is quite out of date; have him update if you can, and see if it goes away.

 

sp00n

To the TPS issue, yesterday on nightbane I had negative DPS.

At the start of the fight where I had 0 damage done:
http://sp00n.pytalhost.com/misc/omen_bug_tps1.jpg

During the fight, when I feinted:
http://sp00n.pytalhost.com/misc/omen_bug_tps2.jpg

And another issue with the pullout bar, after the adds were down and Nightbane was about to land:
http://sp00n.pytalhost.com/misc/omen_bug_bar.jpg


I was the only one running a recent Omen/Threat version, and only one person, Eferra, was also running Omen, although with quite an outdated one (which didn't clear threat when Nightbane landed *sometimes*).

 

Gorce

Hello, one of the tanks in my guild is having a problem dropping off the list still. We are on a US server and no one is running a EU client or anything like that. I have read through 50+ pages and the problem has been mentioned but the resolution has not. I haven't figure out how to reproduce the event yet, but I have noticed it on our Magtheridon, Hydross and while we were testing the meter in Blasted lands. Its simply as stated 10 pages ago, he generates roughly 30k threat then it just stops and he'll get 5k added here or there but not enough to even stay on the list. On his end he is reading normally, he's at the top and tanking happy as a clam and can tell us when we are closing in on him.

Now, he has updated to the newest version at least 6 times while we've been trying to figure this out. When he and I went back to blasted lands to try to reproduce it the mod worked perfect. No matter what we tried the two of us couldn't make it happen. So I am stumped. Please any help. about 70% of the raid is using Omen now and he is the only one with any issue. But being a main tank its a big issue atm.

 

Marroc

ENSURE that he is updating THREAT. It's a library now, so if you're running pitbull or anything else that uses it make sure you either copy omen's copy of threatLib to the main addon directory, or disable other addons that use threat.

As far as I know, the threat 'capping' on tanks was fixed a while ago (I've also not seen it for at least a week or two worth of raiding myself).

 

Demi9OD

Any chance of being able to drag the Agro gain bar out like you can with a tanks? Would be pretty useful on fights with tank transitions or when one of our Ferals is tanking something so I can't have to keep my eyes on peeled on the meter. Also, player and agro gain being the same color is a bit distracting, but I supposed that can probably be modified (not home to check).

 

Apate

Not exactly it, but have you checked out the Threat component of Aloft? It adds an aggro-gain bar to enemy nameplates. Admittedly, I haven't really taught myself to watch it, yet.

See you, auntie.