Warcraft Tank Threat Geeking

[bryant]

I just released my first package to the WoW community, or at least the portion of the WoW community which cares about tank threat and understands UNIX/perl.

4 Tanks and a Void Reaver



I don't think that worked out very usefully, since I charted aggregate threat per second over the entire fight; charting the 10 second rolling average was too choppy. But the point is hopefully clear.

Release announcement with full details here.

Comments

[shiffer.livejournal.com]

This is a chart of TPS over time?

It seems a bit wonky to me for a VR fight. The MT switches means that each tank will have significantly more rage during certain portions of the fight, which should (and did, in the several VR fights I tanked) increase their TPS considerably.

Moreover, three of the four tanks reached their max TPS during the second minute, and then stabilized at a lower point. The only thing I can think of that will account for this is single Bloodlust or a Windfury totem that didn't get renewed. Both assume that the shaman is dead, otherwise the log makes no sense.

[bryant]

Note that it's aggregate TPS, which means that the fluctuations which should be created by the rage differential are smoothed out, and are more smooth the later we get in the fight. If I chart the ten second aggro you can see the spikes more clearly, but the data is too choppy to be useful. I should try a 30 second average chart.

My lack of experience with data presentation is definitely showing. :)

Also, Ghoselle is a druid, so less subject to the limits of incoming rage. On the other hand... I don't actually find myself super-starved for rage as an OT on that fight, and I'm always a bit perplexed when people say they are. I always have enough rage to keep up a Shield Slam, Devastate, Devastate, Devastate cycle; I usually can manage to get some Heroic Strikes in there as well. Pounding generates a ton of rage for me.

I am spec'd into Anger Management, Imp Heroic Strike, and a smidge of Unbridled Wrath. Which helps.

Linnet, the low tank, had aggro during the period in which the other three tanks are going from their peaks downward, and that's early enough in the fight so that the effect of less aggro will be more notable in the graphs.

Linnet and Nurgh both had a shaman, who did in fact die pretty early on.

[shiffer.livejournal.com]

Ah, -aggregate- TPS. That makes a lot more sense.

My experience is that no one is rage-starved on VR, but in general there's still a noticeable difference between just getting AOEed, which produces respectable amounts of rage, and actually having the big guy maul you in person, which should give you infinite rage. The difference is usually 100-200 TPS.

This is a very interesting tool. Was there something specific you wanted to check when you wrote it?

I'm asking because while I'm sure there are analytical uses to this graph, I think I'd actually be more interested in the non-aggregate TPS over time one, which also seems to be the "natural" one to start with.

[dcltdw]

*cheers!*

For some reason, that seems really cool, in a really geeky way. :)

[bryant]

The tool also provides threat generated in the current second, 5 second average TPS, and 10 second average TPS:



That's me; I started out tanking, took a knockback around 150 seconds, picked it up again, took another knockback relatively quickly, and went into Execute mode at the end to beat the enrage timer.

Mostly I wanted to analyze threat curves for our various tanks, since it's been a possible problem spot. I expect I'll be adding in data collection for damage taken next, that being the other thing I care a lot about in a tank.

[shiffer.livejournal.com]

I doff my hat to you sir.

I'll definitely take another look at this if I ever take a position of leadership in a raiding again.

[michele-blue.livejournal.com]

Grats, sweetie. That's awesome!

[head58.livejournal.com]

I have absolutely no idea what any of this means but I truly admire the statistical and data-troll aspect.

[solarmus.livejournal.com]

Hmm, I assume we'd see the most periodic spikiness if we had a fury warrior as one of the tanks who was swapping between sword and boards and DW.

Otherwise the threat plateau level is useful info, especially for comparing gearsets/ability rotations.

[bryant]

Yeah, I think so.

The big problem there is that it won't catch most of the non-white damage threat generated by a Fury warrior -- it can't detect Sunder Armor. Irksome.

[ladyjestyr.livejournal.com]

Le awesome; I'm definitely going to hook my guild's tanks up with this. *attaches wires and electrodes to them*

[indigo.livejournal.com]

Have you tried parsing a VR fight where Dayna was tanking?

[bryant]

Not yet.