Posted by: TheLastDeadMouse | January 5, 2010

Huh… that’s… odd.

WARNING: The following post contains data, math, and all around nerdery. Anyone who doesn’t not wish to be exposed to long winded mathematical rambling, click here.

Now then, the chart you’re looking at is the results from my initial Dr. Boom Growl testing, with the hunter’s RAP on the x-axis and the threat per Growl on the y-axis. At first it follows a pattern we’ve come to know and expect, a base value that’s used below a certain RAP value. With RAP values of 2189, 2592, 2911, 2994, and 3010, testing showed Threat values of 1548, 1560, 1551, 1550, and 1550 respectively, very strongly pointing to a base Growl value between 1550 (the median) and 1552 (the mean). Then, also as expected, after reaching the RAP threshold, somewhere between 3010 and 3029, a modifier kicks in increasing the amount of threat generated by Growl for each additional point of RAP.

What’s the disconnect then? The game is using the wrong formula to calculate Growl. It should be multiplying the RAP in excess of the threshold by some value, then adding that to the base threat. Using the data, the formula for threat with a RAP above 3020 SHOULD be

Threat = 1550 + (RAP – 3020)(1.425)
or simplified to
Threat = 1.425 x RAP – 2829.

Instead it appears that the formula used in game for RAP above the threshold is assuming the wrong base threat. The true formula being used is

Threat = 1258 + (RAP-3020)(1.425)
or simplified to
Threat = 1.425 x RAP – 3121

So what should we take from all this? First from a practical stand point, don’t have a RAP of 3020 to 3300. A RAP in this range would result in less threat from Growl than if you had a RAP of 3020 or less. If you have a RAP in this range, use it as an opportunity to switch out some AP for more stamina. Secondly, either we’re getting more more threat below the RAP threshold than we should, or 300-350 less above it than we should. I’ve created on thread on the official WoW bug forums which can be found here. Feel free to head on over and bump it to make sure we get this thing fixed.

Note: All numbers presented here are estimations with a certain margin of error and no number in this post should be used as a hard rule, just a general guideline.



  1. I love it when people figure things like this out, I could never wrap my head around it.

    Correct me if I am wrong though, but basically your results show that, above the threshold, for every 1 RAP you gain, growl does an additional 0.7 or there about threat?

    From what I am seeing in the data there looks to be an ever so slight upward trajectory going on here:


    [319/-9 = -35
    83/-1 = -83
    16/0 = 0
    19/-352 = -0.05
    24/40 = 0.6]
    49/67 = 0.73
    128/183 = 0.7
    273/387 = 0.7
    133/195 = 0.68
    362/518 = 0.69
    204/297 = 0.68
    167/220 = 0.75
    295/436 = 0.67
    444/652 = 0.68
    445/589 = 0.75

    It is small enough to just be noise I suppose.

    Anyhoo, I will take the 0.7 number and run with it, excellent to know this. But how about TS though 🙂

  2. Excellent stuff. I’ll be linking to this in a few days.

    Threat generation is the difficult line item now. Understanding all aspects of it will allow us to maximize threat.

  3. You’ve got that backwards Doug, for every .7 AP you gain after the threshold you gain 1 threat. Or for 1 AP you gain 1.45 threat, whichever way you want to look at it.

    Once you get above the AP threshold of ~3020, there seems to be constant scaling. If it looks like there’s an upward trend its probably just an artifact of the relatively small sample size.

    • Ah, trust me to get it wrong way around :), I was getting all depressed about it being the wrong way around and having a reducing absolute effect for every additional threat.

      Actually, 0.7 AP to 1 Threat is a great coefficient, much better than the other way. Yeah I looked at it again, it was just wishful thinking on my part. I am tempted now to go do a little run with Thunderstomp, you got me wondering.

      • Don’t feel too much pressure to do a series for Thunderstomp; I’ve got the day off again tomorrow and planned to test the hopefully simpler Thunderstomp threat formula.

        • I have actually started already, I probably won’t get a formula though, but it will be interesting to see if I results are similar.

          So far I have only one data point:
          3852 RAP
          251472 Total damage (-2000 for the start up auto-shot)
          207685 damage from auto-attack (83%)
          41537 damage from TS (16%)

          I am just going to run that at two more RAP levels and see if the percentage moves.

  4. I redid that test because I forgot to take down the dps…

    I got:
    3852 RAP
    287227 Total
    240088 Melee (83%)
    45987 TS (16%)
    320.8 dps

    This was on the IF Heroic Training dummy.

  5. Sorry to double post here but I have some more results in:

    3978 RAP
    251039 Total
    208007 Melee (83%)
    41895 TS (16%)
    304.5 dps

    What is odd here is that the dps went down here despite the 100 or so extra RAP. I was well above the hit cap both times and did not see any debuffs on the dummies that could account for the discrepancy (in fact I had to abort one run due to some random warrior doing a drive by sunder).

    4155 RAP
    253000 Total
    211216 Melee (83%)
    40541 TS (16%)
    315 dps

    Some rogue was putting up Savage Combat (+4% physical damage) during the second half of this test and I ran out of time to redo. It wasn’t very long so it may not have skewed the result too much.

    I will retest the 3852 RAP one later today when I get home, I hope I did something wrong :(.

  6. I did a much longer run at 3852 RAP and got a much more expected result:

    3852 RAP
    635858 Total damage
    524358 Melee (83%)
    107036 TS (16%)
    303.9 dps

    3705 RAP
    251816 Total damage
    208514 Melee (83%)
    42123 TS (16%)
    305.5 dps
    (in this one a pally ran past and put Heart of the Crusader up, which noticeably affected the dps for a while).

    3235 RAP
    251556 Total damage
    205922 Melee (81%) <——–
    45634 TS (18%) <——–
    279.9 dps

    Something tells me there is a limit here somewhere, it looks like diminishing returns kicks in above this level.

  7. One more set:

    2916 RAP
    250135 Total Damage
    201183 Melee (80%)
    47954 TS (19%)
    273.9 dps

    Everybody and there uncle’s stepsister were testing for this run.

    2526 RAP
    252595 Total Damage
    200491 Melee (79%)
    50214 TS (20%)
    258.5 dps

  8. Last one… (a compilation with some more)

    RAP dps Spread

    4155 – 315 – 83/16
    3978 – 304.5 – 83/16
    3852 – 303.9 – 83/16
    3705 – 305.5 – 83/16
    3235 – 279.9 – 81/18
    2916 – 273.9 – 80/19
    2526 – 258.5 – 79/20
    2273 – 265.1 – 81/18
    1932 – 230.8 – 79/20
    1642 – 220.4 – 78/21

    Hope that helps. Sadly it appears that TS scales worse than Auto-attack with RAP (unless I have it backwards again).

  9. I got inquisitive and did three more with higher RAP to see if the trend continued, I even finished off with a Flask of Endless Rage and an Intellect Scroll and Mini-Kings (only on me) for good measure:

    5302 – 351.4 – 84/15
    4782 – 346.9 – 84/15
    4392 – 320.5 – 83/16

    Is it just me or does it look like TS has no scaling with RAP whatsoever? To put in in another perspective, for the lowest set TS was doing 50k out of 252k, for the final one it was only doing 38k out of 251k.


    Here is the series with the dps portion attributable to TS:

    5302 – 53
    4782 – 52
    4392 – 51
    4155 – 50
    3978 – 49
    3852 – 48
    3705 – 48
    3235 – 50
    2916 – 51
    2526 – 51
    2273 – 47
    1932 – 46
    1642 – 46

    So it does scale, but only very weakly: From 1642 to 5302 you gain a whopping 7 dps (or 18 tps) from Thunderstomp.

    Wowser, please tell me I’m horribly mistaken. (and move this text somewhere else if you want, I kinda got carried away 🙂 ).

  10. Oh Doug, we’ve got a miscommunication here. I’m not looking for the damage formula for Thunderstomp. Other theorycrafters have figured all that stuff out long ago. What I’m testing for is the threat coefficient, specifically whether its a flat value or if its a percentage of damage dealt.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s


%d bloggers like this: