I can't find a model to match your test results. The most likely reasoning to me is:
1. Pet expertise is calculated prior to transferring hit
- This makes sense because if pet expertise was based on pet hit rating we wouldn't see this problem
2. Pet expertise for FA and hit rating are calculated separately
- This makes sense because there is a definite anomaly in calculations
3. Calculated Pet Expertise for FA and Hit rating are rounded down prior to being combined.
- This makes sense because it supports why you would see an anomoly like this
Unfortunately, modeling this together leaves your first example (232 hit rating and 1/3 FA) at 25 expertise which should be .25% dodge and your first sample is .5% dodge. Not to mention the other examples would have much lower dodge rates as well. My next best guess then is that expertise is either calculated completely differently for NPC's then it is for players, or their formula for converting FA to expertise is just wrong.
Also, regarding your thought on rating being rounded down, I'm not sure how this is possible since 8% hit would not be equal to 26 expertise anymore unless they are granting extra expertise rating for each hit rating.
While I have no evidence to support this, I suspect that pet hit rating isn't used in the calculation. I say this because in the past pets have been unable to do anything with rating stats - for instance when at level 70 we used drums that give everyone in the party (including pets) a haste rating buff, pets gained nothing from them since they had no math to convert the haste rating into haste - they just gained a buff to their haste rating while their attack speed stayed constant.
If this is still true (and it could be pretty easily tested with respect to haste again), it would mean that the issue is in the inheritance of hit->expertise directly, and not in the pet's expertise rating conversion. I suspect that the inheritance code just tries to give pets raw expertise (not expertise rating) based on our hit, and while the code that works off our hit rating does it correctly, the portion that works off our FA does not. Perhaps each part rounds resulting expertise down, and the resulting total is then lower than if they had been combined first, then rounded down? I'm not sure where things like Heroic Presence fit in.
I suspect that the inheritance code just tries to give pets raw expertise (not expertise rating) based on our hit, and while the code that works off our hit rating does it correctly, the portion that works off our FA does not. Perhaps each part rounds resulting expertise down, and the resulting total is then lower than if they had been combined first, then rounded down? I'm not sure where things like Heroic Presence fit in.
My theory is something like this -- a mathematical error in the hunter hit from Focused Aim to pet expertise code that is not present in the hunter hit rating to pet expertise code. It might involve rounding or it might be using the wrong % or rating number. Considering all the difficulty I had keeping the various hit%, hit rating, dodge% (and whether it was dodge avoidance or actual dodges), expertise rating, and expertise, along with what was rounded when, straight -- and I am still not certain that I have! -- it is no surprise if an error was made.
It wouldn't be the first time a fairly simple math problem was done incorrectly, as in Serpent sting getting the intended full damage for each tick! That was a big damage difference, so they fixed it right away. This is a relatively small damage difference, so they may never fix it, but the exact amount would be useful for the damage simulators. And if we can actually figure out where the miscalculation occurred, it may then be simple enough to fix.
I thought about putting this in the macro thread, but given the nature of why it's needed, the bug thread seems like a more appropriate place for it. I want to clear up some confusion regarding the current bug slowing down pet attacks based on some voluminous log comparisons of my pet attacks versus another almost exactly geared and specced Hunter also using a Wolf. These finding were consistent from fight to fight in any fight where both pets were active the entire time. I feel very comfortable with the results, which are also consistent with what Warlocks have observed. Test dummy results were not used as a stationary target in a non raid setting is less likely to provide a full example of what I believe the source of the bug is in play.
--------------------------------
Bite -
When macro'd and spammed increases the # of bites your pet will use in a given period, I have observed nearly a 20% increase in the number of Bites my pet was using in a mobile encounter with it macro'd into steady shot.
Furious Howl -
Observation 1: When macro'd and spammed does not increase it's uptime.
Observation 2: My uptime was consistently lower then my partners uptime on Howl with Bite macro'd, although not significantly so. Normally this was a fraction of a percent lower. I suspect the slightly lower uptime is a result of the pet being occasionally focus starved from using bite more often. Another possibility was that the Howl cast was being delayed due to Bite's GCD, which is why I did my next test.
Observation 3: Macro'ing Howl before Bite did not impact the results significantly enough to draw any conclusions from.
---------------------------------
Results:
Macro'ing Bite into your shots will increase pet DPS, however it may cause you to take a very slight hit(less then 1%) on Furious Howl uptime. Additionally, having pet abilities in the macro seemed to increase the number of errors I was getting i.e. '...not ready yet' type of errors. I utilized a variety of ways of applying script UI Error message clear, or UI Error message on/off, and disable/enable SFX commands, but was unable to remove the messages entirely. I ended up opting to disable error messages in my steady shot macro with Bite, and reenabling the messages in my Chimera macro. The following line can be added to any shot macro to accomplish manually casting bite.
/cast [@pettarget, exists] Bite
As Warlocks have noted, the reason for the delay is because of the range check pets need to make when using abilities that involve attacking the target. Sometimes your pet may be failing the range check despite being in melee range just because of latency, and this check does not appear to be happening continuously so a failed range check delays the time to your next bite. This bug and the Warlock reasoning is supported by my observations that Bite was affected by this bug significantly while Howl was not. I have not run any comparisons yet with a ranged attack pet. A logical conclusion could be that ranged attacks will still be affected by this bug, although at a reduced rate since a ranged attacker is less likely to fail range checks due to latency, but there are no test results for ranged attacks from Hunter pets I've seen yet that support any conclusion.
I'm not sure where things like Heroic Presence fit in.
I'm pretty sure the reason that Heroic Presence doesn't transfer experitse to pets is that the pet doesn't get hit from their owner from it, but rather from being in range of the buff themselves (I haven't tested this personally but it makes logical sense since the pet gets the buff too). Since the pet receives the hit directly there is no automatic conversion of hit to expertise, since it seems that the hit to expertise conversion happens on the hunter end and not the pet end.
Maybe I'm beating a dead horse here but I wonder if anyone has already investigated how exactly Auto Shot works when moving around. Here are my observations so far:
when starting to shoot something, an UNIT_SPELLCAST_SENT event is fired for Auto Shot.
this is followed by UNIT_SPELLCAST_SUCCEEDED for each Auto Shot that is fired (including the first).
if an Auto Shot should fire, but you are moving UNIT_SPELLCAST_FAILED_QUIET is fired each 0.5 seconds (assuming you are still facing the target. If you are not facing it, this event will not fire).
If you stop moving "too late", the Auto Shot will be delayed by 0.5 seconds (the time from the last UNIT_SPELLCAST_FAILED_QUIET to the next UNIT_SPELLCAST_SUCCEEDED is always 0.5 seconds).
So the question is what "too late" is, when you want to avoid delaying the next Auto Shot. It appears to depend on latency.
Given the possible 0.5 second delay and depending on latency, it might be a good idea to look for a Auto Shot swing timer with a "save zone" like castbars have (if the timer reaches the "save zone" you have already delayed the next Auto Shot).
Edit: I've uploaded the addon i used to test this to wowi, if anyone wants to verify these findings (link)
As of today, post patch, I am noticing a 9% crit from unbuffed pet on the heroic training dummy. As far as I know, the pet comes with 5% crit, to which the 9% coming from Spider's Bite should be added. Am I wrong, or is there a bug here ?
I finished extending my test sets using Focused Aim to 10K+ and as a result have somewhat more refined results. (See data here.)
Taking a somewhat different (and hopefully more enlightening) tack to interpret the results than I did in my original post, I come up with this:
Taking the results from the first test sets, 7% hunter hit + 1/3 FA, shows that we appear to be getting 0.46% pet dodges for one point in Focused Aim. Thus with 7.08% hit rating and 1/3 FA, we are getting 6.04% dodge reduction rather than the full 6.5%. This would suggest we are missing approximately 2 pet expertise, or 16.46 (8.23*2 = 16.46) expertise rating. According to Expertise Rating Calculator, 6.04% dodge reduction is accomplished by an expertise rating of 198; if the cap is 214, then we are missing 16 expertise rating at 1/3 FA. Assuming 26.75 expertise rating for each point in FA would cap our pet, then we are actually getting 10.75 expertise rating per point in FA. 10 (rounded down) expertise rating gives 0.3% dodge and parry reduction.
If we add 0.3% dodge (from FA) to the 0.46% dodge we actually experienced in the test, we come up with 0.76%, very close to the predicted result of 0.79% dodge without any FA.
Let's take the next step and look at what we are getting with 2 points in FA. Assuming that we get 10.75 expertise rating from the second point as well as from the first, we get 21.5 -- this rounds to 21.
With 6.04% hit rating and 2/3 FA, test results show 5.51% dodge reduction rather than 6.5%. This would suggest we are missing approximately 4 pet expertise, or 32.92 (8.23*4 = 32.92) expertise rating. According to Expertise Rating Calculator, 5.51% dodge reduction is accomplished by an expertise rating of 181; with cap of 214, we are missing 33 expertise rating with 2/3 FA. 21 (rounded down) expertise rating gives 0.64% dodge and parry reduction. 0.64% dodge reduction (from 2/3 FA) plus 0.99% (from test) equals 1.63%, even closer to the predicted result of 1.62% without FA.
Now for 3 points in FA: With 5.12% hit rating and 3/3 FA, test results show 5.27% dodge reduction, suggesting a missing 5 pet expertise, or 41.15 (8.23*5 = 41.15) expertise rating. According to Expertise Rating Calculator, 5.27% dodge reduction is accomplished by an expertise rating of 173; with cap of 214, we are missing 41 expertise rating with 3/3 FA. Assuming 10.75 expertise rating from each of three points in FA (10.75*3 = 32.25), rounded down, 32 expertise rating gives 0.98% dodge and parry reduction. 0.98% dodge reduction (from 3/3 FA) plus 1.23% (from test) equals 2.21%, 0.23% lower than the predicted result of 2.44% without FA.
Our predictive model would have expected to find 6 rather than 5 missing expertise at 5% hit + 3/3 FA, and this also would have been in keeping with the test results from 7% hit + 1/3 FA and from 6% hit + 2/3 FA.
Finally, we can look at the overcapped result, 6.04% hit + 3/3 FA. In this test we were experiencing 5.96% dodge reduction, accomplished by 196 expertise rating. We should be getting 157.5 expertise rating from 6% hit, leaving 38.5 that we were getting from 3/3 FA. 38 expertise rating gives 1.16% dodge and parry reduction. Add the 0.54% dodge reduction from the test with 1.16, and the result is 1.70%, fairly close to the 1.62% predicted for 6% hit with no FA present.
It almost seems like, in addition to the rounding issue making the second point worth a bit more than than the first, the third point in FA is actually giving more expertise than either of the first two. If the first two points give 10.75 each, then the third point would have to give 16.5 expertise rating to account for the results of the overcapped test result (interestingly, almost exactly 2 expertise' worth).
If we go back to the 5.12% hit + 3/3 FA results and apply this idea, we get this:
With 5.12% hit rating and 3/3 FA, test results show 5.27% dodge reduction, suggesting a missing 5 pet expertise, or 41.15 (8.23*5 = 41.15) expertise rating. According to Expertise Rating Calculator, 5.27% dodge reduction is accomplished by an expertise rating of 173; with cap of 214, we are missing 41 expertise rating with 3/3 FA. Assuming 10.75 expertise rating from each of the first two points in FA (10.75*2 = 21.50), plus 16.5 from the third point, we get 38 expertise rating for all three points. 38 expertise rating gives 1.16% dodge and parry reduction. 1.16% dodge reduction (from 3/3 FA) plus 1.23% dodges (from test) equals 2.39%, very close to the predicted result of 2.44% without FA.
10.75 expertise rating for each of the first two points in FA and 16.5 for the third is an inelegant solution to say the least, but it does seem to fit the test results (allowing for some RNG). The first two points seem to be *missing* 2 expertise each (giving 1.25) while the third point seems to *give* 2 expertise (missing 1.25). None of the points give as much expertise as they should to cap our pets, that is, 3.25 each.
I have no idea what sort of logic might have led to such a mistake. I remain hopeful that someone else has a better model to explain the test results.
As of today, post patch, I am noticing a 9% crit from unbuffed pet on the heroic training dummy. As far as I know, the pet comes with 5% crit, to which the 9% coming from Spider's Bite should be added. Am I wrong, or is there a bug here ?
You need to clarify and post your info. The crit rate will be reduced by the level difference between the target and source, however it should be closer to a 3% loss. What kind of volume of attacks did you use?
10.75 expertise rating for each of the first two points in FA and 16.5 for the third is an inelegant solution to say the least, but it does seem to fit the test results (allowing for some RNG). The first two points seem to be *missing* 2 expertise each (giving 1.25) while the third point seems to *give* 2 expertise (missing 1.25). None of the points give as much expertise as they should to cap our pets, that is, 3.25 each.
I have no idea what sort of logic might have led to such a mistake. I remain hopeful that someone else has a better model to explain the test results.
Its a stretch, but this is the best I can come up with.
Hit rating is converted to expertise at .0991155 per point(213.122 / 262.32 / 8.197)
Focused Aim should grant 3.33~ expertise per point, cap subtracting expertise granted by 5% hit[(26-16)/3], but instead is granting half that.
Expertise is rounded for each separately after conversion to pet.
Exp from hit/Exp from FA - Missing expertise - results
26/0 - 0 - actual results 0
23/1 - 2 - actual results 2(.46%)
19/3 - 4 - actual results 4(.99%)
16/5 - 5 - actual results 5(1.23%)
19/5 - 2 - actual results 2(.54%)
The main question here surrounds the 232 hit rating = 23 expertise as it is borderline and based on a reasonable doubt that the math was inaccurate enough to account for 22.994796=23. I'll see if I can find a way to repeat the results with exactly 230 hit rating to verify.
Its a stretch, but this is the best I can come up with.
Hit rating is converted to expertise at .0991155 per point(213.122 / 262.32 / 8.197)
Focused Aim should grant 3.33~ expertise per point, cap subtracting expertise granted by 5% hit[(26-16)/3], but instead is granting half that.
Expertise is rounded for each separately after conversion to pet.
I was using 26 expertise cap divided by 8 hit percentage points = 3.25 pet expertise per 1% hunter hit -- by using exact hit rating to exact expertise rating as you have done, it looks even more like rounding might be the culprit.
Or, it could be something as simple as the coder seeing that each point of FA should grant 3.25 expertise, and interpreting that as 0.325% dodge reduction instead -- if decimals were allowed, 10.75 expertise rating would work out to around 0.325% dodge reduction (10 expertise rating = 0.30% dodge reduction, 11 expertise rating = 0.34% dodge reduction).
It still doesn't explain the results for the 5% hit + 3/3 FA results though, unless I just had terrible RNG on my testing. I think you may be on to something with the extra points of hit rating (232 vs 230), although I had thought the rounding down for pets would have made 5.12% hit the same as 5.0% for the purpose of transferring to pets.
Dispels like Tranquilizing Shot are on the spell hit table, so you'd need 17% spell hit/445.94 hit rating or 14% spell hit/367.25 hit rating with Misery/Faerie Fire to guarantee a dispel on a level 83 mob (what level are the shambling horrors?). I'm not sure, but I doubt Focused Aim grants spell hit, so your 179 hit rating only gives you about ~6,8% increased chance to hit with spells, which means 7 to 10% (depending on the presence of FF/Misery) of your Tranqulizing Shots will fail.
Has anyone else noticed autoshot getting reset like crazy on LK?
I was on 10 man LK tonight and spamming my normal steady shot macro and noticing my quartz bar showing my auto shots getting reset 2-4 times per steady shot. So I stopped shooting all together to see if auto shot would go off as normal...and nope...same thing. So after raid I popped over to org to the test dummy...and viola...no issues what so ever.
Has anyone else noticed autoshot getting reset like crazy on LK?
I was on 10 man LK tonight and spamming my normal steady shot macro and noticing my quartz bar showing my auto shots getting reset 2-4 times per steady shot. So I stopped shooting all together to see if auto shot would go off as normal...and nope...same thing. So after raid I popped over to org to the test dummy...and viola...no issues what so ever.
Was there another hunter in your raid? Quartz is currently bugged that if there's another hunter in the same group/raid as you, their autoshots resets the swing timer. In my LK10 group I didn't notice autoshot messing up, so I suspect its the quartz bug you're experiencing.
Has anyone else noticed autoshot getting reset like crazy on LK?
I was on 10 man LK tonight and spamming my normal steady shot macro and noticing my quartz bar showing my auto shots getting reset 2-4 times per steady shot. So I stopped shooting all together to see if auto shot would go off as normal...and nope...same thing. So after raid I popped over to org to the test dummy...and viola...no issues what so ever.
It is caused by the seasonal event perfume/cologne buffs (75 AP flasks). When someone with perfume is standing too close to someone with cologne, their autoswings stop working. We noticed this in our guild on Blood Queen, where lots of melee reported they couldn't use any melee ability. I myself noticed it the same night on Val'kyr Twins, but only after the fight when I looked at my measly 10k dps. It turned out Autoshot only had done 7% of my damage (and as a result that meant less IaotH procs, less Wild Quiver, less Zod procs, less focus for pet etc. The reason I wasn't affected on Blood Queen but affected on Twins was the fact that on BQ I am standing far away from people, yet on Twins we were all hugging.
They disabled the aura on the perfumes/colognes during or around Tuesday's maintenance as far as I know. I've been using the buffs along with others, and the aura has not appeared once during this week's reset.
When the LK kills the raid at 10%, then resses it after the RP event, he seems to neglect ressing hunter pets.
This is a fairly minor annoyance, since (on non-heroic at least) the fight seems to be a free kill at that point, but it does hurt our DPS by around 1K in that phase, unless you happen to have specced heart of the phoenix for some reason.
I'm unsure if your pet's corpse is visible and eligible for a druid battle res at that point.
Was there another hunter in your raid? Quartz is currently bugged that if there's another hunter in the same group/raid as you, their autoshots resets the swing timer. In my LK10 group I didn't notice autoshot messing up, so I suspect its the quartz bug you're experiencing.
I can second that the cause of this bug is due to Quartz counting any hunters auto shot as your own. I and another hunter in my 25 man group noted this bug before the holiday event began. One of use would start dps on trash and as soon as the other hunters began to attack our quratz would quickly start reseting our auto shot timers.
When macro'd and spammed increases the # of bites your pet will use in a given period, I have observed nearly a 20% increase in the number of Bites my pet was using in a mobile encounter with it macro'd into steady shot.
Macro'ing Bite into your shots will increase pet DPS, however it may cause you to take a very slight hit(less then 1%) on Furious Howl uptime. Additionally, having pet abilities in the macro seemed to increase the number of errors I was getting i.e. '...not ready yet' type of errors. I utilized a variety of ways of applying script UI Error message clear, or UI Error message on/off, and disable/enable SFX commands, but was unable to remove the messages entirely.
In a post i read recently, a hunter informed that priest's skill Renewed Hope makes GCD of pet skill Bite started. Maybe it's the reason why such error message ocurred.
Has anyone else experianced their pet running to attack LDW while pulling the trash in her room? Two times now (on 25man reg and 10m H) my guild was clearing trash in her room and instead of attacking my current target, my pet made a blitz to LDW and aggroed her.
02/01/10, 1:36 PM
Similar Threads
