Jump to content


Photo

Cataclysm Mechanics Testing


  • Please log in to reply
322 replies to this topic

#21 ieatpaperbag

ieatpaperbag

    Piston Honda

  • Members
  • 230 posts

Posted 01 October 2010 - 07:20 AM

10. Verify the IP/WP proc rates. I have no particular reason to believe it's changed, but it's worth verifying. Similarly, verify that the proc rate on ranged attacks (i.e. FoK) is what you'd expect.
21. Is FoK normalized?


Sorry I kind of doubled up on the confirming that FoK is not normalized, I'll skip explaining that part.

It appears that the proc rates for IP/WP are the same on beta as on live, at least for ranged weapons.

With Instant Poison on a 3.00 ranged weapon I expect to see 42.86% proc chance (ignore those melee attacks, there was no poison on that weapon)
Posted Image

With Instant Poison on a 1.80 ranged weapon I expect to see 25.71% proc chance.
Posted Image

With Wound Poison on a 3.00 ranged weapon I expect to see 100% (107.14%) proc chance.
Posted Image

With Wound Poison on a 1.80 ranged weapon I expect to see 64.29% proc chance.
Posted Image

#22 Docrev

Docrev

    Piston Honda

  • Members
  • 216 posts

Posted 01 October 2010 - 07:49 AM

Grinding through the data a bit, I think the more elegant way of expressing the damage formula is 302.5 + .1125 * AP, which is increased by Vile Poisons and Master Poisoner to 392.04 + .1458 * AP. This gives perfect agreement with your data in all cases (assuming, at least, your metagem was unequipped or inactive for the 2036 AP test).

The helm was unequipped for the 2036 and 616 attack power trials, so that agrees with your model.

#23 ieatpaperbag

ieatpaperbag

    Piston Honda

  • Members
  • 230 posts

Posted 01 October 2010 - 08:47 AM

1. How much armor do bosses have?
20. How does Imp SS stack with Aggression? How do these interact with the bonus damage on Sinister Strike?
35. Are glancing blows still 24% chance and average-case 25% damage reduction?


With a 2.5 speed 7.4 dps Axe and 5099 AP I expect normal attacks to land for about 929 (I can probably get a larger sample if needed but there were boomkins in every city putting up FF)
Posted Image
The boss level dummy has approximately 31.43% damage reduction from armor. It also appears that glancing blows still have a 24% chance to happen and reduce damage by 25%.

With no talent points and using the same axe
Posted Image

With Imp SS only (5874 AP now with Vitality)
Posted Image

With Aggression only
Posted Image

Both Aggression and Imp SS
Posted Image
(2.5 * 7.4 + 5874 * 2.4 / 14 + 200)*(1 - 0.3143) * (1 + 0.15 + 0.30) = 1217
Aggression and Imp SS appear to be additive.


27. Which finishers count in terms of Restless Blades?

I am not sure how to give "proof" but I can confirm that on beta the following

Count for Restless Blades
  • Eviscerate
  • Envenom
  • Deadly Throw
  • Rupture

Do not count for Restless Blades
  • SnD
  • Expose Armor
  • Recuperate
  • Kidney Shot


23. What about Revealing Strike?

Revealing Strike does not look like it's normalized.

Using the same 2.5 speed 7.4 dps Axe and 5864 AP on the boss level dummy
Non-normalized:
(2.5 * 7.4 + 5864 * 2.5 / 14) * 1.25 * (1 - 0.3143) = 913.33
Posted Image

Using a 1.9 speed 6.5 dps Mace and 5864 AP on the boss level dummy
Non-normalized:
(1.9 * 6.5 + 5864 * 1.9 / 14) * 1.25 * (1 - 0.3143) = 692.67
Posted Image


25. How does the new Vitality damage buff stack with Savage Combat?

It is currently multiplicative on beta.

No talents
Posted Image

With Vitality
Posted Image

Both Vitality and Savage Combat
5099 * 1.15 * 1.04 = 6098.4
Posted Image

#24 Lightshadow

Lightshadow

    Piston Honda

  • Members
  • 133 posts

Posted 01 October 2010 - 02:36 PM

It seems to me that your second test confirms that AR stacks additively with one of Vitality and Haste Rating and additively with the other, but does not distinguish which is which - that is, the gains from AR could be interpreted at 150 * 1.25 = 187.5 ~= 195, but could also reasonably be interpreted is 150 * 1.3257 = 198.9 ~= 195. A repeat of this experiment with 0 haste would be illuminating.


Good catch. Repeated said experiment with only a white throwing weapon equipped (0 haste). 48 FoKs and 70 energy remaining at the end of 2 minutes. Net regen 1650. 1650 - (12.5*120) = 150. It would seem like the scaling factor here is haste not vitality.

#25 niemann

niemann

    Glass Joe

  • Members
  • 4 posts

Posted 01 October 2010 - 03:22 PM

Regarding Potent Poisons and Vile Poisons stacking, testing with instant poison suggests that they stack additive.

Average of around 250 instant poison hits with:
With no talent spec: 797
20% from Vile Poisons: 952
20% from Vile Poisons and 20% from Mastery: 1104.

1104/952=1.16.

It is also worth noting that Envenom doesn't benefit from Mastery directly.

#26 ieatpaperbag

ieatpaperbag

    Piston Honda

  • Members
  • 230 posts

Posted 01 October 2010 - 03:41 PM

28. Are existing DoTs (DP, Rupture, etc.) boosted by the KSp damage buff?


It looks like existing DoTs are not boosted by the 20% KSp damage buff. In this test I ruptured and swapped to a weapon with no DP on it just before a KSp.

Posted Image

#27 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13510 posts

Posted 01 October 2010 - 07:06 PM

I'd like to take a few minutes to discuss a testing technique that's helpful for getting detailed results. It involves the behavior of fixed damage attacks - i.e., things like Gouge, Rupture ticks, Venomous Wounds procs, etc. which always obey the same damage formulas. When you're actually testing such an attack, it will exhibit a damage range exactly one point wide - for instance, it will do 123-124 damage in practice. This is because, when the game calculated the damage for that move, it will find that it does (say) 123.4 damage. At this point, the game rounds randomly up or down based on the fractional part; in this example, it will round down 60% of the time and round up 40% of the time, so the actual average over the long term will be 123.4, but each individual attack will do either 123 or 124.

Hence, if you test once or twice and find that it does 123 damage, that's somewhat useful as it tells us that the internal damage number is between 122 and 124; but if you test until you get both values (123 and 124 in this case), it tells you that the answer is between 123 and 124, which basically halves the uncertainty.

Note that this is true for both crits and hits; for instance, if you look at the second Venomous Wounds test, we know both that the base damage is between 1002 and 1003, and that 1.545 times the base damage is between 1548 and 1549, which lets us state with absolute confidence that the actual damage number is between 1002 and 1002.59, which is better still. Hence, when testing this sort of thing, the most descriptive answer you can possible give is to get both damage number for both crits and noncrits; this gives us the most possible confidence in the answer.

However, sometimes the value comes out very very close to an actual damage number; for instance, in the first Venomous Wounds test above, the expected damage is 1.2 * 1.2 * 1.08 * (302.5+.1125*4347) = 1230.99912 damage. Hence, 99.9% of the time, it's going to do 1231 damage, meaning you can beat on the dummy all day and not see a 1230 - you'll note that he did 53 trials and all did the same amount. And while 53 tests with venomous wounds isn't that hard to do, once you start testing with Gouge... you won't *always* be able to get both numbers for a given attack. However, if you can at least make a reasonable effort to get them - 6 or 8 trials of a given attack - that does help reduce the uncertainty involved. More is of course better, but we'll take what we can get.

I bring this up because while ieatpaperbag's testing in post 23 is indicative that target armor is about 11950, I think it would be good to narrow it down further; in particular, some testing with a fixed-damage attack (gouge for rogues - other classes have other options) would be very helpful towards pinpointing the actual value.

Regarding Potent Poisons and Vile Poisons stacking, testing with instant poison suggests that they stack additive.

Average of around 250 instant poison hits with:
With no talent spec: 797
20% from Vile Poisons: 952
20% from Vile Poisons and 20% from Mastery: 1104.

1104/952=1.16.

It is also worth noting that Envenom doesn't benefit from Mastery directly.


Interesting - particularly since Venomous Wounds implies the opposite. I wonder if the damage range on IP is dirtying the results? Regardless, we'll need more testing - we should probably look at WP and DP in addition to Venomous Wounds and IP to try to get the full picture. As a note to people testing this: beware of Master Poisoner.

It looks like existing DoTs are not boosted by the 20% KSp damage buff. In this test I ruptured and swapped to a weapon with no DP on it just before a KSp.


Interesting. Could you also confirm with Rupture, and perhaps try refreshing DP during KSp? That is, if you have a rolling DP stack and refresh it during KSp, does the following tick do 20% more damage? This will, of course, require getting lucky on poison procs during a KSp, and quickly unequipping your DP weapon once it ends, but I'd be interested to know the answer.

I suppose while we're at it it might be worth verifying the behavior of poison and Venomous wound procs during KSp; I recommend Wound Poison rather than IP for testing this to avoid the IP damage range problems. And I suppose if we're being picky I'm not sure we've ever verified that autoattack damage gets boosted during KSp - we've always sort of assumed it does, but it might be good to check.

#28 Xenden

Xenden

    Glass Joe

  • Members
  • 5 posts

Posted 01 October 2010 - 07:17 PM

18. Does HAT affect Spell Crit?

Hat does affect spell crit as well as melee crit.


No Hat.
Posted Image

With Hat.
Posted Image

I'm not quite sure why it says there was 32 sp, i had no buffs on me, so I assume it is just a bug.
And if anyone is interested this is the spec i used; shouldn't make a difference, but I'm including it for the sake of completeness.

#29 Lightshadow

Lightshadow

    Piston Honda

  • Members
  • 133 posts

Posted 01 October 2010 - 07:42 PM

I can verify that autoattack damage gets boosted during KsP. Totally naked except for a white 1h level 1 sword (with flowers in the OH to prevent attacks).

Posted Image

#30 ieatpaperbag

ieatpaperbag

    Piston Honda

  • Members
  • 230 posts

Posted 01 October 2010 - 09:03 PM

I think it would be good to narrow it down further; in particular, some testing with a fixed-damage attack (gouge for rogues - other classes have other options) would be very helpful towards pinpointing the actual value.


I did 100 gouges with 6595 AP (after vitality, I only had 3 points in precision, forgot about them)
15 of the normal hits were 1030, the rest were 1029 (average of 1029.15)

Posted Image

I did another 50 gouges with no talent points (5735 AP)
22 of the normal hits were 905, the rest were 906 (average of 905.56)

Posted Image

Initially I used (AP * 0.21) + 1 to predict my damage before armor and got two numbers that did not agree with each other or the initial guess from earlier tests. I looked at the cata tooltip in wowhead and guessed at (AP * 0.21) + 117. The first test estimates damage reduction from armor to be 31.4791% and the second suggests 31.4671%. They don't seem to agree to the degree I would have liked for something like armor. What do you suggest for a better number rogue-wise, a larger sample? what are some of the other classes' abilities that could provide better data?

#31 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13510 posts

Posted 01 October 2010 - 09:41 PM

Your data is internally consistent for the moment, so I wouldn't worry about it too much. As a note on further testing: doing repeated trials once you've identified both possible damage values isn't really that useful - there's just too much noise in the probability distributions for the average to say much. But lets look at we can figure out based purely on the fact that crits do 2058-2059 damage with 6595 AP and 1810 to 1811 damage with 5735 AP.

Wowhead asserts that the damage of gouge at level 85 is 117; this is consistent with your proposed formula of 117 + .21*AP, so lets assume for the moment that that is correct.

Per the combat ratings thread, the armor mitigation parameter, P, is 26070 at level 85. Hence, an attack with raw damage D and mitigated damage D' against a target with armor A satisfies D'/D = P/(A+P), or, rearranging, A = P*(D-D')/D'.

At 6595 AP, D = 3003.9 for a crit, so plugging in at our two endpoints for D, we find that 11963.8 <= A <= 11982.3.
At 5735 AP, D = 2642.7 for a crit, so plugging in at our two endpoints for D, we find that 11972.6 <= A <= 11993.6

Hence, combining these data ranges, we can assert with absolute certainty that, assuming our other formulas are correct, the target's armor is between 11972.7 and 11982.3. Eyeballing it based on the positions of the averages within each range, we can guess that the answer is probably "about" 11975, but with a few more endpoint tests with Gouge (or any other fixed-damage ability), it should be reasonably straightforward to narrow down the answer more precisely.

#32 ieatpaperbag

ieatpaperbag

    Piston Honda

  • Members
  • 230 posts

Posted 01 October 2010 - 10:45 PM

14. How does Coup de Grace stack with Revealing Strike? What about Executioner (the Subtlety Mastery)? What about Aggression?


I did all of the following with no gear and 2 white-quality weapons on a lvl 60 dummy that gave approximately 12.56% DR from Armor. I did not take the RvS glyph, Savage Combat or Bandit's Guile in any spec.

No evis talents, I did take Imp SS for convenience and so that I would have the same AP of 783 from Vitality for all except for the executioner test which had 785 AP via Sinister Calling.
Posted Image

Coup de Grace only with combat as the main tree for vitality.
Posted Image

Coup de Grace with Aggression, they appear to be additive
2978 * (1+ 0.20 + 0.15) = 4020.3 ~ 4005
Posted Image

Coup de Grace with Executioner again with 785 AP and base 20% mastery, these appear to be additive as well
2978 * (1+ 0.20 + 0.20) = 4169.2 ~ 4167
Posted Image

The rest involve RvS which does something to eviscerate that I either misunderstood somewhere, or I did something wrong, or has an effect I just plain can not figure out.

No evis talents, just RvS
Posted Image

Coup de Grace and RvS
3096 * (1+ 0.20) = 3715.2 ~ 3719
Posted Image

Coup de Grace with Aggression and RvS
3096 * (1+ 0.20 + 0.15) = 4179.6 ~ 4199 (getting far but this sample only has 25 eviscerates)
Posted Image


Interesting. Could you also confirm with Rupture, and perhaps try refreshing DP during KSp? That is, if you have a rolling DP stack and refresh it during KSp, does the following tick do 20% more damage? This will, of course, require getting lucky on poison procs during a KSp, and quickly unequipping your DP weapon once it ends, but I'd be interested to know the answer.


An old rupture also does not increase in tick damage during or after a KSp.
Posted Image

If DP is refreshed during KSp, it increases to 120% until refreshed again after.

In this one, I did not refresh DP after KSp.
Posted Image

In this one, DP was refreshed after one tick after KSp ended.
Posted Image


[*]Verify the IP/WP proc rates. I have no particular reason to believe it's changed, but it's worth verifying. Similarly, verify that the proc rate on ranged attacks (i.e. FoK) is what you'd expect.


Adding to the expected ranged proc rates in post #21, here are some tests with WP/IP on a single MH weapon, the proc rates appear to be the same as live.

IP on a 2.6 speed Mace
187 / 500 = 0.374 (expected 0.3714)
Posted Image

IP on a 1.3 speed Dagger
93 / 500 = 0.186 (expected 0.1857)
Posted Image

WP on a 2.6 speed Mace
460 / 501 = 0.918 (expected 0.9286)
Posted Image

WP on a 1.3 speed Dagger
230 / 500 = 0.460 (expected 0.4643)
Posted Image

#33 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 02 October 2010 - 03:27 AM

18. Does HAT affect Spell Crit?

Hat does affect spell crit as well as melee crit.

I'm not quite sure why it says there was 32 sp, i had no buffs on me, so I assume it is just a bug.
And if anyone is interested this is the spec i used; shouldn't make a difference, but I'm including it for the sake of completeness.

All but your first 10 intellect count towards SP/Mana pool.
All but your first 10 agilty count towards AP/crit.
All but your first 10 strength count towards AP/block (if str still increases blockvalue).
So it seems you should have exactly 42 intellect naked
Probably some very old design decision that still lurks around

Regarding HaT: There are no +5% spell and 5% melee crit buffs, every crit buff gives %5melee and spell crit, a little step further down the road so you don't only have 2-3 exact 10man raid setups which guarantee you all de/buffs
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#34 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13510 posts

Posted 02 October 2010 - 03:47 AM

Regarding HaT: There are no +5% spell and 5% melee crit buffs, every crit buff gives %5melee and spell crit, a little step further down the road so you don't only have 2-3 exact 10man raid setups which guarantee you all de/buffs


I'm completely aware of the design principal; on the other hand, consider the wording of the abilities:

Leader of the Pack: "increases critical strike chance..."
Elemental Oath: "...bonus to their critical strike chance."
HAT: "Increases the ranged and melee critical hit chance..."

There's a very distinct difference in wording there. And it's not like they haven't shown a willingness to give some classes inferior versions of buffs - for instance, how there are both +10% and +6% spell power buffs. So it might just be a typo. Or we might actually have a weak version of the ability. And rather than making assumptions, it seemed sensible to test it.

#35 Lightshadow

Lightshadow

    Piston Honda

  • Members
  • 133 posts

Posted 02 October 2010 - 05:25 AM

26. Does the AR haste stack as expected with Windfury/SnD/Haste Rating?


AR haste stacks multiplicatively with SnD, WF, and haste rating.

				Observed		Expected	
base speed			2.6	1.6		
6% haste talents		2.45	1.51	2.452830189	1.509433962
AR + talents			2.04	1.26	2.044025157	1.257861635
AR + talents +snd		1.46	0.9	1.460017969	0.898472597
AR + talents + WF		1.86	1.14	1.858204688	1.143510577
AR + talents + WF + snd		1.33	0.82	1.327289063	0.81679327

				2.6	1.5		
talents + gear			1.85	1.07	1.846315535	1.06518204
AR + talents + gear		1.54	0.89	1.538596279	0.8876517
AR + talents + gear + snd	1.1	0.63	1.098997342	0.634036928
AR + talents + gear + WF	1.4	0.81	1.39872389	0.806956091
AR + talents + gear + SnD +WF	1	0.58	0.999088493	0.576397208

By the by, can you say a little more about what you mean (or link to the original posts) about 16 and 24?

#36 shadowboy813

shadowboy813

    Von Kaiser

  • Members
  • 50 posts

Posted 02 October 2010 - 07:17 AM

FoK on the main target is buffed, but not on any of the secondary targets.

Edit: I can also say something more about energy regen. Without a doubt base regen is 10/second and 12.5 a second with vitality. The value of haste was up for debate in the 4.0.1 combat spreadsheet thread, so I did a somewhat longer test. I spammed fok for 120 seconds and recorded both the total casts and the energy remaining at the end of the test. With 32.57% haste I got 58 casts off with 58 energy remaining, for a net expenditure of 1988 energy or 16.566 energy/sec. If haste is multiplicative (e.g. regen = 12.5 * (1+haste%)) then I'd have expected 16.571 energy/sec, well within the margin for human error here. The only alternative hypothesis I've heard advanced was 12.5 + .11 * haste% which would predict an overall regen of 1930 energy, a reasonably large difference.

Having done that I went one step further and did the exact same test (with the same 32.57% haste) but this time also used an adrenaline rush during the test, making sure not to cap and without the glyph. Here I got 63 casts off with 78 energy remaining at the end, for a total regen of 2183. The total difference in energy regained was thus 195 (give or take a little bit given that I might have read the energy remaining bar slightly early/late). Without the glyph I would have expected 150 energy regained if there was no interaction with vitality, but this is significantly higher. It's a much closer match for it being 150*1.25 than it is for just the flat 150, but also pretty clearly not being influenced by haste.


I did a test of energy regen scaling with haste using a different method (prior to seeing this post).

My methodology (assumptions noted where they were made).
I dumped energy with fan of knives, standing outside of melee range with autoattack off. I then spammed fan of knives for 200 executes. I noted the activity on recount and divided this into the expected energy cost of 200 Fans of Knives. I did this at 0 haste rating, and 7 other incremental values of haste. I then plotted these on a spreadsheet and added a trendline using the built in least squares fit.

I then repeated the same experiment as subtlety (no benefit from vitality).

It appears as though vitality stacks multiplicatively with haste-based energy regen scaling.

Here is the chart:

Posted Image

(Note the outlier at around 20% haste on the subtlety line. Nature called about halfway through so that point only represents 100 FoK executes. EACH point represented 6-9 minutes of pressing the same button repeatedly. I was not about to start that data point all over).


I agree that this method has some error introduced that would increase the energy regeneration from the initial energy dump and if anyone is interested, I will refine my method to a more accurate one. By using time-stamps from a combat log, parsing them, and discarding initial executes to a point where the derivative of time-between-executes becomes tolerably close to zero. This should allow a more meaningful data set to be used.

#37 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13510 posts

Posted 02 October 2010 - 08:05 AM

By the by, can you say a little more about what you mean (or link to the original posts) about 16 and 24?


16: http://elitistjerks....17/#post1703736
24: Perhaps it's my imagination, but my impression on previous fights with snare-able targets (Freya comes to mind) was that the uptime on Blade Twisting isn't anywhere near the 90%+ you'd expect based on the tooltip proc rate. And since I don't recall anyone ever actually verifying what the proc rate is, it seems reasonable to at least to a sanity test to see if the proc rate is the 40% the tooltip suggests. Doesn't need to be big or fancy, just a couple hundred hits to verify it's in the right ballpark, confirm that it procs off SS, finishers, and KSp, etc.

#38 niemann

niemann

    Glass Joe

  • Members
  • 4 posts

Posted 02 October 2010 - 09:28 AM

On Venomous Wounds, Potent Poisons and Vile Poisons stacking.

Interesting - particularly since Venomous Wounds implies the opposite. I wonder if the damage range on IP is dirtying the results? Regardless, we'll need more testing - we should probably look at WP and DP in addition to Venomous Wounds and IP to try to get the full picture. As a note to people testing this: beware of Master Poisoner.


Venomous Wounds doesn't gain from Vile Poisons which was also never implied in the earlier test. In my case, Venomous Wounds did 1203-1204 damage with and without Vile Poisons.

Regarding stacking, I tested with one deadly poison stack.

With only 20% Potent Poisons: 265-266.
With 20% Vile Poisons and 20% Potent Poisons: 309-310.

Worst case: 310/265=1.169.

On a more curious note, Venomous Wounds only works when your target is poisoned.

#39 korner

korner

    Von Kaiser

  • Members
  • 59 posts

Posted 02 October 2010 - 09:53 AM

7 How do Opportunity and Improved Ambush stack?

Tested with no proc enchants/trinkets and no meta gem (so crit is exactly 2 * normal nit)

1. Clean Ambush, no Imp. Ambush, no Opportunity.
Posted Image

Min. damage: 1852
Max. damage: 1859

2. Improved Ambush only:
Posted Image

Min. damage: 2130 (minimal crit/2)
Max. damage: 2138 (max. crit/2)

Calculated values:
min: 1852*1.15 = 2129.8  ~ 2130
max: 1859*1.15 = 2137.85 ~ 2138

3. Improved Ambush + Opportunity:
Posted Image


Min. damage: 2686 (minimal crit/2)
Max. damage: 2696 (max. crit/2)


Calculated values:

1. Multiplicative:

min: 1852*1.15*1.30 = 2768.74
max: 1859*1.15*1.30 = 2779.21

2. Additive:

min: 1852*(1+0.15+0.30) = 2685.4 ~ 2686
max: 1859*(1+0.15+0.30) = 2695.6 ~ 2696

So, Opportunity and Imp. Ambush do stack additively.


PS: Regarding Find Weakness:
It looks like it's not implemented on PTR - none of Ambush, Garrote or Cheap Shot does not modify damage of yellow attacks or autoattacks, at least for me.

#40 Aldriana

Aldriana

    Mike Tyson

  • Moderators
  • 13510 posts

Posted 02 October 2010 - 06:17 PM

I can verify that autoattack damage gets boosted during KsP. Totally naked except for a white 1h level 1 sword (with flowers in the OH to prevent attacks).

Posted Image


For this test, can you confirm what weapon you were using? In particular, what speed it was?




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users