Sorry for the spam, but like Avitas, I'm posting as I continue to find new information.

I'm going to look at this from a different angle and figure out what should be expected if my theory is correct.

Given 1000 swing cycles using a large club how many times are Windfury and Seal of Command expected to proc?
20% windfury = 200
37.3% Seal of Command = 373
Windfury procs SoC = 200(.627)(.373) = 46.8 - 62.7% of attacks did not proc SoC on auto, the remaining have a 37.3% chance to proc.
SoC procs Windfury = 373(.8)(.2) = 59.7 - 80% of attacks did not proc Windfury on auto, the remaining have a 20% chance to proc.

Total
259.7 Windfury attacks or 26.0% of normal attacks.
419.8 Seal of Command procs or 42.0% of normal attacks.

My results from 20 test in Blasted Lands (Note: Removed double Windfury proc from results)
489 Swings (386 excluding windfury)
103 Windfury
164 Seal of Command

Windfury
Expected: 26.0%
Actual: 26.7%

Seal of Command
Expected: 42.0%
Actual: 42.5%


These numbers are very close to expected. I would really like another data set from someone in blasted lands to compare to.

Just a heads up, my window for getting the SoC model changed for the next Rawr release is quickly closing... make a decision peoples!

Are you sure we don't? Remember, offhand attacks for DW classes show up as 'swing' too, so your numbers are flattened. And MS warriors have Sword Specialization, which muddies the waters slightly there.

I would vote to just leave it as is until we have more testing done.

Remember, if SoC is proccing Windfury (READ: that means your swing does not proc windfury but the seal proc does proc Windfury, not that SoC is proccing off a Windfury attack) it is a bug. Windfury has been set to not proc on Yellow attacks, and SoC is a yellow attack. So yes, it may be happening in game, but in all fairness it shouldn't be.

On the topic of bugs, I ran the numbers for the "new" JoW with bugs for us and will post a new version of the list tomorrow (have to fix a few things with shamans again first).

This looks promising. I'm going to run a 20ish min test myself when I get home, probably, if it correlates, it would further remove some doubt.



I'd say go with this for the release:

-SoC can proc off WF (nothing changed here)

-WF can proc off SoC (this is largely based on Paragos' results, while not conclusive yet, it looks more likely than not, so we should include it)

-No chain proccing (again based on Paragos' results)


I'm hoping Fiola can comment on this too (mainly Paragos' latest numbers), since it seems there's some disagreement on the first point, however this has to go down tonight folks. I'll run my own tests to weigh into either conclusion.

Toaster, while I agree with you on WF proccing off SoC probably being bug, however what we're trying to ascertain here is "what is" rather than "what should be". It's by far not the first time we meet wonky mechanics that just shouldn't be like that in theory (according to the rules blizzard vaguely establishes on their tooltips), but are later proven to be correct in practice and we adjust our theorycraft accordingly.

If it turns out this is not correct, possibly release a minor revision for Rawr afterwards (Astrylian?)

Sence alot palas have got into this new seal Twisting. Lot of have tested twisting SOC and SOB. They have got some really interesting resualts.

WoW Forums -> Seal twisting is godlike with SOB

SoC procs an extra Seal of Blood effect. Could this mean SOC is rather considered a so called normal auto swing/on the next attack skill instead of special attack?

That could explain as well why WF would proc on SOC procs.

Alright, I ran a 30 min test on a Blasted Land mob.

WWS Log: Avitus - WWS

(Don't ask me why it lists me as a mob)



(One minor correction compared to the WWS: It was only 590 swings, WWS is correct, those 2 extra accidental swings where after the test period had run out and I didn't have SoC or WF since were were trying to de-aggro the blasted lands mob)

Setup:

A shammy dropping WF (and healing stream) and making sure it never fades.

No Judgement applied to avoid any refresh mechanic oddities.

No JoC, CS or any other abilities used. Just Auto-attack and refreshing SoC before it fades.

I was using my regular gear which is hit capped and expertise capped, so no misses, dodges or parries (since it's a low level mob it was enough).

I was using no equipment that has haste procs: No DST (used regular AP trinket instead), no weapon with Mongoose (used [Large Club]) instead.

Time: 1800 seconds (30 mins)

Assumptions and base calculations:

Due to my 74 passive haste, the 3.2 speed weapon was hasted down to 3.06 speed: 3.2/(1+((74/15.76)/100)) = 3.05648 attack speed

At 7ppm, 3.05648 attack speed should give a SoC proc chance of 35.659%


Results compared to "classic theory" (no WF off SoC, no SoC off WF):

White Swings:

Theory: 1800/3.05648 = 589 swings.
Practice: 590 swings (771 swings - 181 WF procs).

SoC procs:

Theory: 7*1800/60 = 210 SoC procs
Theory again: 590 swings * 35.659% = 210.3881 SoC procs
Practice: 265 SoC procs (44.9% proc chance)

Theory is not matching practice.

WF procs:

Theory: 590 swings * 20% = 118 WF procs
Practice: 181 WF procs (30.678% proc chance)

Theory is not matching practice.


Calculation 1 (same as in Paragos' example):

Note: Paragos assumed SoC and WF cannot proc more than once per cycle.

Assuming WF can proc SoC:

20% * 0.64341 * 0.35659 = +4.58867% to proc SoC
New chance to proc SoC: 35.659% + 4.58867% = 40.24767% to proc SoC

Assuming SoC can proc WF:

35.659% * 0.8 * 0.2 = +5.70544% to proc WF
New chance to proc WF: 20% + 5.70544% = 25.70544% to proc WF


Results compared to assumptions from new theory (calculation 1):

SoC procs:

Theory: 40.24767% = 237.461 SoC procs (@ 590 swings)
Practice: 44.9% = 265 SoC procs

Closer, but still not matching practice (slightly under).

WF procs:

Theory: 25.70544% = 151.662 WF procs (@ 590 swings)
Practice: 30.678% = 181 WF procs

Closer, but still not matching practice (slightly under).

I believe Paragos initial assumption of "no chain procs" is probably false, hence the lower numbers. There was also a lot of number rounding being done in his example so it seemed to fit, however I explicitly did not round in my calculations and it seemed to not fit my data.


Calculation 2:

Note: I'll assume SoC and WF can chain proc.

Assuming WF can proc SoC:

Proc chance will remain at base (35.659%) but White + WF will be considered.

(590 white + 181 WF procs) * 35.659% = 274.93089 SoC procs
-> 46.598% to proc SoC off auto-attack

Assuming SoC can proc WF:

Proc chance will remain at 20% but White + SoC will be considered.

(590 white + 265 SoC procs) * 20% = 171 WF procs
-> 28.983% to proc WF off auto-attack

Results compared to assumptions from new theory (calculation 2):

SoC procs:

Theory: 46.598% = 274.93089 procs
Practice: 44.9% = 265 SoC procs

Even closer, though not a perfect match (slightly above).

WF procs:

Theory: 28.983% = 171 WF procs
Practice: 30.678% = 181 WF procs

Even closer, though not a perfect match (slightly under).


Conclusion:

It seems pretty obvious that "classic theory" where nothing procs off each other is not correct.
There is still a slight discrepancy (96.4% match to actual SoC, 94.5% match to actual WF) between this "new theory" and actual results, however this can be attributed to only having run a 30 min test (600 swings) rather than a more conclusive 10000 swing test.

I can also pretty much conclude that yes, SoC and WF have to have some sort of relationship allowing them to proc off each other. Since I avoided all other outside influences/variables in my test, this is the only remaining explanation. I consider this a conclusive result (mutual proccing) and a done deal.

Furthermore, there's a high indication, though not conclusive, that they can chain proc as illustrated above. Why we don't notice it in game could be attributed to the spammy nature of the combat log and not being able to tell the "proc source" of procs. There also might be a conclusion inbetween 1 proc each per cycle and being able to proc indefinitely, but I can't figure it out atm, there's definitely something however.

I remember in the old days of Reckoning bombs, if you used SoC, in the 5 swings you unleashed there was ALWAYS a SoC proc (I never had one without). I also remember quiet distinctly freak occurrences of 2 SoC procs from a 5 swing reck bomb, though they were very rare. There might be something regarding allowing chain procs, but really limiting their chance of occurring.


Anyway, I'm "mathed out" for now, this is my bid on this whole issue, I'd love to read some input or constructive criticism. Hell, anything to solidify this or even throw it overboard and find an even better theory, as long as we find a conclusive answer.

Thanks for the test, Avitus.

I went through every line of your combat log. I saw two instances of double windfury procs, once at
19:24'20.135 and once at 19:35'26.243. It is important to note that the second occurance of the double windfury proc had NO SoC proc involved, so it was definitely not SoC which caused the double proc. I'd like to put forth that the windfury proccing itself bug does not necessarily mean windfury procs windfury 20% of the time. It could be due to timing or some switch not flipping to let the occasion double proc slip through.

As for your larger numbers, I agree that my current calculation of predicting expected results is not exactly right, and here is why.

Suppose RNG is being kind and you get more than average windfury procs. That is going to increase your overal SoC proc rate as well. The opposite is true for a lucky SoC rate increasing your windfury procs. Suppose you get lucky on both, which increases your rates for both and you get really high numbers for both. This would have a self-limiting factor if they couldn't chain proc. (I noticed almost invariably in your logs that if you got one, you got the other) So, one going well helps the other get closer to the expected average that my caluclations provide. The only time things go badly is if you get really unlucky on autoattack procs for both windfury and SoC. Neither one would bring the rate of the other closer to my calculations.

The possibilities for WF and SoC rates in comparison to mean are above/above, above/below, below/above, below/below. Three of the four possibilities are going to help increase the other, so you're more likely to get above my calculations than below.

I am sure there is some fancy statistical math to compensate for such things, but I have to be honest, that is way beyond my knowledge.

I still believe in the current hypothesis, but feel the calculations for average could use some math experts. However, the current model is not accurate and this one is a lot closer, so it wouldn't hurt to put it into rawr in the next release.

Hi all,

Very new Ret pally here, still only Level 65, but going hard!

I'm generally not a big fan of macros, but a Ret pally has so few abilities that are used, I thought it might be easy to whip one up for farming purposes.

Looking for feedback:

I think it is fairly explanatory, its intended to be a "spam" macro.

Hold ALT or CTRL whilst in combat to recast SotCru if you switch targets.
If you switch targets, the castsequence will reset.

My only concern is the macro CD is 8 seconds whilst Crusader Strike is 6.

Looking for testing and opinions, as I said I'm very new so feedback is welcome. I doubt this macro is raid DPS viable, but I'd love to hear your experiences!

Cheers.

That's a nice write-up there, Avitus. I don't have the time/will to to do a math post as well organized or thorough as that, but here's my 2 cents on the data set you collected.


Data, using 72 haste rating (4.7% haste) w/ 3.2 AS weapon
Normal swings: 590
SoC procs: 265
WF procs: 181


Base expected proc chance (model where WF and SoC don't affect each other)
SoC: 35%/37% (haste/no haste)
WF: 20%

Apparent proc chance
SoC: 265/590 = 45%
WF: 181/590 = 30.6%


I. Apparent proc chance if WF procs SoC, SoC procs WF
SoC: 265 / (590 + 181) = 34.4%
WF: 181 / (590 + 265) = 21.1%

II. Apparent proc chance if - WF proc'd from SoC can't proc SoC, (using non-hasted proc chance)
SoC: 265 / ( 590 + 181 * .63) = 265 / ( 590 + 114) = 37.6% (note: We'd be "aiming for" 35% if we expect the hasted proc chance)
WF: 21.1% (same as I; applying chain proc rule to WF->SoC->WF would increase the apparent proc chance)


At a glance, both models I and II are likely from this set of data. Let's examine the likelihood of either using the Std.Dev rules from before.

I. SoC swings - 771; WF swings - 855
Expected SoC (37%, non-hasted): 285
Std. Dev for SoC (non-hasted): 13.4
Expected SoC (35% hasted, for completeness): 270
Std. Dev for SoC (hasted): 13.2

Actual proc # is 1.5 SDs from expected for nonhasted SoC and around .4 SD from expected for hasted SoC.


Expected WF (20%): 171
Std. Dev: 11.7

Actual proc # is around 1 SD from expected WF proc #.


II. SoC swings - 704
Expected SoC (37%, non-hasted): 164
Std. Dev for SoC (non-hasted): 12.8
Expected SoC (35% hasted, for completeness): 246
Std. Dev for SoC (hasted): 12.7

Actual proc # is 0.1 SD from expected non-hasted SoC and 1.5 from expected for hasted SoC.


Summary:
Based on the multiple sets of WWS data we've collected so far, WF procs off ALL SoC attacks.

Based on all the data we've collected, SoC proc chance seems to be affected by WF. Looking only at this one data set. SoC either uses hasted proc chance and can chain proc (SoC->WF->SoC), OR it uses non-hasted proc chance and cannot chain proc.


I'm very sure there are better mathematical tools (student t test, some other funky stat stuff) for analyzing all of our data, but I'm not a math/statistics major and am busy with my other classes. = P

Beta 14 posted!

DOWNLOAD: Rawr - Release: Rawr b14


Beta 14:
- Added the ProtWarr and Healadin models! These are our first versions of these models, so please report any bugs you find with them!
- Mac support! See the readme for details.
- Major improvements to the intelligence of the Optimizer
- Added a new feature in the Optimizer: Build Upgrade List. This feature will take longer than the normal Optimizer (so you'll want to run it at a lower thoroughness, most likely), but will produce a chart of how much value you could gain if you had each item that you don't already have available.
- Added a Direct Upgrades chart, which shows what the biggest upgrades for you would be, considering just direct upgrades, no other gear swaps. For a comprehensive upgrades evaluation, use the new Build Upgrade List feature of the optimizer.
- Significantly improved performance all across Rawr, especially in the Optimizer
- Added a Delete Duplicates function. By right clicking on an item or using the Item Editor, you can Delete Duplicates for an item, which will delete all gemmings of the item except for the one you selected, and any that are equipped.
- Reworked how Rawr stores calculation options. This means that when you load your characters created with previous versions of Rawr, the Options tab will be reset to default, please be sure to fill that out again.
- Rawr now correctly handles mainhand and offhand enchants, as appropriate by equipped items.
- Improved the loading performance of the Item Editor on successive loads. The first time you open it, it should be faster than before, and the second+ time should be nearly instant.
- Fixed a bug that made the Fill Sockets functionality of the Item Editor not work.
- Fixed a bug that made the chart render as a big red X occasionally.
- Made some changes that may help with the UI for users of high-DPI resolutions. High DPI is still not supported (as Windows' support for it is retard and broken), but this should help somewhat.
- Fixed a leak of control handles in the item selector, should improve performance and prevent crashes related to this.
- Fixed the Armor Penetration values of Imp/Expose Armor.
- Reloading the current character from the armory will now load enchants and talents, in addition to items.
- The Item Selector should no longer extend off the screen when there's room available, on very low resolution displays.
- Added support for loading characters from KR and TW realms.
- Load Upgrades from Armory will no longer create duplicates.
- Added some warnings to the Optimizer to help people use it properly (such as pointing out when you've forgotten to mark any choices in a slot as available). These can be disabled in the Options.

Rawr.Retribution Changes:

-Added a slider to choose bloodlust uptime %
-Added a slider to choose Drums of Battle/War uptime%
-Added a slider for target armor, lists bosses with that armor value
-Added a slider to choose the number of Ferocious Inspirations
-Added a slider to choose Expose Weakness AP value

Note: No change was made to SoC and WF mechanics at this time. Upon further review, a change will be made for B14.1

Would you human pallies concede that Fiola's summary a few posts up is the final decision on the matter? If so, I will code it tonight for release in Rawr B14.1.

if you have three mobs, and you JOW each mob by a different pally, and you AOE said mobs, will the AOE'ers get JOWx3 (Since it is 3 seperate mobs)

Just wondering.

BW

You can safely add WF procs off ALL SoC attacks to Rawr.


Assuming JoW was on 3 AoEed mobs, you could get 3 procs of JoW.

There's no concession, Fiola is agreeing

And I can pretty much throw out the alternative theory (that SoC goes off base speed), since it's been proven multiple times that it doesn't if you run pure SoC tests (without WF) with haste gear, it remains at 7ppm.

What to code:

I think we're now all in agreement that both SoC and WF proc off eachother (though further results would solidify this more, so anyone reading, feel free to run a few tests and supply more data).

It also seems a high indication of chain procs.

What this means for coding: Just take the "calculation 2" from the mammoth post I made (with the test results) a few posts up and use that


Regarding the new Rawr: Very nice work

There's some things however that still seem to be missing off the lists (major cleanup of buffs tab, fixing all buffs that give spellhit/spellcrit, SoC libram working for SoB, Solarian Sapphire Battleshout, Trueshot aura).

Also it seems that Ferocious Inspiration got implemented wrong, it's 3% to "all" damage, not just physical attacks.

Anyway, I'll help with the coding as soon as I have more time (long vacation coming up in 2 weeks), also taking further specifics to PM/IM.

Cheers

The buffs tab is Rawr-wide, not specific to retribution. Hence, it will include all buffs that any rawr module needs. The solarian sapphire is something that is still being debated among the developers as to how to implement. I thought I fixed the libram thing, but I guess I just forgot. Trueshot aura I totally forgot about, let me get back to you to make sure its something that just I have to change, or is a program-wide effort.

As far as FI implementation being wrong, if its wrong now it was wrong before, all I did was replace a static number with a variable. I'll look into it, and we'll definitely talk regarding SoC/WF changes.

No, it's filtered to just buffs relevant to Ret... I just checked, and all the code in CalculationsRetribution.HasReleventStats(stats) is commented out, and just 'return true' in its place, aka saying all buffs are relevant. Just need to switch that back to checking whether there are relevant stats for each buff.

Regarding FI... It's implemented correctly. FI really is 3% to *all* damage, not just physical. Works fine for hunters, warriors, mages, spriests, everyone.

Ferocious Inspiration <-- "all damage increased by 3%".

Yea I'm retarded. I'll have that fixed too.

Sorry I'm still learning the code :p

Yeah, I was agreeing. Not asking for any concessions. X D


Final thoughts on haste/nonhaste:

Going to the other data set (no haste, 3.2 AS)
490 hits (Total; 386 normal swings)
104 windfury
164 SoC

Chain proc model (WF off all SoC, SoC off all WF, 37% SoC proc chance)
expected SoC = 490 * .37 = 181
SD = 10.7; actual is 1.6 SD away from expected
expected WF = (386 + 164) * .2 = 550 * .2 = 110
SD = 9.4; actual is 0.64 SD from expected

Non-haste model (WF off all SoC, SoC off of normal->WF, 37% SoC proc chance)
expected SoC = (386 + 104 * .63) * .37 = 167
SD = 10.3; actual is .3 SD from expected



Now for some other random thoughts.
If chain proc model is true, we expect 8.4 PPM of SoC + WF regardless of gear level and amount of haste. (we assume SoC does not scale with haste, so PPM stays constant, and WF's boost to PPM is static)

From inspection, we have the
[bonus procs from WF] + [bonus procs from haste & WF] - [the # of SoC->WF->SoC procs that were included in the previous terms, but which we do not expect in this model]


damage difference between the 2 models
7 PPM *1.2 * (haste%) - 7PPM * (1+haste%) * (0.2 * PC)
[bonus procs from haste & WF] - [the # of SoC->WF->SoC procs]

8.4 PPM (7 PPM * 1.2) is in both equations and cancel out, leaving us with the two terms above.



Say we have a Cat's Edge (3.5 AS, 40.8%). Then the equation simplifies to
8.4 PPM * (haste%) - 0.57 PPM * (1 + haste%)
= 7.83 PPM * haste% - 0.57 PPM

At 0% haste, damage difference is 0.57 PPM - Model 1 says we can get WF + SoC + WF procs (chain proc), whereas Model 2 says we don't.

At 7% haste, the damage of both models is equal.

At higher haste levels, the damage of the 2nd model is higher. (Because 1st model says SoC is static, 2nd model says SoC scales with haste)



SUMMARY:
At <7% haste, the "SoC scales with haste, cannot proc twice" model expects less damage. At 7+% haste, the "SoC does not scale with haste, but can double proc" model expects less damage. Given ambiguity between which of the 2 models is correct, you won't go wrong by conservatively under-estimating your damage.


And thanks, now I'm late for class. X D


Edit: polished conclusion. 7% haste is around 105 haste rating, correct?

Blessings:

I find that I personally need wisdom or I'll run into mana issues, despite poting. I need salv or I'll risk pulling aggro or having to stop dpsing in order to not pull aggro. Having mana issues or being threat caped, reduces my dmg. So I usually have these 2 up. But then I miss out on Might and/or Kings, which also reduces my dmg.

If your raid have a max of 3 pallies, which blessings do you get? What if only 2 pallies are around? All physical dpsers get might and kings and I somehow, find myself not having them most of the time.



This btw, brings me to another thing I'd like to talk about: Inscriptions. It seems each toon will have 4 inscriptions available to change his spells and abilities. Blizz said they want to help ret pallies with our mana issues. My guess is that they would affect mana regen/consumption (something like what shammies have), soc (hopefully back to 100% dmg), cs (reduced cd) and possible something related to consecration or sotc/jotc. What do you guys think?

on a boss you shouldn't really have mana issues considering JoW will always be up (kings or might, to each their own). on trash i'll give myself BoW because most mobs aren't getting a judgement and i'll end most pulls autoattacking. kings is around 150ap and 1% crit. hard to turn that down for a boss.

If you're raiding as 3 paladins, do this:

Pally #1 throws BoKings on everyone
Pally #2 throws BoSalv on everyone (along with BoWis on Hunters and BoL on tanks)
Pally #3 (you) throws BoWis/Might on everyone and deal with 10 minute buffs accordingly (ie enh shaman[s], feral druid[s], and yourself)

Keeping the buffs perfect really doesn't matter in your case until bosses, otherwise I wouldn't bother with the 10 min buffs unless you're able to keep track of them.

Also what Gevlin said. Have another paladin judge Wisdom at the start of the fight.