Retribution: Updated for 3.3 - Page 95

• Zurm · 12/21/09, 3:43 PM

Originally Posted by Angrygrimace

Is haste really so valuable to a Ret. that any piece with haste on it is automatically a "Ret. Piece" and should be passed and I should be forced to just blow 50 badges on the Badge cloak and delay getting my Tier bonuses?

It's more that ArP is bad for us and haste is...less bad. Crit is still better (significantly). That being said, with warriors approaching the ArP cap and DKs caring less about it than they used to, haste pieces are certainly not "ret" pieces.

That being said, the [Saronite Gargoyle Cloak] is probably the best he can get... is the same true for you?

• frmorrison · 12/21/09, 4:02 PM

Originally Posted by Angrygrimace

Is haste really so valuable to a Ret. that any piece with haste on it is automatically a "Ret. Piece" and should be passed and I should be forced to just blow 50 badges on the Badge cloak and delay getting my Tier bonuses?

Note the expertise on the Badge cloak may be useful to you, but in fact Haste is better for Unholy DK than Ret (largely due to the extra benefit you get to pets Ghoul, AotD, and Garg being affected by it). Warriors get the least benefit from haste from the three. To better compare the two, Ret's dps benefit is 1 Haste about equal to 1 Attack power (since other than getting SoV up, it only affects auto and the resulting Seal hits and makes Cons and Exo cooldown a little faster), while a Unholy gets 1 haste equal to 1.75 attack power.

A group's loot decisions are out of the scope of the EJ class mechanics forums but discussing the worthiness of a stat is fine.

Angrygrimace · 12/21/09, 4:03 PM

Originally Posted by frmorrison

Note the expertise on the Badge cloak may be useful to you, but in fact Haste is better for Unholy DK than Ret (largely due to the extra benefit you get to pets Ghoul, AotD, and Garg being affected by it). Warriors get the least benefit from haste from the three. To better compare the two, Ret's dps benefit is 1 Haste about equal to 1 Attack power (since other than getting SoV up, it only affects auto and the resulting Seal hits and makes Cons and Exo cooldown a little faster), while a Unholy gets 1 haste equal to 1.75 attack power.

A group's loot decisions are out of the scope of the EJ class mechanics forums but discussing the worthiness of a stat is fine.

Well, I'm Expertise capped as it is in my raiding set (he has less than 10 Expertise in his raiding set), so I believe that the Saronite Cloak is probably a better choice for me than the Ocean Serpent one.

My primary interest is primarily in the idea that he claimed I was not understanding the way Seal of Vengeance worked and that haste was the best stat for Ret. Paladins, which made little sense to me based on every bit of research I've done. My understanding is that I should be advising him to gear straight strength for everything, just like most other plate classes, because even if he stacked enough haste to gain more SoV procs, the actual proc would just be much lower since his attack power would be lower.

Glutton · 12/21/09, 4:35 PM

I don't think we've ever had a clear or satisfactory explanation of the discrepancy, but SimulationCraft values haste much more than Rawr. Depending on your location relative to haste maxima and minima, you may see it provide a higher DPS value than crit. Using my gear as an example, SimulationCraft graphs the two as having nearly equal value.

1 Haste = 1.81 AP
1 Crit = 1.74 AP

Kagtuuk · 12/21/09, 4:46 PM

Originally Posted by Angrygrimace

My main raiding character in our 10 man only guild is my Death Knight, specced Unholy. The Retribution Paladin in our guild (who recently switched from being PvP only) essentially declared that [Saronite Gargoyle Cloak] is "his," and that I would should be forced to either pass on the upgrade or go buy [Might of the Ocean Serpent] with 50 Frost badges instead of going for the 2 set T10 to get the set bonus and that I'm "screwing" him by not just giving him that cloak.

For what it's worth -- I don't know about your guild mate, but my Ret spec gear is roughly equivalent to yours (item-level wise). According to RAWR, the best possible non-25 man cloak for me is [Recovered Scarlet Onslaught Cape] worth 357.76 DPS, as opposed to the Saronite Gargoyle at 327.7 or the Ocean Serpent at 287.66. So from a badge loot perspective he should probably be considering the Recovered Scarlet as opposed to the Ocean Serpent.

Update: You linked the non-heroic cloak, but in case [Saronite Gargoyle Cloak] is what you meant to put, it's worth 358.24 DPS (without a socket bonus...), or ~.5DPS more than [Recovered Scarlet Onslaught Cape].

• frmorrison · 12/21/09, 4:52 PM

Originally Posted by Glutton

I don't think we've ever had a clear or satisfactory explanation of the discrepancy, but SimulationCraft values haste much more than Rawr. Depending on your location relative to haste maxima and minima, you may see it provide a higher DPS value than crit. Using my gear as an example, SimulationCraft graphs the two as having nearly equal value.

1 Haste = 1.81 AP
1 Crit = 1.74 AP

I believe Rawr used Redcape's spreadsheet as its basis, which is where I get my stat weights from that I listed in the post: 1 Haste = 1.05 AP
1 Crit = 1.42
Those are big differences, maybe caused by a different set of gear?

Exemplar · 12/21/09, 5:05 PM

Haste is a funny stat. For the most part it provides 0 DPS increase except when it hits plateaus.

For casters hitting enough haste for 1 more spell is more DPS. Until they can cast that extra spell - no boost. Same goes for us and autoattack (or possibly a special attack from Exo/Cons GCD reduction).

So not only does gear impact its efficacy, but modeled length of fight.

Redcape and I tend to use 5 minutes as a standard model. In such a relatively short timespan the difference between current and one extra auto+seal is a large jump. So each point of Haste is 0 DPS until you reach a threshhold. I believe we both average out haste to come to a reasonable weighting. Longer models (Rawr and Simcraft?) could come up with smoother calculations for a single point of haste.

Glutton · 12/21/09, 5:20 PM

The numbers I linked were from 10,000 iteration run of a 5 minute Patchwerk encounter.

I could never prove it, but my theory was that over a large time scale the spell GCD reductions on Exo, Conc, and Divine Plea can allow for greater usage of Crusader Strike. I once did multiple tests of SimulationCraft by adding steps of +/- 10 haste, and there was a pronounced jump in CS usage at a certain level of haste.

Exemplar · 12/21/09, 5:43 PM

Updated version of Bellator spreadsheet available with all currently known loot (and socket bonuses I could track down).
Bryntroll proc modeled as accurately as possible.
Shadowmourne proc not modeled, however an average 200 Str bonus has been factored in.

Links are, as always, in the original post.

Redcape · 12/21/09, 7:08 PM

Re: value of haste

In my spreadsheet I do not model individual autoattacks at all. Thus the value of haste on my sheet is never going to jump due to getting one more autoattack in, which is exactly the way I want it. Rebalancing your stats around those haste breakpoints you see in some simulators is a terrible mistake because those breakpoints are entirely dependent on the exact fight length chosen to model. Unless you are able to predict that a fight will last precisely 300 seconds (or whatever length your model happens to be) then those breakpoints are important to ignore when gearing and to avoid when theorycrafting.

The second source of breakpoints is when the GCD reduction from haste happens to dramatically improve your ability usage due to changing the spell GCD. This is something my sheet models based on your current haste level but the returns you see modeled do not include the average benefit of GCD reduction because it is so spiky. I tested the effects of adding huge amounts of haste to find the effects of GCD reduction on overall dps and it was absolutely miniscule, certainly small enough that natural variation and human frailty would overshadow it dramatically. Generally the effects of 2000 haste (ignoring the DPS benefit from more autoattacks) was in the neighborhood of 2.

Rawr, my spreadsheet and Bellator's all seem to agree: Str/Hit are king, Expertise is after those and then Crit/Agility. Haste is significantly lower than those others and ArPen is even worse. If another simulator is dramatically disagreeing on the benefits of haste then I would suggest that it is most likely due to some kind of bizarre spike valuation based on spell GCD that does not accurately measure up to reality or that the simulator is mistaken.

The best test would be to take the autoattack and seal damage that Simulationcraft suggests, manually figure out the benefit of haste and figure out if the numbers work out.

Edit: I sat down and worked out the above calculation for my sheet and the numbers were precisely on: The sheet lists 94 as my rating for haste (94 dps per 100 haste) and when I pull out the autoattack + AA-Seal damage and manually figure out the value of adding 1000 haste I get a dps increase of 939.73. Bang on. Try the same with Simulationcraft and see if it works out.

• Zurm · 12/22/09, 8:27 AM

Originally Posted by Exemplar

Shadowmourne proc not modeled, however an average 200 Str bonus has been factored in.

Minor nitpick, this should be 180 STR not 200. The split second you hit 400, the bolt procs and you lose the stack, so typically you will be between 0-360 STR, with an average of 180.

Also, Rawr is NOT based off any other program. Back when I worked on it in the middle of BC, I re-wrote it entirely from scratch. Then for 3.0, Ermad/Endoscient re-wrote it in its entirety once again. We did this specifically to try and avoid continuing bugs that may exist in other modeling tools.

• frmorrison · 12/22/09, 8:36 AM

Originally Posted by Zurm

Minor nitpick, this should be 180 STR not 200.

Indeed the proc should average to 180. I put in an extra 20 (on my Rawr file) to attempt to account for the shadow damage once you hit the full stacking.

Exemplar · 12/22/09, 9:15 AM

The non-proc model of Shadowmourne will be updated to 180 in my next release. Thanks for the info - I had mistakenly figured it stacked to 10, then 11th proc was the direct damage and loss of stack. You know... what would make sense. Silly me, didn't read the tooltip close enough.

Proc modeling will naturally wait until someone has one and we can determine how fast it stacks and procs. It's vaguely moot, as even without the proc (or 180 Str average) it's clear and away top DPS of known weapons. Modeling the proc will just let us know (within normal modeling margins of error) the exact DPS of the weapon.

dedmonwakeen · 12/22/09, 11:57 AM

Dimples in the SimC haste curves are not really from "fitting in" one more ability. There are a couple common misconceptions about the standard 5 minute sim length:

(1) Each iteration is limited by boss health, not actual sim time. Halfway through the first iteration, the boss health is inferred. At the end of every iteration, the boss health is adjusted (with a dampening factor) based upon whether the fight ended early or late.

(2) DPS is calculated by damage done divided by time spent DPS-ing. In multi-actor sims, if the boss dies while a player is in mid-cast, then the divisor used to calculate DPS is the time of the player's last action (or DoT tick).

There was a recent bug-fix that restored the RNG component of lag calculation in SimC. This was playing havoc with the haste calculation for spell-casters. Fixed lag values result in deterministic clashes for DoT-refresh and Cooldowns. These deterministic clashes in turn result in discontinuities in the DPS-Haste curves.

However, most Paladin offensive abilities are "attacks" rather than "spells". GCDs triggered from attacks do not benefit from haste reduction which is why one should see the same number of occurrences for CS/DS/Ex/HoW/etc regardless of the amount of Haste equipped. Indeed this is what SimC shows for me when I compare a T9 baseline against +100 Haste and +1000 Haste. The only thing changing is the number of melee attacks and SoV procs.

Originally Posted by Redcape

I sat down and worked out the above calculation for my sheet and the numbers were precisely on: The sheet lists 94 as my rating for haste (94 dps per 100 haste) and when I pull out the autoattack + AA-Seal damage and manually figure out the value of adding 1000 haste I get a dps increase of 939.73. Bang on. Try the same with Simulationcraft and see if it works out.

Redcape, given the nature of the 1/(1+x) haste formula, would you have expected that the DPS contribution of Haste to be so linear between +100 and +1000?

Grigorim · 12/22/09, 12:06 PM

one should see the same number of occurrences for CS/DS/Ex/HoW/etc regardless of the amount of Haste equipped.

Not entirely true. As Bluedeep noted, it would be possible for hasted Exo/Cons GCDs to allow for a larger number of Crusader Strikes. Could this be at all related to the bug-fix you mentioned?