With Xav's WWS log I could spot an error with Thunder Clap. The Improved Thunder Clap talent also increases the coefficient by 130%; that means it goes up from 12 % to 15,6 %.
Here is the new version along with new values. I'll delete all previous data to not confuse anyone.
With Xav's WWS log I could spot an error with Thunder Clap. The Improved Thunder Clap talent also increases the coefficient by 130%; that means it goes up from 12 % to 15,6 %.
Here is the new version along with new values. I'll delete all previous data to not confuse anyone.
Yep, I knew something was wrong with the TC numbers. I couldn't pinpoint it like I said, but it just wasn't really adding up with what you'd expect based on the threat modifiers, etc.
While 5200 AP in your current example is high (unless procs, etc are in effect), I put in 4500 AP and got the following, which may be more in line. I don't know my exact average AP in a boss fight, but I think it's right around 4500, maybe 46-47 fully buffed, I'll pay closer attention.
I assume the conc blow/shockwave numbers are bugged and I just chose the higher number, they were only off by 100. Anyway, I assume this (properly?) has Thunder Clap ahead of Devastate, which is what I was nearly certain of before. I was going crazy trying to figure out how it could possibly be less, so if there's other stuff we're missing, or others have insight/clarifications, go for it.
Edit: Good call on the shockwave/conc blow reasoning, parry/dodge. fixed ranking (4500 ap, again)
I didn't download the Yomon's sim but in the case we took into account that TC/Shockwave can't be parried/dodged wouldn't we also need to account for the lower GCD on concussion blow? (which I thought would have always put it ahead of shockwave)
Just checking, but did you factor in that Rend ignores armor?
Also, given the boost to rend with slow weapons (and assuming deep wounds), are we to conclude that slow weapon speeds are better than fast weapon speeds for tanking now? In other words, has anyone done a calculation on the trade-off between fewer heroic strikes vs. stronger rends and deep wounds?
Also, given the boost to rend with slow weapons (and assuming deep wounds), are we to conclude that slow weapon speeds are better than fast weapon speeds for tanking now? In other words, has anyone done a calculation on the trade-off between fewer heroic strikes vs. stronger rends and deep wounds?
This was addressed on the first page of this thread. Whenever you're ever capping rage out (rage overflow due to damage intake), a fast weapon will be better.
When you use a slow weapon it's even harder to dump rage, so the point at which you cap your rage comes sooner. Having a variety of tanking weapon speeds would end up being the best, such as 1.5, 2.0, 2.5, or something.
This was addressed on the first page of this thread. Whenever you're ever capping rage out (rage overflow due to damage intake), a fast weapon will be better.
When you use a slow weapon it's even harder to dump rage, so the point at which you cap your rage comes sooner. Having a variety of tanking weapon speeds would end up being the best, such as 1.5, 2.0, 2.5, or something.
I'm well aware of the tankspot post on slow vs. fast weapons, showing that faster weapons generate more TPS.
However, at the risk of receiving another posting infraction, I would like to point out that that post was made before people seriously considered putting rend into their rotation. In the example above, a slow weapon would provide (3372-2488)/12 = 74 TPS more per rend than a fast weapon. This may not be enough to justify a "slow weapon with rend spec" rule, but I think it is worthy of a calculation...
I've gotten the impression from the discussion here that devastate isn't an every GCD ability any more. The concern I've seen brought up is the S&B procs that come from it. Maybe I missed more clarification in a later post, but Revenge also procs S&B, so devastate is more of a 'get the sunders up fast with [Glyph of Devastate]' and then the [Glyph of Revenge] combined with a fast weapon for fast high dmg HSs and revenge based S&B procs is a priority. Devastate and TC would then be when refresh needed abilities, and HT, Conc, and Shockwave being thrown in when their CD is up.
Have I understood correctly where the discussion is leading?
I'm well aware of the tankspot post on slow vs. fast weapons, showing that faster weapons generate more TPS.
However, at the risk of receiving another posting infraction, I would like to point out that that post was made before people seriously considered putting rend into their rotation. In the example above, a slow weapon would provide (3372-2488)/12 = 74 TPS more per rend than a fast weapon. This may not be enough to justify a "slow weapon with rend spec" rule, but I think it is worthy of a calculation...
If the rough numbers in the tankspot post are to be trusted, he (shiz98) has a fast weapon as generating about 120 more TPS than a slow one (1.6 vs 2.6), in an infinite rage scenario. This is without counting the difference in rend. If a slow weapon's rend is a 70-80 TPS boost, it's still less than the gains you'd get from heroic striking more with a fast weapon.
Edit: Chlorin/zapps has some better illustrated numbers, which again show the fast weapon being superior, and again more of a TPS gain than rend. Actually, the crossover point may have changed, cause their numbers are a bit out of date, too.
Now, someone much better at math than me, and more thorough is certainly welcomed (and encouraged) to run the numbers so we have more concrete things to work with.
And I really think I need to get that 3.0 guide up, we've got a lot of nice new material addressed so far.
I'm going to try and get them to check this thread out (and the other) again, and see if we can get updated numbers.
so if there's other stuff we're missing, or others have insight/clarifications, go for it.
Things that might be wrong, or could make an impact on the calculation are the following:
- Armor Penetration: Thunder Clap scales better than other abilities, but in a normal encounter, no tank wears items with armor penetration, so it won't be needed, will it?
- Mongoose:The crit chance in your character sheet is increased by Mongoose proccs; I could add this as a averaged out crit chance and add a mongoose option
- Cruelty: Does it affect Thunder Clap, or not?
- expertise dodge cap: in the FAQ in Class Mechanics Forum is it listed as 6.25 %, but I though it wound be 6.5% pre WotLK (tanktips.com says 5.75% dodge, 10.75% parry, old EJ threat 13.75% parry as last tested cap), what is true?
- expertise parry cap: in the FAQ it says 16%, but I though it wound be 11.5% (13.5%?)
I used 6.25% and 11.5 % in my calculations - so what is wrong, what is right, I am confused - or does the parry chance of bosses just vary so often?
I do not know why I used 5300 ap in my example, it is just too high. However, yesterday my stats were 4714 ap without proccs, crit chance 13.91 with Mongoose, without Elixir of Mighty Agility.
Another thing to add to the slow vs. fast weapon discussion is, that slower weapon have a higher defensive value.
With a 1.6 Weapon (Last Laugh) you attack 1.3 times each second, with a slow weapon 2.5 (Broken Promise) you attack 1.07 times each second - that is about 22% more parry hasted attacks with Last Laugh.
Because parry hasted attacks are only dangerous if they occur more than once within 1-2 seconds the multiplied chances give a slow weapon a superior safety against parry hasted attack.
As an example, assuming 13.75 % parry chance, 8.5% expertise:
Last Laugh:
- double parry hasted attack in 3 attacks:
- triple parry hasted attack in 4 attacks:
Broken Promise:
- double parry hasted attack in 3 attacks:
- triple parry hasted attack in 4 attacks:
4% expertise:
Last Laugh:
- double parry hasted attack in 3 attacks:
- triple parry hasted attack in 4 attacks:
Broken Promise:
- double parry hasted attack in 3 attacks:
- triple parry hasted attack in 4 attacks:
Edit: My math was wrong. Maybe one who is good at calculating probabilities can do the math with those stats for an encounter of 300 second length.
This math only shows the chance of the cases 2 parries in 3 attacks and 3 parries in 4 attacks, with the added numbers of for example 3 parries in 5 attacks (what might be deadly as well) the chances are increased furthermore, increasing the defensive value of slow weapons.
You can also see how good expertise really is.
So if a encounter is not threat sensitive I wound wear a slow weapon, and so I will be wearing Broken Promise at some Ulduar Encounters, even if Last Laugh will drop.
To get back to Rend, like Xav said on tankspot.com, it is advisable to drop Rend with a fast weapon completely from the rotation, as it is even with imp Rend, and Mangle/Trauma a insignificant threat increase. If you focus on your other abilities only, you are likely to do more TPS.
With a slow weapon it is okay to use Rend, but it does not increase your TPS a lot. Yesterday I did squeeze in Rend every time there was the opportunity to, and it reached only as much as 3 % of total DPS, as where Deep Wounds, as a comparison, reached 9%.
Last edited by Yomon : 01/09/09 at 4:18 AM.
Reason: wrong math
Another thing to add to the slow vs. fast weapon discussion is, that slower weapon have a higher defensive value.
With a 1.6 Weapon (Last Laugh) you attack 1.3 times each second, with a slow weapon 2.5 (Broken Promise) you attack 1.07 times each second - that is about 22% more parry hasted attacks with Last Laugh.
This is only true some of the time as Blizzard is turning off Parry-haste for more and more boss mobs these days.
To get back to Rend, like Xav said on tankspot.com, it is advisable to drop Rend with a fast weapon completely from the rotation, as it is even with imp Rend, and Mangle/Trauma a insignificant threat increase. If you focus on your other abilities only, you are likely to do more TPS.
I second that, last Naxxramas I was trying to keep rend up almost all the time with Last Laugh and the threat increase or decrease seemed so minimal that it's not even worth bothering, I had all the melee buffs, unfortunately I can't provide a WWS as the text file bugged out.
I've also tested out Heroic Strike glyph and the numbers seem to match up Xav's post on tankspot. Glyph of Devastate seems better 90% of the time as there are many situations where you will want the debuff up on targets fast, stuff like Sartharion with 3 drakes on 10 man comes into mind while tanking the adds and not the drakes, quickly applying the debuff can be quite useful to the people working on the drakes.
However from a pure TPS point of view even though I love my Improved Charge, you can theoretically spec both 2/2 Improved Rend and 3/3 Heroic Strike. That would mean that Heroic Strike glyph would actually be making 111,6 TPS (with a 74,4 threat per rage rate). Not that it's useful information but it's not that far behind in a fight like Patchwerk where your heroic strike crit is 30% due to it being close to an infinite rage situation. Again it's a matter of preference as Devastate glyph will prove useful in more situations such as in rage-starved ones as many of you have said.It is also superior in an infinite rage situation as well according to the several posts and calculations,if that is the case this is a no brainer.
Originally Posted by nakoda
I've gotten the impression from the discussion here that devastate isn't an every GCD ability any more. The concern I've seen brought up is the S&B procs that come from it. Maybe I missed more clarification in a later post, but Revenge also procs S&B, so devastate is more of a 'get the sunders up fast with [Glyph of Devastate]' and then the [Glyph of Revenge] combined with a fast weapon for fast high dmg HSs and revenge based S&B procs is a priority. Devastate and TC would then be when refresh needed abilities, and HT, Conc, and Shockwave being thrown in when their CD is up.
Have I understood correctly where the discussion is leading?
Pretty much, yes. Last pages are about usage of rend with a slow weapon being worthwhile over other abilities. Devastate isn't quite worth it even when with SaB proc when Shield Slam is only 1 GCD away as it will only be a rage benefit, not threat benefit. So in my opinion it is still good to open up with a Devastate in hopes of a fast SaB proc as it is not only good threat (Devastate /w S&B procc possibility 2393 according to post #76 even though it assumes an AP of 5k+) but considering you would want to use Glyph of Devastate it is a bit of a minimal DPS increase for your entire raid to get those sunders up fast.
EDIT:A short explanation what Devastate /w S&B means. If you use Devastate as the first or second special after Shield Slam it has a higher value than the third special, because Shield Slam will be up after the third special anyways and a S&B procc wound be wasted. A S&B procc is calculated as a Shield Slam instead of a Devastate
As it seems I didn't clarify this properly earlier: that calculation is very insufficient for use in an ability comparison. The value of an S&B proc isn't dependent on just the damage difference between SS and Dev but also by how you reduce the SS cooldown with the proc (i.e. first or second special) and what other abilities you delay in doing so. The error caused by not accounting for any of that is large enough that the posted threat value for Devastate with S&B Procc is entirely meaningless.
Things that might be wrong, or could make an impact on the calculation are the following:
- Armor Penetration: Thunder Clap scales better than other abilities, but in a normal encounter, no tank wears items with armor penetration, so it won't be needed, will it?
- Mongoose:The crit chance in your character sheet is increased by Mongoose proccs; I could add this as a averaged out crit chance and add a mongoose option
- Cruelty: Does it affect Thunder Clap, or not?
- expertise dodge cap: in the FAQ in Class Mechanics Forum is it listed as 6.25 %, but I though it wound be 6.5% pre WotLK (tanktips.com says 5.75% dodge, 10.75% parry, old EJ threat 13.75% parry as last tested cap), what is true?
- expertise parry cap: in the FAQ it says 16%, but I though it wound be 11.5% (13.5%?)
I actually recently started running some tests on the parry chance. Stacking expertise and running DW Last Laugh/Broken Promise.
13.00% - Received a parry in the first minute
13.75% - Received a parry in about five minutes
14.00% - So far with over 6000 attacks, I have received no parries. I am going to continue to test this until I have performed near 20k attacks with zero parries before I assume that 14% would be the cap.
I should be done with the tests by tomorrow evening so I will post the final results (and parses) when complete.
The difference between Thunderclap and Devastate is small, in this example I'll use Thunder Clap prioritized over Devastate.
In a 18 sec cycle you use at least 3 Revenge and 3 Devastate, the possibility of 2 S&B proccs is quite high, so that you can use one Shockwave every three cycles.
It is unlikely to have four S&B proccs in four cycles, so that you can use concussion blow every five cycles.
That means Devastate threat with S&B procc possibility will be:
0.6x (Devastate + (Shield Slam - Thunder Clap)*0.3)
+ 0.2x (Devastate + (Shield Slam - Shockwave)*0.3)
+ 0.2x (Devastate + (Shield Slam - Concussion Blow)*0.3)
Now is that correct? Shall I implement it that way?
The value of Sword & Board is too situational to allow us to fix a static value to the Devastates that can give us useful procs at all.
Generally, we can work using the assumption that a Devastate is the "baseline" threat per GCD: it's what you'd definitely do if no other abilities are available. If we ignore miss chances then that yields your threat increase from the chance to get Sword and Board as (Shield Slam - Devastate) * 30% * (Cooldown gain on Shield Slam) and your threat loss from delaying some other ability or abilities as (Other - Devastate) * 130% * GCD / (Cooldown on other), summed for all delayed abilities. Any sort of mean value approximation is simply going to be so significantly off that it doesn't give us any useful information as to when we should and shouldn't choose Devastate over other attacks.
The spreadsheet and tables I posted earlier in posts 58 and 75 do the relevant comparison for single abilities on cooldown which is useful but still a simplification. I think the general conclusion we can draw from them is that Devastate is always worth it immediately after Shield Slam and often worth it 2 GCDs after if no more than one other significant cooldown is delayed by doing so. Revenge is always worth it in either the 1st or 2nd GCD slot after Shield Slam, regardless of how many cooldown abilities are delayed by using it. I don't think it's feasible to fix Revenge as always after Shield Slam, the stochastic nature of both it and Sword and Board will make such a rotation unsustainable.
The difference between Thunderclap and Devastate is small, in this example I'll use Thunder Clap prioritized over Devastate.
In a 18 sec cycle you use at least 3 Revenge and 3 Devastate, the possibility of 2 S&B proccs is quite high, so that you can use one Shockwave every three cycles.
It is unlikely to have four S&B proccs in four cycles, so that you can use concussion blow every five cycles.
That means Devastate threat with S&B procc possibility will be:
0.6x (Devastate + (Shield Slam - Thunder Clap)*0.3)
+ 0.2x (Devastate + (Shield Slam - Shockwave)*0.3)
+ 0.2x (Devastate + (Shield Slam - Concussion Blow)*0.3)
Now is that correct? Shall I implement it that way?
I think to model it accurately becomes is even more complicated than this. You're not replacing a Shockwave with a Shield Slam when S&B procs, you're delaying its use. Assuming you prioritize Shockwave over CB, for example: if you proc S&B during the Shockwave rotation, and you replace a Shockwave with Shield Slam, this just means you replace your next CB with Shockwave. And subsequently you replace your next Thunderclap with the CB that you delayed. And replace the next Devastate with that Thunderclap. This is assuming a strict Shield Slam > Shockwave > CB > Thunderlcap > Devastate priority for now, for simplicity.
At first glance this might seem like it could be measured with (Shield Slam - Shockwave) + (Shockwave - CB) + (CB - Thunderclap) + (Thunderclap - Devastate), which is of course the same as Shield Slam - Devastate, which was the original modeling, but there's also the effect of chain S&B procs to consider. Take for example the effective threat of a devastate which procs a S&B which pushes back a revenge. Consider:
- Shield Slam, Revenge (S&B procs), Shield Slam, Devastate (no S&B), (GCD in question)
Now, in the questionable GCD, we could use a Devastate, we could wait 0.5 seconds and use a revenge (probably not ideal, though worth mentioning), or we could use a long cooldown move. If we devastate, and it procs S&B, we're pushing back our revenge an extra GCD. The threat on that Devastate's potential S&B is worth slightly less than one in the typical Shield Slam, Revenge, (GCD in question) scenario.
Yes indeed, I now see the whole problematic. But maybe we don't have to model it accurately - only that accurate that we can surely say Devastate with Sword and Board proc is worse or better than a special move.
We can say Devastate with S&B proc possibility is better than Thunderclap or Rend with fast weapons and it is worse than Revenge, Shield Slam, Concussion Blow, Shock Wave and maybe Heroic Throw (it does reset the swing timer, doesn't it?).
Things that come close are Rend with a slow weapon, and Rend with a fast weapon above 75% - and there we are at the next problem: Rend is good on paper, but is it good in practice?
Yes indeed, I now see the whole problematic. But maybe we don't have to model it accurately - only that accurate that we can surely say Devastate with Sword and Board proc is worse or better than a special move.
We can say Devastate with S&B proc possibility is better than Thunderclap or Rend with fast weapons and it is worse than Revenge, Shield Slam, Concussion Blow, Shock Wave and maybe Heroic Throw (it does reset the swing timer, doesn't it?).
Things that come close are Rend with a slow weapon, and Rend with a fast weapon above 75% - and there we are at the next problem: Rend is good on paper, but is it good in practice?
I don't think we can even claim that definitively, as the S&B proc possibility varies in value depending on the remaining cooldown on Shield Slam. A S&B proc in the first GCD after a Shield Slam is very valuable. A proc in the last GCD before your next shield slam has zero value (in an infinite rage situation). The problem comes in trying to measure that value for the 3 GCDs after a Shield Slam.
I think tank skill priority may have reached Enhancement Shaman levels of complexity, where we need to develop a sim to determine priority with a given gearset rather than relying on rules of thumb -- especially since your individual priority will fluctuate based on your stats and weapon speed.
It's all academic at the moment, as nothing (except possibly Malygos) requires any amount of sustained TPS that pushes current limits, but it's an interesting question and possibly more useful going forward into Ulduar and subsequent tiers of content.
Something related that may be of more immediate interest to a lot of tanks is a priority for achieving maximum DPS while tanking. While it would be similar to a max-TPS priority list, there would be some potential differences.
Yes indeed, I now see the whole problematic. But maybe we don't have to model it accurately - only that accurate that we can surely say Devastate with Sword and Board proc is worse or better than a special move.
We can say Devastate with S&B proc possibility is better than Thunderclap or Rend with fast weapons and it is worse than Revenge, Shield Slam, Concussion Blow, Shock Wave and maybe Heroic Throw (it does reset the swing timer, doesn't it?).
Things that come close are Rend with a slow weapon, and Rend with a fast weapon above 75% - and there we are at the next problem: Rend is good on paper, but is it good in practice?
One problem I have with using rend, especially with a fast weapon (I haven't gotten a broken promise yet), is that it's another in a long line of keybinds. Which would take some muscle memory training to get used to hitting. If it's only a small gain, is it really worth it? IMO, no.
I think you're overcomplicating things a bit. I agree that it's complicated to figure out exact tps numbers or making a model, but that doesn't mean you can't figure out what "cycle" will produce the highest threat on average. I think it's all a question on looking forward at least 1 gcd. Using Yomon's values above for fast weapons, we get:
1.a Is Heroic Throw off the cooldown? Use Heroic Throw
1.b Is Shockwave off the cooldown? Use Shockwave
1.c Is Shield Slam off the cooldown? Use Shield Slam
2. If either of the big three above comes off the cooldown on the next gcd there is no (tps) reason to factor in a S&B proc. If that is the case, use the highest threat move off the cooldown:
- Concussive Blow > Revenge > Rend above 75% > Thunder Clap > Devastate
3. If none of the big three comes off the cooldown on the next gcd and Revenge is off the cooldown use Revenge, even if CB are available. (0,7*(Revenge+CB)+ 0,3*(Revenge+SS)) > (CB+Whatever is on cooldown will average lower than a 3 out of 10 chance to proc S&B + 7/10 Rev+CB)
4. If revenge is on cooldown, too, it's a question of wether a Devastate S&B proc will produce more threat than using Devastate and the highest threat move coming off cooldown on the next gcd.
4.a If it is CB, the choice is between (X + CB) and (0,7*(Dev+CB)+0,3*(Dev+SS)). This gives that average tps for both Rend 75% and Thunderclap will be higher than hoping for a Devastate and a S&B proc .
4.b If it is Revenge, the choice is between (X + Rev) and (0,7*(Dev+Rev)+0,3*(Dev+SS)), regardless off wether it procs S&B. X gives a higher average tps for CB, Rend 75%, but not for Thunderclap
4.c If it is Rend 75% is coming off the cooldown, you want to use CB instead of Devastate, but not Thunderclap.
4.d If it is Thunderclap, you want to use CB or Rend 75% before Devastate.
4.e If the only move of cooldown on the next gcd is Devastate, you'll want to use CB or Rend 75% before Devastate, if possible, but not Thunderclap!
Conclusion (using Yomon's values for fast weapons):
- Use HT, SW, SS when they are up and in that order
- If not, use Revenge if it's up, unless HT,SW or SS is up next in which case it's better to use CB
- If not, use CB if it's up
- If not, use Revenge if it's up, even though HT, SW or SS is up next
- If not, use Rend 75%, if it isn't on already
- If not, use Thunderclap, but only if the next gcd cools off either the big three or CB
- If not, use Devastate
- Don't use Rend under 75%
I really don't think that abilities should be ranked higher based on S&B procs. The chance to proc, %30 is too low to be reliable. At the moment any S&B procs should be seen as a bonus to our threat, just like a crit would be. Even though you may regard a certain ability higher based on an increased chance to crit.
Single Target Threat Priority (when Thunder Clap and 5 Sunders are on the target)
1. Shield Slam
2. Revenge
3. Devastate to maintain sunders
4. Shockwave
5. Concussion Blow
6. Heroic Throw
7. Devastate
Thunder Clap
I see it primarily as a debuff mechanism and from my WWS logs its equal in threat and does not crit as often or for as much as Devastate.
Deep Wounds/Devastate v Rend
1. Deep wounds is automatic and does not take up a GCD, plus its potential to scale is much higher as a good run of crit RNG will cause it to stack.
2. Rends threat is generated over 15secs, its threat generation would have to be far superior to devastate to see it as generating more threat over time. At the moment with Shield Slam/Revenge being so frequent and there being a much smaller window to add in an additional GCD ability plus the fact that devastate crits I see Rend as being inferior for the time being.
Rend may be useful at the start of an encounter if you spec for it and have trauma/mangle, however most of the time I'm trying to maximise my threat while putting up my sunders and then my thunder clap de-buff.
I tried to model threat for S&B procs a while ago but ended up giving myself a terrible headache. I think one of the things that was most confusing was that S&B procs didn't give us a free SS after the proc (and hence the SS threat - next ability threat *0.3 calculation) but only pushed the next Shield slam forward by 1-2 GCDs (and delaying the rest of the 'rotation' by 1-2 GCDs resepctively). I tried to model a 3 minute long 'rotation' with S&B but I gave up because I was confusing myself too much.
In the end when I modeled my threat for a better rotation I just took the assumption that since my rotation was delayed the only ability that would be 'lost' by the end of the fight would be the filler ability: Devastate. I therefore just took shield slam threat - devastate threat *0.3 as the threat value for S&B abilities.
In order to model S&B threat properly I think someone good at programming may have to compare between a prioritized rotation without S&B procs for a set amount of time, and compare it to a rotation with S&B procs in the same amount of time, and see the difference between the abilities that were replaced by the more frequent shield slams.
I think you're overcomplicating things a bit. I agree that it's complicated to figure out exact tps numbers or making a model, but that doesn't mean you can't figure out what "cycle" will produce the highest threat on average. I think it's all a question on looking forward at least 1 gcd. Using Yomon's values above for fast weapons, we get:
1.a Is Heroic Throw off the cooldown? Use Heroic Throw
1.b Is Shockwave off the cooldown? Use Shockwave
1.c Is Shield Slam off the cooldown? Use Shield Slam
This is demonstrably false: you're making the assumption that if an ability does higher threat on application then it should have priority, but this isn't the case. Heroic Throw might hit hard, but it's on a one minute cooldown and even delaying it by 4 GCDs is unlikely to cause you to loose an HT over the course of the fight whereas doing the same with Shield Slam would definitely cost you one of those over the course of the fight. Shockwave, on the other hand, is on a 20 second cooldown and any delay is 3 times as significant compared to HT. Shield Slam is on a mere 6 second cooldown, delaying it is 10 times as significant as delaying HT.
To make this perhaps more clear, let's take a simpler example that we can calculate in full: imagine we have only two threat moves, ability A doing 5 threat on a 3 GCD cooldown and ability B doing 4 threat on a 2 GCD cooldown. I imagine most people think 5 > 4 so A always has priorty, which leads to a 3 GCD cycle of A > B > 0, for (5+4)/3 = 3 threat per GCD. However, the correct decision here is to prioritize B, as that gives a rotation of B > A > B > 0 for (5+2*4)/4 = 3.25 threat per GCD. The reason for this is essentially because delaying a 3 GCD attack for 1 GCD is just a 33% loss whereas doing the same for a 2 GCD attack is a 50% loss. 50% of 4 > 33% of 5.
This gets considerably more complicated as you add more attacks and a baseline 1 GCD ability like Devastate, but the eventual rule of thumb will be to prioritize the ability with the highest (Ability Threat - Devastate Threat)/(Ability Cooldown), i.e. make sure the short cooldown abilities that do significantly more than Devastate are on cooldown as much as possible. This means your base priorities, ignoring Sword and Board completely, will be something like SS > Rev > SW > CB > TC > HT > Dev.
Sword and Board further muddies the water because we're still not weighing 30% of Shield Slam threat Concussion Blow threat or somesuch; we're weighing an increase in Shield Slam frequency, with the significance of the increase depending on when the proc attempt is done, against a smaller decrease in the Concussion Blow frequency. To again take a simple example, let's imagine that Concussion Blow is the only ability that's off cooldown immediately following a Shield Slam. Using Devastate here is a 30% chance of halving the cooldown to your next SS, for 15% of an SS over the course of the fight. The corresponding delay on CB is 1.5 seconds plus another 1.5s 30% of the time, for a 1.95s average, which is a 6.5% chance of loosing a whole CB. 15% of SS > 6.5% of CB, so you should prioritize Devastate. If CB had, say, an 8 second cooldown then this would not be true.
TL;DR: It's not enough to look at threat per application to determine ability priorities, you need to factor in cooldown length, what other cooldowns you effectively lengthen by delaying ability use and for S&B what cooldowns you shorten.
01/07/09, 11:11 AM
Similar Threads
