[Math] Bloodlust/Heroism timing - Page 2

MatsT · 03/09/09, 7:34 AM

1: You claim that it's a mathematical wash when you use bloodlust.

2: We say you are wrong. We agree that the effects of late bloodlusts are overrated and it's probably a practical wash, but since classes scale differently with haste it's mathematically better to use it at a certain point.

3: You claim that we are wrong, it's a mathematical wash.

4: We prove that we are right, it's not a mathematical wash.

5: It doesn't matter, it's too hard to model, it's a practical wash anyway.

I'm glad we finally agree. But please don't state that it's a mathematical wash and that it doesn't matter.

Edit: Sorry for double post, screwed up my edit.

Wraithlin · 03/09/09, 7:35 AM

If I am dismissive it is because you have to place a rational limit on what you can and cannot usefully discuss. The haste gains either side of execute range may well be assymetric, but until you can measure that and demonstrate that:
1) Raid dps scales better below 35% than above 35% for every raid composition.
2) The increase in raid dps of synchronizing about execute is greater than that of reliably synchronizing all other external cooldowns.
Then its throwing about examples plucked from the air. The fact that the model makes assumptions does not make it wrong until you demonstrate that the assumptions are meaningfully incorrect. Stating that it makes assumptions stating the obvious.

If someone can demonstrate one of these two things then there is something tangible to discuss. Debate is good when one party is bringing something new to the table, debating is not good when it is just endless rehashing of old ground because some people can't use the search function and read through an old post. So far, there is nothing that has been said in this post that is not already said in Bink's Blog post and comments. As I said in my first post, by all means bring some WWS parses that have been meaningfully broken down to support your comments, but don't post based on faith and examples from the aether.

[e]
When I say "its a wash" I don't mean that the diference is zero, just that it is small enough not to matter.

songster · 03/09/09, 7:44 AM

Originally Posted by Wraithlin

So you would have to be sure that, overall, your raid dps scales better with BL below 35% than it does above. Thats non-trivial thing to model

Sure, but it's not impossible either, and it should be possible to get a reasonable napkin math answer.

For instance, the standout obvious factor is that mages have an execute talent, and also have by far the best scaling from percentage haste, in that it affects 100% of their damage. Rogues have no execute talents, but also scale badly from percentage haste, in that it almost exclusively affects their white damage.

Now, which other classes have execute talents, and how do these classes scale with haste? It should certainly be possible to get a yes/no answer to the proposition "Classes with execute talents tend to scale better with haste than classes without execute talents".

Wraithlin · 03/09/09, 7:48 AM

Originally Posted by songster

Now, which other classes have execute talents, and how do these classes scale with haste? It should certainly be possible to get a yes/no answer to the proposition "Classes with execute talents tend to scale better with haste than classes without execute talents".

It would be helpful to try and quantify any effect as well.

songster · 03/09/09, 8:15 AM

Originally Posted by Wraithlin

It would be helpful to try and quantify any effect as well.

Well, let's start a list. I'll go with what I know.

Class/spec	Execute?	DPS gain from 30% haste	Notes
Rogue/Mutilate	No	~20%	White damage and IP damage scale, yellow damage and DP damage do not scale except via Focused Attacks
Rogue/Combat	No	~20%	White damage and WP damage scale, yellow damage and DP damage do not scale except via Combat Potency
Rogue/Subtlety	Yes (20% bonus below 35%)	~20-30%	White damage and WP damage scale, yellow damage group-dependent - if the group crits more from haste, then HaT damage will scale with haste.

I'll fill in better results for the first two once I've had a chance to hack the relevant spreadsheet to calculate it. HaT is poorly modelled and so you won't get a better estimate.

Wraithlin · 03/09/09, 8:35 AM

To be even more annoying; is it possible to estimate the gain in dps (as a %age) both above and below the execute break-point.

MatsT · 03/09/09, 8:51 AM

As you said, modelling exact numbers for the gains seems extremely complicated.

Mages definitely scales better than average with haste, since casters simply does.

Affliction warlocks is also a caster, but dots in general scales worse with haste. I'm not very familiar with affliction rotations and mechanics, but from these threads it seems that they should scale very good because of drain soul.

Warriors should scale very well with haste as long as it doesn't make them hit faster than every 1.5 seconds.

Not sure if we're gonna be able to calculate any real numbers. I suspect that the small dps gain of timing it is outweighted by easier cooldown timing. In addition, the two most important factors in most real fights is probably when you can get 40 seconds of uninterrupted dps, and when healing is most intense (i.e. can you drop a healer if you bloodlust during the most healing intense part). For a fight such as Patchwerk or Brutallus you should probably try to pop it somewhere between 35% and boss death minus 55 seconds, but for most more complicated fights no.

Daurs · 03/09/09, 9:41 AM

You have calculated only the "use"able CD's as far as I see, but what about trinket procs with internal CD's and high proc chances?
Shouldn't there be a bonus for popping heroism in the beginning aswell, since then all people will get their 10 second trinket procs giving massive DPS boosts aligned with heroism aswell, while if you pop it later on during the fight, this might be the case for only half the people?

Endus · 03/09/09, 10:43 AM

Originally Posted by MatsT

Ok then, i guess the example it is then...

Joe the Wizard output 100 dps. When the boss drops below 50%, his damage increases to 200 dps due to cool talents. Greg the Lizard Lord can also output 100 dps, but unfortunatly he has no talents that increases his damage at low boss health. When Greg pops his bloodlust, Joe's dps is doubled. Greg does not benefit from bloodlust at all. In conclusion.
Above 50%:
Joe: 100 dps without bloodlust, 200 dps with.
Greg: 100 dps without bloodlust, 100 dps with.
Below 50%:
Joe: 200 dps without bloodlust, 400 dps with.
Greg: 100 dps without bloodlust, 100 dps with.

Now consider the two cases of popping bloodlust early and late in the fight. Let's set boss hp to 100k

Early bloodlust:
40 seconds of 200+100dps, meaning 12k damage.
190 seconds of 100+100dps, meaning 38k damage.
167 seconds of 200+100dps, meaning 50k damage
Total: boss dies after 397 seconds.

Late bloodlust:
250 seconds of 100+100dps, meaning 50k damage.
40 seconds of 400+100dps, meaning 20k damage.
100 seconds of 200+100 dps, meaning 30k damage.
Total: boss dies after 390 seconds.

Now, can you please explain to me why it's not better to pop bloodlust below 50%? Extra bonus if you can do it without insulting me.

I would say that this is mostly because you're exaggerating the impact of Bloodlust beyond all reason, and turning a difference that isn't practically visible into one that is.

Same examination. Both do 100 DPS. Joe does +30% in Bloodlust, and +50% in Execute phase. Greg gets +20% in Bloodlust. So for Joe, the numbers are 100 (normal), 130 (normal-BL), 150 (Execute), and 180 (Execute-BL). Greg is at 100, or 120, depending on BL. Setting boss HP to 50k to reflect my lower numbers and because a 6+ minute fight as you simmed is beyond anything we're seeing, in a straight beatdown.

Bloodlust early;
40s of 130+120=10k damage.
112.5s of 100+100=22,500 damage. Boss is now at 35%
70s of 150+100= 17,500 damage, boss dies.
Total time = 222.5s

Bloodlust late;
162.5s of 100+100=32,500 damage. Boss is at 35%
40s of 180+120=12k damage
22s of 150+100=5500 damage, boss dies.
Total time = 224.5

My magic numbers show it being worse. Which just goes to show, making stuff up with no practical basis for it isn't helpful, because if you exaggerate in the right way, you can prove anything.

Roywyn · 03/09/09, 11:03 AM

Thinking about the models used, it seems the important factor is whether Heroism is a bigger percentage DPS increase for a class above or below execute range. If it's all the same then the Heroism timing wouldn't matter.

Heroism is all the same to Mages, same for Rogues. Death Knights slightly prefer a late Execute, because more haste > more Killing Machine procs > more Frost crits > those (and Oblit) hits are improved by their execute abilitly.
(Warriors likely prefer a late Heroism if that gets their swing timer closer to 1.5s to allow better execute spam. I'm not sure how desirable Executing and Execute spam is right now for Warriors, however.)
Hunters and Paladins prefer an early Heroism, since their Execute ablities are on timers and woven into their cycles. So, they don't scale with haste, meaning that the absolute benefit from Heroism is always the same, hence the relative benefit is better at the beginning.

Also, Warlock Executes are getting overhauled for 3.1 and many other classes change as well. So this can easily change. Taking Shizzle's table and trying to get that information into it would look like this:

Class	spec	Execute?	DPS gain from 30% haste	Notes
Rogue	Mutilate/Combat	No	~20%	Whites damage and poisons scale, yellow damage and poisons scale via Focused/Potency
Rogue	Subtlety	+20% damage <35%	~20-30%	White damage and poisons scale, yellow damage group-dependent - if the group crits more from haste, then HaT damage will scale with haste.
Mage	Fire/Frostfire	+12% damage <35%	~27%	LB is a DoT; LB/Pyro will hit the 1s GCD cap
Mage	Arcane/Frost	No	~29/30%	Mostly 1:1 haste scaling, Arcane minorly impacted my mana cost incease
Warlock 3.1	37+ Demo	Hasted Soulfire cycles <35%	~25% above, ~30% below execute	Numbers are guesstimates; anad without pet, which doesn't matter for us
Warlock	33+ Affli	+12% (shadow) damage <35%, (Drain Soul)		Same scaling above and below Execute without Drain Soul
Hunter/Paladin		Kill Shot/Hammer of Wrath below 20%		Those have a cooldown; early Heroism is a bigger % increase (same absolute)
Death Knight	Frost	+12% IT/HB/OB/FS below 35%		Death Knights very slightly favour a late Heroism (Haste improves IT/HB/FS crit)
Warrior		Execute below 20%		Heroism scales by getting your swing timer to 1.5s so that you can Execute every GCD

songster · 03/09/09, 11:33 AM

Originally Posted by Roywyn

Thinking about the models used, it seems the important factor is whether Heroism is a bigger percentage DPS increase for a class above or below execute range. If it's all the same then the Heroism timing wouldn't matter.

Not true, it also matters how much of your raid DPS comes from classes which scale poorly/well with haste.

To see why, let's take JoeRogue and BobMage. JoeRogue does 100 DPS, and gets a 20% boost from Heroism. BobMage does 100 DPS, and gets a 30% boost from Heroism. BobMage also does 50% extra damage when the boss is under 50%. Let's look at some example raids.

1) A raid of 25 JoeRogues.
Total raid DPS at the start or end = 25 * 100 = 2500
Total raid DPS at the start or end, with Heroism = 25 * 100 * 1.2 = 3000 (20% increase)
Heroism is a 20% boost when used at the start or the end of the fight, so timing is irrelevant.

2) A raid of 25 BobMages.
Total raid DPS at the start = 25 * 100 = 2500
Total raid DPS at the start, plus Heroism = 25 * 100 * 1.3 = 3250 (30% increase)
Total raid DPS at the end = 25 * 100 * 1.5 = 3750
Total raid DPS at the end, plus Heroism = 25 * 100 * 1.5 * 1.3 = 4875 (30% increase)
Heroism is a 30% boost when used at the start or the end of the fight, so timing is irrelevant.

3) A raid of 24 JoeRogues and 1 BobMage
Total raid DPS at the start = 25 * 100 = 2500
Total raid DPS at the start, plus Heroism = (24 * 100 * 1.2) + (1 * 100 * 1.3) = 3010 (20.4% increase)
Total raid DPS at the end = (24 * 100) + (1 * 100 * 1.5) = 2550
Total raid DPS at the end, plus Heroism = (24 * 100 * 1.2) + (1 * 100 * 1.5 * 1.3) = 3075 (20.6% increase)
Heroism is a very slightly larger percentage boost when used later in the fight.

4) A raid of 15 JoeRogues and 10 BobMages
Total raid DPS at the start, without Heroism = 25 * 100 = 2500
Total raid DPS at the start, with Heroism = (15 * 100 * 1.2) + (10 * 100 * 1.3) = 3100 (24% increase)
Total raid DPS at the end, without Heroism = (15 * 100) + (10 * 100 * 1.5) = 3000
Total raid DPS at the end, with Heroism = (15 * 100 * 1.2) + (10 * 100 * 1.5 * 1.3) = 3750 (25% increase)
Heroism is a slightly larger percentage boost when used later in the fight.

4) A raid of 1 JoeRogue and 24 BobMages
Total raid DPS at the start, without Heroism = 25 * 100 = 2500
Total raid DPS at the start, with Heroism = (1 * 100 * 1.2) + (24 * 100 * 1.3) = 3240 (29.6% increase)
Total raid DPS at the end, without Heroism = (1 * 100) + (24 * 100 * 1.5) = 3700
Total raid DPS at the end, with Heroism = (1 * 100 * 1.2) + (24 * 100 * 1.5 * 1.3) = 4800 (29.7% increase)
Heroism is a very slightly larger percentage boost when used later in the fight.

Obviously it's not a large effect, but it's non-zero. The reason the effect exists is that in a case where you have a mix of rogues and mages, the rogues (who scale badly with haste) do proportionally more of the raid DPS at the start of the fight, while the mages (who scale well with haste) are a higher proportion of raid DPS at the end of the fight.

However, in practice the scaling factors for all classes are almost certainly between 20 and 30% DPS increase from 30% haste, and the execute bonus is not a full 50% (as used in the example above). Thus I would imagine that there is no more than a few percent change in the proportional benefit of Heroism when used early or late. Moreover, there may well be classes that act against the tendency demonstrated above - i.e. a class that has an execute bonus, but that scales poorly from haste. An example of the latter would be a bugged HaT rogue - they have an execute bonus, but get almost no bonus from haste since the majority of their damage is yellow damage and they are capped by the 1 second GCD spamming Eviscerates.

MatsT · 03/09/09, 11:46 AM

Originally Posted by Endus

I would say that this is mostly because you're exaggerating the impact of Bloodlust beyond all reason, and turning a difference that isn't practically visible into one that is.

I exaggerated to make it more obvious that the increase is non-zero. Also, your numbers is slightly off. 100 dps * 50% execute bonus * 30% bloodlust bonus means 195 dps, not 180. In the end, this gives us.

No bloodlust: 232.5 seconds.
Early bloodlust: 222.5 seconds.
Late bloodlust: 222.1 seconds.

This is a quite small ~0.4% dps increase overall, given that your numbers are a good estimation.

Distei · 03/09/09, 11:59 AM

As interesting as this topic is, I'd also say it's kind of pointless. As the autor assumed, 40sec with 30% more haste shortens the fight by about 10 seconds - let's not discuss whether that's perfectly right, I think it's not that far off.
Assuming a 5 minute fight, 10 seconds are 3.33% of the fight (in a static fight also 3.33% Boss-HP).

Now, even IF saving Bloodlust for the 35% phase is beneficial it would be in the range of 1% of the fight duration... at max - so about 2-3 seconds.
Sure, if we had encounters like they were in Sunwell these 2-3 seconds might be important but since we don't have such fights the only difference it might make is that you finish a whole Naxx-Run 3*16 = 48 seconds earlier.

And even if we had that hard encounters, using Bloodlust in the right moment (no movement, weaknesses, DDs have CDs ready) probably makes a bigger difference than those 2-3 seconds we would get if the under 35% theory is true.

Xenophon · 03/09/09, 12:30 PM

There is also a less quantifiable but important benefit associated with early bloodlust:

To bloodlust effectively, you need to be able to sync it with (most) cooldowns, but also make sure that you are not reducing the number of cooldowns used by "saving" them for the bloodlust. Now, in an ideal situation all players would know the expected duration of the fight and plan accordingly for a late bloodlust. In reality, however, there are cases in which players defer an early cooldown use in order to ensure that it is up for a late bloodlust, and effectively lose the oppertunity to use a cycle of cooldowns. An early bloodlust ensures that cooldowns are used at (near) the earliest possible oppertunity, maximizing the number of cooldown cycles possible in any given fight.

songster · 03/09/09, 12:46 PM

Originally Posted by Distei

As interesting as this topic is, I'd also say it's kind of pointless. As the autor assumed, 40sec with 30% more haste shortens the fight by about 10 seconds - let's not discuss whether that's perfectly right, I think it's not that far off.
Assuming a 5 minute fight, 10 seconds are 3.33% of the fight (in a static fight also 3.33% Boss-HP).

Now, even IF saving Bloodlust for the 35% phase is beneficial it would be in the range of 1% of the fight duration... at max - so about 2-3 seconds.
Sure, if we had encounters like they were in Sunwell these 2-3 seconds might be important but since we don't have such fights the only difference it might make is that you finish a whole Naxx-Run 3*16 = 48 seconds earlier.

And even if we had that hard encounters, using Bloodlust in the right moment (no movement, weaknesses, DDs have CDs ready) probably makes a bigger difference than those 2-3 seconds we would get if the under 35% theory is true.

Aye, in a raid situation it's not really relevant. Where Heroism / BL really does become relevant is in the 5 man content, particularly the DPS race achievements. The individual fights rarely last longer than 180 seconds, some as little as 120 seconds. Shaving 10 seconds off that is somewhere between a 5% and 8.5% boost. To put that into perspective, that's about what you get from upgrading 1/3 to 1/2 your equipment by a full tier. That's substantial.

5-man stuff may be old hat to all of us now, but there's a lot of people out there still doing it and gradually working on the achievements. There are a couple of 5-man achievements (e.g. Less-Rabi) that I would recommend people don't bother trying without a shaman - even if they're in Naxx25 gear! That's silly, especially given that 5-man groups are inherently less likely to have a shaman.