It seems that some classes would scale linearly with heroism (i.e. ~30% increase in DPS) while others would scale non-linearly (either >30% increase or <30% increase). I'd suggest that as a starting point for a more mathematical (or empirical) analysis, I would break all DPS classes into one of four categories:

1. Melee classes with CD limited special attacks
Fury or arms warriors, ret paladins and enhancement shaman all fit in to this category. Heroism increases their damage from white attacks, but the CD on their special attacks means that their overall increase in damage will be less than 30%

Note: A warrior in execute range gains a non-CD special attack, moving him to the next category.

2. Melee classes without CD limited special attacks
Rogues and feral druids are both imited by the energy regeneration model but both can benefit from an increase in white attacks (rogues through combat potency, druids through 2Pc T4). Both classes also maintain a DPS rotation that is centered on rupture/rip uptime (with rogues also needing to maintain SnD uptime). By increasing the number of attacks that can be done between refreshing rupture/rip, heroism should give a >30% increase in total DPS for these classes.

Hunters would also fall into this category, although the haste can put a strain on a shot rotation. Depending on the mechanics of the fight, and supporting talents, it seems the haste can be as detrimental to a hunter as it is helpful. I've seen math and argument both ways, but at best it would seem that DPS should scale linearly with attack speed.

A warrior in execute range would expect to see a a >30% damage, as their increased white attacks generate increased rage which they can convert into execute damage every GCD.

3. Magic classes with support spells
Fire mages, moonkin, shadow priests, affliction warlocks all use DPS cycles which trade a support spell for their main nuke. By increasing the number of nukes that can be cast between refreshing the support spell, these classes should expect to see a greater than 30% increase in DPS.

4. Magic classes without support spells
Shadow destro locks and elemental shaman both use a primary nuke as their main source of damage. The heroism buff should translate into a 30% increase in DPS.

I realize this is a broad oversimplification, as I did not take any cooldowns into consideration, but I believe the basic premise is sound. Classes that operate around maintaining uptime on a debuff should see a >30% increase in DPS, classe that spam a single attack should see a linear increase in DPS and classes that will see an increase in their lower DPS autoattacks should see a <30% increase.

WF totem. Now all melee are in this category.

The chance that a proc based effect (WF, trinkets, Etc) will occur is only increased by the number of extra attacks possible. Since we're specifically talking about white damage, a 30% increase in white attacks will only result (on average) in a 30% increase in WF procs. Since the yellow attacks still maintain their CD, the total benefit from heroism would still be less than 30%.

Category 3 needs to be broken down further, depending on whether the support spells (assumed to be duration- or cooldown-limited) have a higher or lower DPCT than the standard filler nuke. This is where the distinction occurs between fire mages and DoT classes: a mage has to scorch once every 30(?), and that is the only reason they use scorch, so haste decreases how much of their cast time they waste on this non-damage spell. With shadow priests and affliction warlocks, we refresh our DoTs not because of the utility but because they're actually higher DPCT than the main nuke. Adding haste means we spend less time on our higher DPCT spells and more on our lower-damage filler. While our filler damage gains more than 30%, some easy variable-juggling easily shows that if our overall damage gains more than 30%, we would have been better off not casting the DoTs in the first place. Spells limited by built-in cooldowns rather than DoT duration operate in exactly the same way, which realistically means only Elemental shamans that weave Chain Lightning.

Personally, I'm not convinced by the utility of those particular categories. For example, having a six second cooldown on MS is in a very realistic sense equivalent to needing six seconds of energy regen to mutilate (plus some additive terms for finisher frequency). I would define classes by two numbers: what fraction of their damage is hasteable, and by how much it increases their damage. For example, a mutilate rogue has about 55% of their damage hasted, and it increases that damage by 30%. A combat rogue has something more like 60-65% of their damage affected by haste, and if you can prove the extra combat potency is small enough to ignore it's also 30%. A fire mage has 100% of their damage increased by 30%, an affliction warlock has about 50% of their damage increased by 33ish% (hasting the DoTs gives more time to nuke, my napkin math says 10%), and so forth.

Having a special attack without a cooldown is significantly different than not having a special attack with a cooldown.

Since this has already been bumped, I'll note we usually swap in the heroism early in the fight and let the DPS shaman use theirs later. Most fights tend to get more intense at the end, the dps gains at the end are usually higher, and less swapping then is a bit easier.

Any chance we could get some actual math in this thread instead of "zomg my class should get moar lusts"?

So far it seems to have been shown that destro locks gain slightly below 30% from bloodlust, while mages get an average of 165% (feel free to correct me if I'm wrong). There's still no math for rogues, warriors, hunters, moonkin (with and without IS or moonfire) and elemental shaman. I'm sure noone really lusts affliction locks and shadow priests if they can help it, so that's not as important.

Well it's a four month old thread, and another thread was just locked that dealt with the same topic (Bloodlusts). It's been talked to death, in other words.

Moonkin Math:

If moonkin has zero lag, SF-spam DPS is proportional to haste. At any haste level, I'll call the SF spam DPS 100%.

Moonkin maintaining MF IS FF, zero lag, 1640 +spell (buffed), 4-piece t6, no t5, 21.4% crit in caster

FF = 0% dps, one GC every 41s.
IS = 93% dps, one GC every 13s
MF = 154% dps, one GC every 16s. (2-piece tier 6)
SF = 100% dps, the rest of the time.

I added 1s to the DoT durations, because refresh won't occur until in-progress SF completes.

Number of GC's per 30s spent on the instant casts is

30/41 + 30/13 + 30/16 = 4.91
DPS is (30/13*93%+30/16*154%) / 4.91 = 103%

At zero haste there are 20 GC's over a 30s period. 20-4.91 = 15.09 are spent on SF

(15.09*100% + 4.91*103%)/20 = 101% SF dps over 20 GC's.

As Haste goes up (at least until 50%), the number of GC's that occur every 30s goes up proportionally, and the extra GC's are spent casting SF. As Haste approaches 50%, the "Rotation" DPS approaches the "SF spam DPS". However, the rotation DPS was already only slightly above the SF spam.

In other words, for this rotation, Moonkin DPS goes up almost, but not quite, as fast as haste goes up. The difference is less than 1%.

Things like exact gear selection, lag, or rotation changes would impact this. In general though, Moonkin DPS should be pretty close to (Haste + 100%).