Monday, July 18, 2011

The Case Of Haste v. Crit

...the honorable Judge Throughput presiding.

Actually, mastery should be in there too, of course, but the title seemed punchier with just two parties. This post is really concerned with haste versus crit and mastery, though.

As any healer who wastes spends time reading about World of Warcraft knows, haste achieves its throughput at the cost of mana longevity*. In order to realize the gain of haste, you need to push more spells out in the same space of time...which means more mana burned through more quickly.

 *Disclaimer coming in the next section, druids.

However, haste provides a consistent, predictable gain to every spell (unlike critical strike). And, unlike mastery, the gains are unambiguous: when you measure heals per second (HPS), every second you remove from the denominator is going to increase the HPS ratio. It is throughput in its purest form.

This has led many people to make statements along the lines of: "If raw throughput is what you're looking for, haste is ideal. If mana efficiency/longevity is still something you need, then it becomes trickier." Explicit in this is the belief that haste is the clear winner for throughput and implicit is that critical strike/mastery are better for mana longevity.

I certainly am not going to debate the second part. Haste is clearly not the best stat to stack from a mana longevity perspective. It is the healer equivalent of the proverbial "burning the candle from both ends." You're going to get a burst of production, but at the cost of sustainability.

But Is Haste Really King For Throughput?

Let me make this disclaimer right off the bat: This entry regards non-HoT healing classes. Haste has a pretty significant effect on HoTs and without any adverse effect on mana longevity. HoT classes need to be dealt with as a separate category when it comes to haste, so none of this is meant to address druids, holy priests or even holy paladins (while perhaps not thought of as a HoT healer, their heavy reliance on Holy Radiance tends to throw them into this category).

That leaves discipline priests, primarily, and potentially restoration shamans (this article can apply to a shaman who largely tank heals and doesn't make heavy use of Healing Rain).

So for these healers, for whom haste is primarily a means to reduce cast times (though lowering the global cooldown is also potentially useful), how should we value haste even if we do not consider the longevity aspect?

Simulations and spreadsheets will generally show haste providing the biggest throughput gain (unless you are a disco priest that primarily spams shields). However, there is one major problem with this, which is not generally priced into tables of throughput per stat, though can be verified with certain simulations.

Any interruption in casting instantly devalues haste-boosted throughput. A lot. This includes purposeful, chosen interruption (a period with little to heal or a desire to regenerate some mana) and forced interruption (due to mechanics or movement). This also means that if you run out of mana mid-combat (or slow your casting due to the danger of doing so), the benefit of all your haste plunges nearly to zero.

The reason for this is obvious: Haste is all about pushing more spells into a space of time. For an encounter, this means pushing more spells into the entire encounter than you could have otherwise. To do this--to push more spells into the entire encounter using haste--you'd need to be chain-casting from start to finish. If you do not currently do that, you can already "haste yourself" by filling those periods of inactivity. If mana issues prevent you from doing so, haste will not help you. Haste will only ensure you cast the spells you already do a bit faster, actually lengthening your periods of inactivity.

Even if mana issues don't prevent you from chain-casting, if you ever stop casting to move, there's a good chance you're losing most or all of your haste gains. The amounts of time you're acquiring from haste, at these gear levels, are objectively quite small, measured in seconds over an encounter length. Those numerically small gains are significant from a throughput perspective (after all, healing power is all measured in small percentage gains)...if you're not losing those seconds here and there by not casting.

Most people aren't chain casting powerful spells from pull to kill (if you're using weaker spells in order to keep activity up, you're simply trading one form of throughput loss--inactivity--for another--low throughput spells). Even if they have strong mana regeneration and are good about finding things to do on the move, it's the very rare healer that keeps pushing out large heals with no interruption at all. If you can do that, then haste will perform pretty similarly to the simmed value and should, indeed, be tops (assuming you're not a shield spammer!).

However, for most people, haste is simply not as strong a stat as is often claimed. That is not to say it's a poor stat or useless. Far from it: there are those who actually don't care about encounter-long output and are there to provide bursts of healing over small periods of time, so inactivity later is not a concern. However, even for those people, it bears noting that haste actually has a "ramp up time" for value. It takes a reasonably significant amount of time (as compared to the encounter length) to realize the benefit of haste; that is, to have saved enough time to be able to push out one extra spell due to haste. This amount of time is on the order of a minute or more, so if you're attempting to use haste to power "bursts" over shorter periods of time than a minute, you may not find that the gains are worth the opportunity cost (of the secondary stats you could have had in place of the haste you stacked).

The Final Act

At the end of Wrath of the Lich King, it made sense to have a ton of haste. Mana pools were large, unspoiled and filled with plentiful and delicious fish. You could stand on one end of a healer's mana pool, look out over it and not see the other side even on a clear day. It felt like mana was infinite.

A maxim took hold then, known as "ABC," or "always be casting." The first step was to, as I put it above, haste yourself by filling every period of inactivity. Even if there was nothing to heal, start a cast on someone (like the tank) just in case damage happened before your cast completed. If you were a discipline priest, you were encouraged to fill any free moment putting a shield on someone. Do something--anything. Inactivity was the mind-killer.

In such an environment, haste makes plenty of sense. If you're straining to find ways to use your mana to increase your throughput, having lots of haste allows you to spend even more mana (that your cup runneth over with) to squeeze more spells into the encounter.

That is not the situation healers found themselves in, in tier 11 raids and, largely, it still isn't the situation for most healers today in tier 12 raids (as partially evidenced by the outcry from many priests about the lack of spirit cloth drops from Firelands bosses). If you do feel you're in that position, knock yourself out with haste and be sure to always be casting. However, for most healers, that situation may only arrive in the final half year or year of this expansion.

Until then, you'll want enough haste to feel "comfortable" that you're not going to lose your tank during a Greater Heal or Greater Healing Wave cast, but you don't want to overdo it. It's not head-and-shoulders better for throughput and has an adverse effect on sustainability over an encounter. Look into adding critical strike and mastery to your portfolio. As in the stock market, diversification is your friend.


  1. I feel like when you need to have a disclaimer to the effect of "This only affects my specific spec and maybe this other spec, sorta, if they heal in this exact way", then maybe you'd be better off framing the entire post as "Why Disco Priests are Special Flowers When it Comes to Haste". :P

    Otherwise, you're probably right. There may be other stats that give you more throughput overall, if you're a healer that relies primarily/solely on direct heals and you have long periods of downtime. However, don't discount the reduced casting time during moments of intense healing so easily. How many times do we complain that someone dies *just* before a heal lands? Maybe with higher values of haste that heal would have landed and saved the day. I mean, a crit does 0 healing if it doesn't land.

  2. Tank healing resto shamsns aren't that uncommon. And it's actually more generally applicable to holy paladins who specialize in tank healing (also not uncommon) and resto druids who do a lot of single-target healing (from what I've read on EJ, as I haven't raided with a druid personally). I just chose to exclude them because both druids and paladins *can* realize great benefit from haste in ways that my post isn't meant to address.

    The thing is, it doesn't require "long periods of downtime" to devalue haste. As soon as you stop casting for even a second, you massively devalue haste. Haste only shows a throughput gain (aside from HoTs) if you chain cast. Any stoppage of casting and all you got from haste was a long period of not healing.

    As to your second point, about a heal that doesn't land in time, the logic can be used in favor of any stat. Had the healer crit on the previous heal, that extra healing may have allowed the tank to survive one more hit. A similar type of argument can be constructed for mastery but since the effect of mastery differs from spec to spec, it's not possible to construct one generalized one like for haste and crit.

    As I wrote, my post is not to argue that haste is useless, even outside its effect on HoTs. In an infinite mana environment, haste is the king of secondary stats by a large margin. So, depending on how you play and how your mana expenditures play out over a fight, haste can be very valuable.