Ask Mr. Robot is a gear optimizing site — you import your armory, push a button, and Mr. Robot recommends what gems, enchants, and reforging you should do. The site allows you to enter your own stat weights, in addition to using its own.
Recently, Revulva from the Mr. Robot team wrote a long dissertation on some research into haste plateaus for MM and SV hunters, and concluded that haste plateaus don’t matter. You can read the SV one here, and the MM one here.
In the research he attempted to make a more real world analysis, rather than a theoretical mathematical analysis. Thus for his modeling he used real gearsets (so adding haste means subtracting another stat, & vice versa) and used Female Dwarf and Simulationcraft to model the dps.
Now, we know that Female Dwarf isn’t always great at modeling and weighting haste, but I was very curious about his Simulationcraft results. One thing that quickly jumped out at me was this sentence:
Now, the theory goes on to suggest that adding more Haste after this point will not increase your damage until you reach a point where casting a 4th CS, delaying your ES, does as much damage as 3 CS and waiting to ES on the 6 second cooldown.
This alarmed me a bit because this is not the theory about haste plateaus at all. However, that doesn’t mean his research is wrong — it was an alarming statement, but he was still looking at stat weights with and without plateaus. So I figured I’d try to get to the bottom of it. I contacted Revulva and we chatted on Skype last night for over hour. I found him delightful and had a great time talking theorycraft. Before I get into the details there, let me briefly outline the theory of haste plateaus:
Haste Plateau Theory Recap
The theory says that haste plateaus are a practical thing that informs two hunter decisions: the big one is that haste plateaus suggest when it’s a good idea to adjust your rotation, pushing back your signature shot a bit to get in an extra stead/cobra/arcane shot.
The second part is that haste plateaus suggest when the dps value of one point of haste changes. After a soft plateau and up to a hard plateau, the value of each point of haste is — in theory — more than the value of one point of haste after the hard plateaus. That’s because after the soft plateau haste is increasing the rate of your signature shot fire — in addition to all the Other Stuff: increasing auto-shots, increasing focus, increasing pet’s melee attacks, increasing pet focus.
But after the hard plateaus, haste is still doing all the Other Stuff. It’s just no longer also boosting your signature shot. Thus haste always increasing your dps, but at certain points it increasing your dps more than other points.
The theory then suggests that the difference in haste weight was enough that it altered reforging priorities (by a small amount). And of course there is also the rotation break points.
Revulva on Haste Plateaus
One of the first things we talked about was that quote where I worried he was misunderstanding the point of the plateaus — but I think instead he just misspoke when he wrote that. He clearly understood haste plateaus.
While he said it’s been a long time since he played a hunter, Revulva does spend a lot of time using various tools, including his own, to model dps for various classes for Mr. Robot. Once of the first things that jumped out at him was how hunters were concerned about haste plateaus, and other classes weren’t, even though they had similar enough mechanics that if plateaus mattered for hunters, they should matter for the other guys too.
He also commented on how annoying it was, from the perspective of a Mr. Robot programmer, to have stat weights change at various plateaus. So he wanted to see what impact the plateaus really had on dps. Is this something that other classes should be paying attention to as well? Is there something unique about hunters? Or are hunters caught in a mathematical theorycrafting model that doesn’t translate to the real world (of warcraft)?
How Often Do You Use Your Rotation?
One of the points that Revulva brought up is that the theorycrafting for haste plateaus models the base rotation — for SV let’s say this is Cobra, Cobra, Cobra, Explosive — that in reality you use this base rotation far less that most people think. For example, if we’re talking about a SV rotation we obviously get to throw out the impact of those haste plateaus during Rapid Fire and Ancient Hysteria. But not only that, Revulva points out, every time you refresh Black Arrow you just saved time compared to a Cobra Shot, and killed your haste plateaus. Also every time you get to fire an Arcane Shot — and of course Lock & Load.
The point is that sure, you use your base rotation far more often than any other rotation, but you also use it far less often than you think — possibly even less than 50% of the time. That means the impact of the haste plateaus is far less than raw math suggests, because the math is based on that base rotation.
And of course the SV rotation is far more stable than the very fluid and yet constrained MM rotation.
So here’s what Revulva found after doing his Simulationcraft models:
- The stat weight of haste did change before and after plateaus. Just like hunter theory, haste is worth more after the soft plateaus and before the hard plateau.
- The amount that it changed was small — though he did note that it was a larger change at certain points for SV than for MM.
- However, he found that even during the low value for haste, it was still basically equal to crit in terms of dps contribution. Therefore by extension haste would be better than crit between the soft & hard plateaus.
- In practice, with a simulated rotation and real-world gear changes, he found that being above or below the haste plateaus didn’t have a significant impact on dps (though I believe that does depend on modifying your rotation to the optimal one for your plateau).
- It is not worth chasing haste plateaus. If you happen to be at one, great, adjust your rotation. If not, that’s great too.
- Haste does not scale linearly.
- He also noted that the difference between haste and crit was very small. Mastery was universally the worst stat, but again the difference was small.
- I’m going to stress this again: the difference in dps contribution between the hunter secondary stats is very small. If you reforge all for haste or all for crit you probably won’t even be able to see the difference.
Revulva also made the comment, “and for all I know the simulators could be wrong…” which brings us to a good point. We know that it’s really hard to model the effect of haste. In fact, modeling all of this is just plain hard in general — and even when we do get it right, either through FemaleDwarf or Simulationcraft or Frostheim doing math the hard way on a series of equation-filled pages on the screen — even when we get it right, what we’re modeling is always assuming some kind of perfect scenario.
My notepad math on haste plateaus, for example, doesn’t take rotational changes into account (other than to mention that it’ll happen and guess at a percentage of how often). FD does a better job of that, but clearly gets confused at certain break points where you can add haste and yield lower dps. And I don’t know what Simulationcraft’s flaws are — other than it’s still a mathematical simulation.
Revulva pointed out that humans are far better at making these calculations that simulators are. As long as you know your priority system and are always shooting, humans tend to be much better than any simulator at maximizing their rotation. They don’t need to do a bunch of rotation math — they just know that if the signature shot is available they use it; if not they use another filler shot. And that basic behavior is really hard to program.
I think what this really demonstrates how incredibly close the dps values of our secondary stats are — especially haste and crit. Once the values are that close the smallest differences in modeling or testing can cause different results.
I agree that it’s not worth chasing haste plateaus — I’ve always said haste plateaus aren’t something that you should pursue, but instead something to just pay attention to and adjust your rotation/reforging based on them. On the other hand, I disagree that they aren’t important: from a reforging viewpoint (a Mr. Robot viewpoint) perhaps not — but from a player viewpoint haste plateaus are very important since they tell us when to change our base rotation — when to push back our signature shot and when not to.
Personally, however, I’m still finding that crit is a bit better than haste, particularly for MM (fuzzier for SV and BM). But after talking to Revulva I’m certainly going to go over the math again this weekend.
Here’s the important summation:
- Haste plateaus are important because they tell us when our base rotation should change.
- Do not chase a haste plateau. It’s not important to be there.
- The difference between crit and haste (even when not at a good point in the plateaus) is incredibly minor. Honestly, go with whichever you prefer. Even if you make the “wrong” choice, you won’t notice.
- Revulva should totally be a guest on the Hunting Party Podcast. With Zeherah.