#### Re-examining Haste Breakpoints

by

, 22-11-2012 at 12:49 AM (40209 Views)
One of my “pet projects” over the last few weeks has been looking at Affliction Warlocks. This involved expanding my understanding of the class along with explaining both how I arrived at the conclusions I’ve made, and what those conclusions are, in the hope that people would understand theorycrafting a little more. As per usual with me this diverged slightly into an exploration of haste breakpoints.

What are haste breakpoints?

In order to get everyone on the same page, it’s time to catch everybody up. Haste breakpoints are the points where you go from X ticks to X+1 ticks on a particular damage or heal over time effect. They also apply to channeled casts and “cast & forget” AoE spells. Basically anything with a time component that isn’t a direct cast.

The haste rating values that you’ll have seen in one or two earlier blog posts are the ultimate end of all this breakpoint work, giving you the exact values you’ll need to hit particular points, and then progress no further. Most Casters (either Damage or Healing) will refer to these on a semi-frequent basis.

The above doesn’t really explainwhywe like the breakpoints. The logic behind it is that of efficiency, ie: getting X haste means when I cast spell Y, I get more damage/healing from it. The percentage gain will be (Z+1)/Z where Z is the number of ticks before the breakpoint. Lifebloom getting an extra tick is only a (15+1)/15 = 6.67% gain, vs Flame Shock’s (8+1)/8 = 12.5%.

This efficiency is pretty crucial to healers, improving the healing returned from every point of mana spent. There is a downside to haste for healers, which is where it increases mana consumption, so nailing those efficiency points can be quite critical. For DPS casters, on the other hand, this efficiency requirement just isn’t there as managing mana as a resource isn’t necessary for maintaining damage output.

I’ve got all that, now what?

For healers, it’s just a question of which breakpoints to aim for. For DPS casters, I’m afraid it’s time to throw the idea of haste breakpoints, or indeed, breakpoints in general, out the window.

Yes, that’s correct: no more breakpoints.

There will still be certain points where stat priorities will change, as you can see in this graph for Affliction Warlocks (based on the default T14H profile in SimulationCraft). X is the change in haste, while Y is the change in DPS.

What I’ve done here is taken a trend line for Mastery, used Excel to generate a formula for it, calculated an error range (which isn’t visible on this graph) and then shown where the DPS resulting from a change in haste is outside that error margin. I’ve changed the graph around a little, so anything above zero on the Y axis is where Haste is worth more than Mastery, and vice versa under zero.

Most of you will be looking at this going "I can see the break points as those spiky points in the graph", and you'd be correct in a way. The trick is realising that those spikes relate to a rapid change in DPS resulting from a change in haste, rather than a point where you should aim for.

In this graph you can see the increase in DPS with each increase in Haste (every data point is +10 haste on the previous) in red, while a 9 point moving average is in black. It's this black line I want you to look at, noticing that for most of the entire 8000 rating range, it's fairly linear. Yes, there are a few spikes, and even 3 dips, but overall the gain in DPS is fairly flat.

This means that whole "efficiency" argument that works for the healers isn't justifiable for DPS casters, as we'd expect to see frequent large drops in the value of haste if this were the case. In other words: adding additional haste past a breakpoint keeps increasing your DPS by a fairly normal value.

This last graph is a plot of the actual DPS values from adding/subtracting X of either Haste or Mastery, and as you can see they are both fairly close (remember in these plots, a lower Y value when X is negative means that it's a bigger DPS loss to take away that stat). You can see where the large spikes from the earlier graphs are located, but here they are fairly small by comparison as we're looking at absolute DPS figures rather than a relative change.

The entire 8000 rating range covers a DPS range of 118,000 to 142,000. This makes the 0-400 dps spikes on the first graph rather small, even if we take into account the 0.05% error, which is only 70 DPS on average. That 400 DPS is only 0.28-0.33% of the total DPS, so it may not be a worthwhile investment of time to optimise your haste levels to such specific points.

So how much haste do I need then?

This is the wrong question to be asking. The better question to ask is "how do I prioritise stats now?". It's not a simple answer though, as you'll want to focus Intellect then Hit (up to the hit cap) and then, depending on actual values, slightly massage the secondary stat values.

Using Affliction as an example, the initial weights produced go Int > Mastery > Haste > Hit > Crit if we take them at face value. However, as discussed in a previous post Hit should really be the second stat to prioritise. Of the remaining three, Haste & Mastery are obviously fairly close, while Crit falls behind in a distant fifth. The resulting priority is as follows:

Int > Hit (to cap) > Mastery >= Haste > Crit

When using this to evaluate gear, gems and enchants, changing M>=H to M>H is OK as long as when you're reforging you change it the other way to M=H, as you've already put enough of a Mastery bias on your gear by how you've selected it, and forcing more Mastery on will likely result in your stat weights getting messed up (resulting in that classic reforge X>Y leads to X<Y cycle). It's mostly about making sure you don't go mad on stacking one secondary stat when another is so close in value.

Haste breakpoints are still useful, but they are points where the value of haste is multiplied for a very short duration rather than points where haste is devalued as you pass them.

TL;DR

In short, if you're really close to a breakpoint you can try to get past it, but otherwise ignore them. You should also treat very small differences between stat weights as being no difference at all.