The next question is when, if ever, should immolate be clipped when you get the emp imp buff.

Assume an Immolate tick on live does

*x* damage. Then (ignoring Chaotic meta for simplicity) a crit does 2

*x* damage, and the expected value of the DoT from one cast of immolate is 5

*x*(1+

*c*) where

*c* is your crit chance without Emp Imp. Similarly if the direct damage is

*y*, a crit does 2

*y* and the expected damage is

*y*(1+

*c*).

If you cast Immolate under the effect of Emp Imp the DoT will do 10

*x* damage, vs. 5

*x*(1+

*c*) that it would have done over that time otherwise, for a gain of 5

*x*(1-

*c*). The direct damage will also crit for a gain of

*y*(1-

*c*), totalling (5

*x*+

*y*)(1-

*c*). (Notice the negative scaling with crit here. Emp Imp tends to do that.)

The time you spent casting this Immolate could have been spent on an Incinerate, however. If Incinerate does

*z* damage on a hit, 2

*z* on a crit, then you gain

*z*(1-

*c*) by casting it under the effects of Emp Imp, and this is how much you lose by casting Immolate instead.

Accounting for the loss of a tick or two of Immolate is a little more difficult. It's tempting to say that you lose

*x*(1+

*c*) damage for each tick on the DoT that you're overwriting, but notice that this assumption yields a DPS loss for immediately overwriting Immolate with an Empowered Immolate, even if you weren't planning on casting any Incinerates--which is obviously wrong. The way you actually pay for these ticks is in having to re-cast Immolate a little sooner, losing 1/5 of the cast time of an Immolate from your Incinerate spam for every overwritten tick; e.g. if your Immolate takes 1.25s to cast and you overwrite 1 tick, you lose .25s of Incinerate spam on average. (I'm obviously simplifying some things to make it mathable.)

If you aren't haste capped, the ratio of cast times of two spells is the same as the ratio of their base casts. So one tick of Immolate is worth 2/15 of an Incinerate, or (2

*z*/15)(1+

*c*)

So the net damage of clipping

*n* ticks of Immolate in order to reapply it with Emp Imp is:

(5

*x*+

*y*-

*z*)(1-

*c*)-(2

*nz*/15)(1+

*c*)

Now at this point we could start sticking in some real numbers from real-world parses. But I'm going to try and keep going. Plugging in spellpower coefficients as the damage values and letting

*n* = 1 to give a rough idea. This is sort of a best-case scenario for the clipping strategy; being haste-capped, as under the effects of Backdraft, will make Incinerate look comparatively better, as will having to clip more ticks than one. According to wowwiki, the spellpower coefficients in question are

*x* =

*y* = 0.2,

*z* = .9143.

(5*.2 + .2 - .9143)(1-

*c*) - (2

*n* * .9143 / 15)(1+

*c*) =

.2857 (1-

*c*) - .1219 (1+

*c*) =

.1638 - .4076

*c*So we see a slight positive scaling with spellpower from this strategy balanced by a strong negative scaling with crit chance. The break-even point should be around 40% crit (.1638 / .4076 = .4019); higher crit makes it never be worth clipping Immolate to apply an Empowered Immolate, while lower crit could make it theoretically a damage boost to do so.

It looks, in other words, like there are indeed situations where you gain damage by clipping Immolate. I'm loathe to actually recommend this practice though, so I'm kind of hoping that putting actual numbers in the equation reduces the crit threshold to an absurdly low level, e.g. "If you have more than 15% crit you should never do this".

Edit: A couple things occurred to me after I posted this. This post is long enough already so I'll only mention them briefly:

a) Set bonuses that give Incinerate extra crit% make the strategy of clipping Immolate look better, in a way that could be calculated by a minor modification of my formula.

The well-known problem situation, where there's too much time left on Imm to re-cast without clipping, but too little to finish an Inc cast before Imm falls off, also plays into this question. I'm guessing that if you manage to be in this situation and *also* happen to get an Emp Imp proc, the best thing to do is clip Imm, but this is very much a guess.