Here's a fun one. With a Deep Arcane spec build (like 58/3/0), you can apparently cap your spell haste.

How you say?

Well, because the primary spell rotation is: Arcane Blast, Arcane Barrage; and the AB debuff duration isn’t affected by spell haste. Thus you can easily get into trouble because the stacking AB debuff gets refreshed with enough spell haste even if you cast ArBr in the meanwhile.

Specifically, when the AB(2.5)+ArBr(1.5) cycle gets down to less than 3.0 seconds, you get AB debuff stacking.
Thus 4.0/3.0 = 1.33 = 33% haste = haste capped. This can easily happen with nominal amounts of gear, procs, and buffs.

After that point, adding more haste is a DPS loss, at least according to Rawr. I believe this is because stacking AB debuffs are just DPM unsustainable, forcing you into lower dps rotations later on. I don’t know if this REALLY turns out to be true given MBAM procs, which could clear the AB debuff.

Interestingly, AB2ArBr actually becomes part of the recommended Rawr spell rotations with high amount of bonus spell haste from procs/buffs. AB2ArBr seems to be LESS dps than ABArBr, because ArBr is that much better DPS than even AB+AB2. I'm still not sure how that one makes any sense at all given that it would ALSO be affected by spell haste capping issue.

The c*c/1+c model holds up very well versus the data I generated, not just for fireball spam but for spam of any of the spells allowed. It's also perfectly good for a scorch/fireball rotation, since nothing in that doesn't proc the buff.

For a living bomb rotation, or something more complex it's still pretty easy to work out the impact, at least to 1-2 significant digits. Well easy with a spreadsheet anyway. The approach is something like this.

1. Pick a number of castings for your chosen rotation. I'll do a simple example to illustrate what I'm talking about, with numbers chosen for math convenience. Lets say that whatever your combination of haste+latency you can manage reliably 9 fireballs and one scorch in 30 seconds.

2. Calculate the % chance of hot streak. for that rotation . Lets pretend it is 50% crit rate, so 1 in 6 castings will proc hot streatk.

3. For ease of math, lets expand ths to 6 cycles, ie 180 seconds, 60 castings. You'd expect to get 10 hot streak procs, so opportunity to cast 10 pyroblasts in there. Now we're looking at ~15 seconds more casting so you should work in another scorch too, just to make sure the stack stays up. Call it 20 seconds total with extra lag from trying to chaincast instant spells and interruptions to your flow by tracking hotstreaks. (a skilled player could do better, but we're working with round numbers here)

4. add up the total damage of casting of 54 fireballs, 7 scorches and 10 pyroblasts, factoring in the effect of crits but NOT the dots. You can include ignite damage in this as it is based on the crit damage.

5. Divide the total damage by 200 seconds for a base DPS figure without dots.

6. Assume the fireball DOT is running continuously. Just divide total damage by 8 and add to the base DPS figure.

7. with 10 pyroblasts+12 second dot, we have in theory 120 seconds of dot uptime, or 12 out of 20 seconds have the DOT running. Take total dot damage and divide by 20 and add it to the base DPS figure+fireball DOT dps. If you want to get fancy, this should actually be reduced a bit, as some pyroblasts will "clump" within a 12 second window and it's unlikely to be up the entire 120 seconds.

You can add a bit more precision to this estimate by simulating the scorch rampup properly. Beyond that the inaccuracies in this estimate are probably larger than any further precision you'd apply.

Now we want to add living bomb to the above? No big deal. Work out how much time the GCD's for living bomb add to the equation, figure out how many you are going to cast, and normalize the dots and burst damage over the new time period. Same with any other spell mixed in that doesn't proc the buff. The key thing is in whatever rotation to calculate how many pyroblasts get procced and when added to casting time, what% of each spell is running then add DPS of each spell, multiplied by their % of the rotation. Then add in any dots based on their expected uptime.

This kind of analysis is good enough to spot differences of 5-10%. It may not absolutely match what you see in the field but as a relative measure it's probably ok. It is probably not good enough for 1-2% differences, especially given how small actual fight sample sizes are. (a 200 second fight might have 10 pyroblasts as expected, or 16 or 6...will usually have 8-12)

The biggest failing of the approach is it assumes you can truly calculate casting time and that the procs can be weaved in without messing up the scorch rotation (either too much scorching or too little). Most of the time you don't get a round number of procced pyroblasts either, so you are usually factoring in fractional castings.

I like doing stuff like this myself so I can try rotations or builds RAWR has not considered. For most things though RAWR does an excellent job, and on anything it has considered I'd believe it over my own spreadsheet.

Since the armory for some reason is showing me as a 2/28/31 with all my points in fire in the final tiers somehow, that's a pretty good question! Though, as you could deduce from the post I was a heavy fire spec with firestarter.

Not an optimal spec by any means, but it was the best I could come up with on patch day, mostly pve focused with some random pvp talents tossed in for BG hilarity; 02/59/00