Value of crit?

[Kver]

Hi!

Having recently read Voltaa's post somewhere on the value of crit at really high ilevels(>582), I was wondering how close it actually is to haste? From a simple mathematical perspective(Especially during skull banner!) it would seem to me that crit is actually really powerful, which is a bit of a no brainer, but I was wondering whether anybody has actually done maths somewhere on this? It might prove to be a very interesting discussion considering the abundance of crit on some items.

[Cycobi]

Realistically speaking because crit will never be our main stat, stacking it second won't get us to absurd crit levels. Whilst I haven't done maths on this topic yet, I can tell you straight away that Arcane won't get to a level of crit while checking all the other boxes (maxing Mastery and maintaining a good level of haste) to make a huge difference.

Crit is strong for Arcane (2.1M+ Crits on Barrage/Blast) but because we lack high crit levels (I'm running 5K Crit alongside 22.8K Mastery and 10.5K Haste), realistically speaking you're probably not going to be able to get higher than MAYBE 8K crit without falling below 9.7K Haste or losing some Mastery, bearing in mind that this value of crit is a guesstimate, if I can be proven wrong this is fine and I can look into it further.

5->8K crit at 600 rating per % means that you'd realistically gain 5% crit, pushing buffed levels to around ~30-33%,. Then, considering the fact we don't have a modifier like Critical Mass or on-proc Shatters (like BF, FoF), realistically speaking 33% isn't really enough considering what you'd actually get with the extra Haste.

Haste will *always* be beneficial for Arcane, technically even past the GCD cap because extra haste is extra mana regen thanks to Nether Attunement, which gets amplified by RoP.

The two stats are pretty close at the moment in terms of weight, but I personally don't see us getting to a point where we'd be able to check all boxes and run high levels of crit. If someone wants to/can do the maths to prove me wrong I'll back down, but I don't see it being viable, honestly.

[Akraen]

(this isn't a counter argument, just a curiosity)

Any idea why the Chinese arcane mages are doing mastery>crit>haste?

[Cycobi]

Extra ilvls (bearing in mind even after the update they still get more as they got the same update) means more free crit from intellect. This makes it easier as they're running around in gear which is, off the cuff, 588 (well, 580 but yeah you get the point). This means that our FULL BiS is their old BiS.

I'm not disputing that at some iLvl crit will be better than haste, it's just not likely to happen until we're in full BiS, or close to, and as you can see from Voltaa's BiS sheet, this has been factored in.

[Akraen]

Extra ilvls (bearing in mind even after the update they still get more as they got the same update) means more free crit from intellect. This makes it easier as they're running around in gear which is, off the cuff, 588 (well, 580 but yeah you get the point). This means that our FULL BiS is their old BiS.

I'm not disputing that at some iLvl crit will be better than haste, it's just not likely to happen until we're in full BiS, or close to, and as you can see from Voltaa's BiS sheet, this has been factored in.

For frost sometimes what's best isn't what people use, due to even the best human reaction times being insufficient to respond (a few oddball mages prove it wrong, hence my lust for haste). I was curious if that happened to arcane, maybe crit isn't better as you say, but they like it for some reason anyway.

Komma did some translating of frost guides one day in IRC and we seemed to deduce that the Chinese really don't do any theorycraft. They seemed to refer to my guide in a weird way, like they completely misunderstood it and now all the Chinese frost mages are doing things wrong.

[Cycobi]

Eh, there's no real "reaction" based play for Arcane, it's all about knowing mana costs at different stacks under and out of Arcane Power, and with stacks of 2 set. If you know all them it's just forward planning and reacting to procs, but even under GCD cap that only comes every 1s, so it's (relatively) easy to react to.

Crit is very strong for Arcane because of how hard we hit, but I just don't see the value in it until much later on due to the fact we won't "stack" it as we have other pre-requisites to meet first.

[Vog]

Based on my simulations in Excel, crit is still significantly below mastery and haste for single target dps, even in BiS and with the Burning Primal Diamond.

[Cycobi]

Sims in Excel? o_O
The last sim I ran (not in BiS, admittedly) had crit pretty close to both Haste/Mastery.

[Frosted]

Based on my simulations in Excel, crit is still significantly below mastery and haste for single target dps, even in BiS and with the Burning Primal Diamond.

If you have a spreadsheet that runs simulations, you should probably share it. My guess though is that you don't mean simulation, but probably something formulation based. Which would still be interesting to see how you did that.

I guess with Arcane though it wouldn't be terribly difficult to model the rotation (if you assume a kind of static rotation) and find values for stats in a spreadsheet.

[Komma]

This is probably going to be a long post, so let me just post my conclusion first: For single target fights, which reflect a majority of bosses, I don't think crit will outweigh haste, even at very high gear levels (~588). This applies especially if you're running a 4T16 setup, which offers significantly more crit rating than a 2T16 with HWF off-set setup. The following paragraphs are just a justification of why I believe that is the case.


I believe a good starting point is taking a look at simc output. The damage breakdown per spell from simc results are pretty close to what we notice in logs. The biggest inaccuracy is usually in Arcane Barrage damage, which is slightly higher in simc due to poor 4-stack surfing. We can infer from this that the relative ET between the main abilities, ie. AB, AM and LB, are pretty close to reality. This means that there's at least a bit of confidence in the stat weights. Just to be safe though, I will discuss later how some of noted simc inaccuracies affect the results.

Now, if crit is going to outscale at any point, I assume it would be at a BiS 588 2T16 setup (Voltaa's BiS setup) which offers the least crit, and each point of crit rating is offering the largest percentage return. A 250K iteration scaling sim which gave the following:

Crit: 7.25 DPS per rating
Haste: 9.78 DPS per rating

Both mage armor and frost armor gave similar results: simc suggests that haste has a 20-30% margin over crit. Just to be safe I plotted a scaling graph, making sure that this wasn't due to haste breakpoints.

How would the inaccuracies in simc affect this result? Let's look at them one by one. Here's a list of problems that I know with the current release of simc:
1. Cloak damage affected by mastery even though it shouldn't (Fixed in dev build). This bug overvalues crit, since haste no longer affects RPPM procs.
2. Poor modelling of "4 arcane charge surfing". This means that AB/AM damage and ET% is lower than reality. I messed around with Lhiv's TCOM to figure this one out. After accounting for the effect of a 588 PBoI amp, with the BiS 2T16 profile, I found that for AB and AM, the damage gained per haste rating is still higher than crit. This means that if we fix 4-stack surfing, haste would gain more ground over crit.
3. Lack of trinket snapshotting. As far as I know, this doesn't favor either haste or crit.

To conclude, this means that the above simc inaccuracies undervalue haste. If we fixed them, the gap between haste and crit would only be bigger.

When you consider how crit and haste interacts with our rotation, it makes sense. A 100% crit arcane mage plays exactly the same as a 25% crit arcane mage. Haste on the other hand, combined with nether attunement and the 75% increased mana regen from RoP, means that the mage can play more aggressively with arcane charges. Until we reach high enough haste where we get hugely diminished returns (ie. passive GCD cap at 18K), the fact that crit has more expensive rating conversion will always put it behind.

[Vog]

Based on my simulations in Excel, crit is still significantly below mastery and haste for single target dps, even in BiS and with the Burning Primal Diamond.

If you have a spreadsheet that runs simulations, you should probably share it. My guess though is that you don't mean simulation, but probably something formulation based. Which would still be interesting to see how you did that.

I guess with Arcane though it wouldn't be terribly difficult to model the rotation (if you assume a kind of static rotation) and find values for stats in a spreadsheet.

I was gonna hold off on releasing the spreadsheet until it is more polished, but decided to share it now in a separate thread.

[Frosted]

If you're using the most recent SimC release the Arcane mastery / cloak bug fix is included.

[Kver]

Those are good points Cycobi, but I feel that if you are at the level of secondaries that you can reach the haste cap and still have a few k stats left to allocate... there is a lot of damage hidden in there, and it would be nice to figure out what to do with it, I personally have decent faith in crit, but there isn't a lot of math to defend me.

[voltaa]

So this week I figured I would give a crit build a try, reforging all haste into crit leaving me at 2.7k haste and 13k crit.

The results were mixed and some of what I had expected to happen, such as this being a garbage IJ build, turned out to be true, but the big surprise to me was how well this performed on Garrosh. I expected a total failure but ended up doing a decent parse.

Overall it was a bad raid night as a guild so some of the data is absolutely skewed for the worse (protectors got a heal off, my keyboard locked up for the forst 30s of shamans, three thok fixate, guildies were playing like fools) but before the heal on protectors I was sitting at 900k, which is a decently high pull. The two fights that really stood out though were garrosh and sha, where this build seemed to do exceptionally well.

http://www.warcraftlogs.com/reports/vm8 ... amage-done" target="_blank