I was looking at Rawr's By Spell breakdown for Frost and I noticed that, although I have close to 30% crit buffed, Deep Freeze, Frostfire Bolt and Ice Lance all show more hits than crits. Since these spells should usually only be cast with Fingers of Frost, the crit rate should be close to 90%. Am I missing or misreading something here or is this a bug?

Rawr4 Repository: US.Proudmoore.Mangará.Mage

Has anyone else confirmed this? To date the 359 item is still set to the 75 sec CD and the 372 to 100 sec in the rawr item cache as of this morning.

I am just curious as to the best way to set up my rawr for my mage. I have been using rawr now for about a month, and love it, but as I have been tinkering with it a bit I see that there is a large amount of options that I am not using. I am wondering if there are things that I should be changing in order to get the most accurate results? As it is right now, I do not change much from the normal settings when I load my mage from the rawr add on. about the only thing I actually change is my gemming so that I can use JC gems, and my enchant to my rings because I am an enchanter as well.

Also I have noticed that there are times that I have used the optimizer to optimize my mage for DPS and it has suggested to reforge, gem and change gear that actually lowers my DPS and my damage done in the calculations. Is that normal? I was not able to find a good reason why it would have asked me to do that by looking at the information provided in the calculations.

Thanks in advace for the help.

If you're just looking at general optimizations you don't usually need to modify settings a lot. Most of the settings are there to either allow you to specify buffs, specific fight conditions or to help in answering theorycrafting questions. For example you could use it to find out how the stat values change in different conditions such as double dotting for example. Some options are there for example to allow you not just to figure out what gear to use to optimize performance, but also to find out how/when to stack cooldowns and similar.

Regarding the optimizer giving lower result, that is normal to some extent. It is not actually suggesting to equip it, it just shows you the best result it found in that run. Since the optimizer is probabilistic at its core it can't guarantee that it'll always find the optimum. I think it's better to show what it finds in that case than just dismiss it because it can help in some cases to understand why it might be having troubles in finding optimum.

You can see which armor is chosen in the solution section under Spell Cycles.

Armor could be added as part of the casting state, but it needs more information than that. Otherwise it would ignore the fact that it takes a gcd to switch armor. I think the easiest solution would be to model a single swap situation. A generic multiple swap support optimization could be added, but it would come at a high performance cost, similar to some of the advanced constraints in advanced solver.

I noticed that to get proper optimization results (for my frost mage) you need to do the following things:

1. select advanced options and there all the necessary stuff (sequence reconstruction, segment cooldowns/mana, variable segments) for "display"
2. check the reconstructed sequence for obvious errors (like dps at 0 mana)
3. play around with advanced constraints and the other options there, until the reconstructed sequence makes sense. I found that more often than not I have to select everything and increase advanced constraints to max, which, unfortunately, makes rawr very slow.
4. before optimizing, select the same options on the "for comparison" side.
5. optimize (that takes a LONG time now....) and see what you get from it

Unfortunately, this often leads top optimization times in the 1-2 hour range and often leads to Rawr hanging (I guess out of memory).

What I also found: If you let Rawr optimize your glyphs along with your gear Rawr always hangs. I guess the parameter space is just too large then.

Can you elaborate a bit if/how much the optimization results using full advanced constraints differ from not using advanced solver? Even though the reconstruction is usually not good without advanced constraints it should usually not affect optimization and comparisons significantly in quality.

When you're talking about stat values are you basing that on the relative stat values graph or the mage specific stats graph? In general the stats graph will give a more accurate picture. Another thing to try if you see negative dps values is to try disabling incremental optimizations. If that fails that's a problem and I'll need to see the character xml where that happens to investigate further.

Reconstruction running oom has no relation with the actual solution not having enough mana. If you sum the mps figures in solution spell cycles multiplied by time spent using each you should see that there is actually enough mana. It's just reconstruction that is having some problems in rearranging all the pieces in a linear sequence that fits all requirements. Now most often there will be some things wrong with the solution, but they're usually related to cooldown stacking. But it shouldn't happen that it's miscalculating mana use.

The main point to take away from this is that the sequence reconstruction is a side result that is computed after the solver is already finished and has no bearing on the dps. The advanced solver should give a more accurate result at performance cost, but ideally the results should not be so different as you're experiencing.

I tried again with my current gear both ways and the difference is negligible. What I remember is, it was just on the breakpoint between molten and mage armor, i.e. sequence reconstruction would show it running oom with molten armor, yet Rawr still insisted on using it. Now it is properly deciding to use mage armor.
Thank you also for the clarification concerning sequence reconstruction.

There is still one issue concerning the stat values. I used Rawr to optimize and then load that itemset into chardev and run SimC with it. I forced SimC to use mage armor and also removed the prepotion.

From Rawr (stats graph) I get the following values:
Int 2.88
Haste 1.31
Crit 1.18
Mastery 0.67

dps: 20798

While SimC (10k iterations) gives:
Int 3.05
Haste 1.26
Crit 1.00
Mastery 1.16

dps: 22146

I can understand that the absolute numbers may be off, since both programs follow different routes, but I wonder why the relative values should be so different.

Edit: SimC without water elemental
Int 2.67
Haste: 1.14
Mastery: 0.96
Crit: 0.96

That does look like something to investigate. I'd first check by manually increasing mastery by 100 on one piece of gear and compute the dps in both to see if there might be a problem with the way relative stats are computed. If that looks ok the next step would be at looking in how the spell breakdowns change when you increase the stat.

I think this mainly comes from pet freeze which is not modeled in Rawr currently. I'll see if I can come up with a reasonable model for it.