I'm expecting to see this kind of jump in the relative haste value at 12.5% or 15%, but not at 14%? Anyone have any idea what's going on? I suspect that this is still representative of the extra Combustion tick, but I can't work out where my "hidden" 1% haste is coming from (Lifeblood?). Note that in the example shown, I've only selected self-buffs, auto-armor is off, molten on.

Also, I noticed that Lifeblood (or mirror images) never shows up in the sequence reconstruction.

This is more of a feature request, but maybe some time in the distant future when you feel bored Kavan, a tally for "reforging cost" could be added to the optimizer, a nice but not necessary feature.

Lifeblood and Heart of Ignacious are giving 2085 haste on a 2 minute cd. Added to your passive haste puts you at 3463. Tyrian's fire mage post gives the a step value for haste at 3476 (extra tick on LB/pyro). That should cause a step gain for adding 10 haste rather than a step loss for removing 10, but it's probably the right idea. Since I see much bigger steps in dps when I push my character across the 15% threshold, it might only be adding a tick to lusted combustion.

Regarding the strange Arcane results. I've had that happen on a few rare occasions but I haven't been able to determine so far if it's fixable or not. It most definitely is due to the complexity of quadratic solver, but it must be combined with some strange edge conditions to generate those results. If you can provide the exact character xml where you had that happen that would be great if it happens more or less consistently. So far I've only had it happen intermittently so I wasn't able to completely track it down.

Some suggestions regarding what to do if that happens, or what I've observed to fix the odd behavior. First try Arcane Light setting in advanced options. Even though it considers less options (no possible rampups), I've observed it produce better results in several cases compared to full model with all cycles.

By nature of the quadratic solver it cannot guarantee to generate globally optimal solutions. There is an advanced option for global restarts which attempts to inject random perturbations in order to have higher chance of reaching global optimum. I wouldn't have this on by default, but it can help in understanding how much volatility there is. Just note that the perturbations are by nature stochastic so you'll no longer be guaranteed to get the same results on exactly the same gear on repeated runs.

Even so this fluctuation is present anyway on gear sets that are very similar. Incremental optimizations are designed to minimize this as much as possible, but it is definitely not eliminated completely. If you look at the stats graph on arcane profile you'll usually see some spikes which are indication of this fluctuation as result of solver not finding globally optimal solution always. If you disable incremental optimization you'll see a lot more spikes usually.

I'm not sure if what I described above is causing the problems you're having. I think it's more probable that those are caused by some bad constraint or numerical instability in some edge condition, mainly because the fluctuations due to nonoptimality are usually on the order of up to 10 dps only and the screenshot with gems has differences of over 100 dps. Also when you get those strange severl k dps increases in optimizer they are usually not confirmed by the final evaluation which would indicate that there is some problem in combination with incremental optimization.

I believe there was some talk of getting an extra DoT tick of Combustion, probably when you are under Heroism or Bloodlust. I know mine did the same thing this weekend when I picked up a couple of new pieces of gear.

Assuming you are fire, the haste number that rawr wants is the raw haste you get from items(516 for first soft-cap) so put in something like haste>=516 and then haste<=520 and you'll get something right near the 12.5% haste mark that you want. Do know that various theorycrafting has suggested that crit is not as good as we thought due to ignite munching and it is probably beneficial to at least to get to 15%(812) for the extra combustion tick.

Interesting thought, FYI though the haste buff from Lifeblood and the Heart do not stack.

Dark Intent is currently not available in Rawr, but I'll try to add support for it. In terms of the dot amplifier part how would you prefer that is supported? Should it be a parameter that lets you enter average number of stacks you get or average dot crits per second of the warlock similar to how Focus Magic parameter works?

Ok I've seen this mentioned in a number of places so I think it's time to make some statement about it. If you do some changes to your gear because Rawr suggests it and you don't see the expected results there can be many reasons for this. The first conclusion most make, and it definitely is one possibility, is that Rawr is wrong. But there are other options to consider such as rng, buffs and when you introduce real raid events it becomes even more tricky.

So how do you determine what is really going on. Tools like Rawr and simulationcraft operate on averages. Rawr uses theoretical averages based on mathematical models while simulationcraft gives you averages based on thousands of runs. Now you make one small change in gear and go do a couple of runs with it. If you look at simulationcraft damage spread for Fire you'll see that there's a 15% deviation in either direction from average. When you make a change that the tools say will result in 100 dps increase on average when the actual value can vary up to 8k between repeated attempts with no change you have to be very careful what you're attributing that to.

I hope the above was a strong enough warning to take rng very seriously into consideration. Now let's assume that we're accounting for rng and we still think there is something wrong. What can you do to help figure out why that is the case and help us improve our tools in making better predictions. The best thing you can do is to save combat logs and look at detailed breakdowns of damage by spell. Both Rawr and simulationcraft give you their breakdown by spells as well and comparing against this will give you the best indication of where the source of the problem is.

Check things like crit percent of individual spells, check contribution to total damage of individual spells, check damage ranges of spells, check dot uptimes and anything else that might be relevant. Now do this before and after the change to gear and compare it. What kind of changes do you see in practice vs what is predicted by the tool? In general this will be hard to distill because of the above mentioned rng, but this is the best thing you can do in pinpointing the problem.

In some cases if you're comparing results on actual bosses the reason for difference might be something like aoe or other raid mechanic that is not modeled. In that situation there's not much you can do other than to help in developing better models that support those situations. In other cases you might find some insight that will improve our understanding of existing mechanics and will allow us to create better models such as for example model for average Combustion damage. In general the more controlled environment you have, the less procs and other events, the easier it will be to validate results and make improvements. As we become more confident on the validity of models in simple scenarios we can move further on to more complex situations.

Ideally in the end we'll be able to make accurate predictions that will help in gearing decisions for everyone. Noone is out there to intentionally deceive you, what you're getting is the best result we have given the current understanding of mechanics.

I noticed some odd Rawr behavior this morning and wondered if someone could explain it to me. I use the same basic 3/35/3 fire talent spec that is listed in Rawr, with the exception of 2 points into Cauterize instead of Improved Scorch since mana is no longer an issue. However, Rawr is showing the spec with Imp Scorch as 2k dps higher than the same spec with Cauterize instead.

How does reducing the mana cost of scorch increase dps, outside of maybe not evocating once? I wouldnt think a single evocate would equate to 2k dps.

I'm running into a problem with gem selection. I run two separate character saves: one with every item item I own and most enchants and gems selected, and a slimmer one with only my "best" items and the "best" enchants/gems selected. I do this for comparison purposes so that I can get better results with my slimmer save, but bigger picture results with my bigger save.

For some reason I was getting much better results on my slimmer save (a 600+ dps difference), so I was trying to figure out why, and it turns out that it was equipping a gem that was way far down the list on the gems pane. On my big save, it was so far down the list that I hadn't bothered making it available. But on my slim save it was made available a long time ago and was so far down now that I didn't unselect it, so I had lucked out in having it selected.

http://i.imgur.com/Z7yqY.gif

My question is, why is it listing this gem so low on dps, yet optimizer is equipping it in 4 out of 9 of my normal gem slots? How can I avoid this problem in the future without basically selecting every gem that's available?

Edit: Looks like the problem was due to an old version.