It's a leftover from previous talent trees; just a tooltip error. It does function properly, and will be fixed soon anyway.

Think it's just using an outdated cookie cutter template. Not a tough job to manually shift some points around though

Most would probably prefer to throw those extra points into arcane stability instead, along with a point from arcane subtlety - but it's really a matter of personal preference where those extra points should go. So I guess someone just picked something random, and implemented the template, without putting too much thought into it.

With the latest beta being released I figured it's a good time to give a short introduction to the new features in the advanced solver. As an example I'll show how you can get sequence reconstruction on a sample 5 min fight with Arcane build.

Start with your favorite character and set up all the fight parameters as usual. After you're done select Sequence Reconstruction chart. Most likely you'll either have nothing or something completely wrong.

Go to Advanced tab and make the following changes. Enable global optimizations, enable reconstruct sequence, probably leave hasted evocation disabled until you get some experience without it. Change SMP search method to depth first, change computation limit to 3000. In the display column select all integral mana consumables, segment cooldowns and advanced constraints. Check debug segmentation and check variable segment duration.

If you've tried advanced solver before you're probably familiar with random hangs and waiting indefinitely between making changes to the options. You no longer need to fear that, you can make as many changes as you want and everything will work. One thing you'll notice is that reconstruction chart says "Sequence reconstruction data is not available" and if you switch to Stats tab you'll see everything under Solution shows .... This means that advanced solver is still working.

Click on the Advanced Solver Log button. You'll see a log and a button for Cancel Calculation. This is especially useful if you set up a very high computation limit (I often set it to something like 10000000 when I want to solve until optimum is proven or I cancel it, make sure not to set it anywhere that high if you're not using depth first!). Whenever the solver finds a new solution it will write a line like "Probe value = ...". This will tell you the value of the solution it currently found (a lower bound) and round when it found it. If round exceeds computation limit the solver will terminate as if you pressed cancel calculation.

Another thing in log is the backtrack messages. You can use these to help you get an idea how long the solver will need until it's done. A good heuristic would be if you get backtrack at N in round R that it'll take a total of the order of R*2^N rounds.

By now the solver should reach the limit and chart with best found solution so far should be shown. One thing you should know is that quality of solutions very much depends on segmentation. If you see that sequence is not executable (i.e. running oom, score < 100) that is a good indicator that you'll need to fine-tune the segments. Most likely you'll see that in the chart one evocation is missing. Here it's important that you enabled debug segmentation so that when you check spell cycles on main tab you can see what is going on. The advanced constraints work best when evocation must be placed on segment boundaries. If you see that the evocation that wasn't placed in the chart is in a relatively large segment then you'll have to split that segment manually. A good first attempt until you get a feeling for it is to look at sequence reconstruction and take a note of the timestamp when evocation happens. For example if evocation is from 3:00 - 3:05 then it might be a good idea to put a manual segment split at 183. You do this on the advanced tab under Additional Segment Splits. You might have to do this a few times as the solver tries to "outsmart" you by wiggling the evocation position around. You can also try to choose one of the fixed segmentations, but take into account that more segments you have the more calculations need to be performed.

If you're still looking to find a good lower bound and you see that it'll take you ages to complete the solution you can start giving the solver some hints about cooldown placement. For example you might want to start with a use cooldowns right at the start hint.

To add such hints click on Edit Custom Cooldown Restrictions. You have instructions on how to use it in the editor itself but let's just give an example for our case. We'll say to put AP, IV and Heroism right at the start, but not for mana gem effect (in most optimal cases you should wait until you use some mana before you pop it). This would looks something like this:

0-15: Arcane Power
0-20: Icy Veins
0-40: Heroism

You can give further hints like use AP as soon as cooldown is ready. It's just as helpful to give hints for where some cooldown should not be present or where it should be.

Let's say with enough hints and proper segmentation you've managed to find a good lower bound. Now you want to know if this is for sure the best way to stack cooldowns. First before you lose the solution make sure to use copy to clipboard and store the solution somewhere. Take note of the total damage it got. You know that anything lower than that is not optimal. To save solver from unneeded work you can set Lower Bound Hint to something just a bit smaller than that value. You'll also have to remove all hints from cooldown restriction editor because you can't know if those are necessarily optimal. Then just set computation limit to some very high number and let it run over night. If the solver completes (before reaching the limit) and it didn't find anything better then you know you have the true optimum.

Wrong revision. Please delete.

Unfortunately frost replenishment is still not ideal on the PTR in that each frostbolt delays the next replenishment tick by 1 second. You typically only get about 60-70% of the possible ticks because of this.

Here's an example:

The time between replenishment ticks ranges from 1.156 to 2.119 seconds. The longest time is probably due to erratic latency (the PTR is overseas from where I am, as far as I know, so the network latency is about 5-6 times higher than normal for me).

That's 10 ticks in roughly 16 seconds = 62.5% efficiency.

Compare that to a Retribution paladin on live (from WWS):

That's 17 ticks in roughly 16.1 seconds. Not 100% regular either, but one tick per second on average.

That is probably not going to happen. There are a lot of combinations and making an automated naming mechanism is too big of an effort for the gains. It takes some time to get used to the notation and when you're trying to convert a control string to cast sequence for the first time it seems very hard (I know it did for me), but once you do it a few times it gets a bit easier. The closest thing to what you want is to create a manually built library of control string => cycle name mappings. I won't be doing it, but if others think it would be needed I can offer help/suggestions.

Kavan, in v2.2.0b5, Lightweave Ebroidery appears to still be giving its 3.0.8 buff, instead of its 3.1 buff, even with 3.1 Mode enabled.

Also, how is Demonic Pact being calculated? What spell power is the 10% being based off of? A warlock with more than 2800 spell power will outscale Totem of Wrath, for instance.

Could you point me at some documentation as to how to do this? I still can't figure out how I'm supposed to interpret the string of digits I get in each possible rotation, which makes this feature useless for me.

Try the latest beta, the description of what each digit means is on the results message box.

I see the tooltip for molten armor in the latest beta rawr states the spi changes but I don't seem to notice the crit % increasing as I add spi gear or buffs.

Is this functioning correctly or is it user error?

The Automatic Armor switch was the key, thank you Enthorn.