There is a problem with socket bonuses and switching between different saved versions of the same character. As in, the socket bonuses stay in effect. I was testing the 'hit is good, equip the sack' advice in the other thread, and after testing a copy of my druid changed to full ICC gear and playing with orange gems to dive to the armor pen hard cap. When I loaded a version (with a slightly different talent spec) with the sack, socket bonuses on my boots and bracer and such stayed in green text (I assume that means the bonus is active) despite the gemming changing back to straight red armor pen.

I don't know if this was a display bug, or if the bonus was actually counted. Didn't think to check before I fixed it.

Also, the 'other stats' pane that pops up when you mouseover your 'stats' pane in the upper left seems bugged or stuck. It displays set numbers for hit and expertise % regardless of gems or gear changes.

EDIT: Loading a new character also causes it to forget which gem is in the gloves BS socket.

Yeah I also noticed certain stats getting "stuck" in the extra stat pane... I'll try to figure that out. It seems to be using the correct numbers in the simulation even though the display is incorrect.

I was able to reproduce the socket bonus problem with one character but not another... I'm wondering if it is a blacksmith-specific problem. I'll try to figure it out and post it along with the next update.

I've posted a small update that addresses some bugs. The full list of changes is at WoW Simulator Change Log.

The bugs that The Grog encountered should be fixed now.

I'm still working on improving the item estimates, as well as the Moonkin module, which is nearly complete.

It is not that easy. One item vs another is just few points difference, and the variability with default settings is around 60.

I tried 3 runs with 10000 iterations (variability +-6), with different gemming:
full str: 10072
full agi: 10082 (65% crit before mongoose, idol, DV and DBW procs, thus partially past soft crit cap)
full arpen: 10146

The estimations were not very helpful (yet). Maybe it would help if you explained how are the estimations done? If its not secret. As it is now they are wrong for my profile, and testing each item with 10000 iterations is not practical.

Also while i understand why the tool would underestimate agility due to soft crit caps, i dont see why it would do the same with armor penetration as i dont have hard cap yet, and i have no arpen-proccing trinket. Any chance you forgot to update DBW which is not proccing armor penetration anymore? When i had full arpen gemming, it said that strength is the best stat to gem, and it switched back to armor penetration when i regemmed half of the gear.

Yeah I would expect the improvements to estimates to value crit slightly higher, but not sure why it thinks attack power is slightly better than armor penetration in the estimates, whereas the actual simulations are slightly favoring armor penetration. I'll have to play around with it.

The estimation technique is not a secret -- we use a very simple approach: I run a bunch of simulations with various combinations of stats. Then to estimate your DPS with a particular item, I equip the item on your character, and use your current stats to do a linear regression based on the pre-calculated data (or another way of saying it, I use your current stats to do a weighted average of all pre-calculated data points). Then, I apply a set of tweaks for idols, set bonuses, and a few other special things.

It's a 100% statistical method -- we figured hey, we have this simulator and an automation harness for it, why bother coming up with formulas? Just run a bunch of simulations and do some simple math. It turns out that the math is a little *too* simple right now, but I'm working on improvements that should handle soft/hard caps on stats more gracefully.


If you look at the current estimates, and if you look at your own results from actual simulations, the value of strength, armor penetration, and agility is *very* close in both cases... the order is reversed, but the magnitude is in the right ballpark. I want to tweak it a little more to be sure, but I'm pretty comfortable saying that in an actual fight, you won't notice any significant difference if you gem one way or the other at higher gear levels (taking soft and hard caps into account of course).

The tool looks fantastic first of all, great job.

I played around with it very briefly and here are a few observations:
- Berserk use doesn’t seem to be optimised. The simulation I just ran through appears to use Berserk with a low energy pool (50) and 5 CP stacked. At least last time I checked the optimal Berserk use was to amass some energy before using but looking at some sims it seems your threshold is set to store energy to 50 as soon as Berserk is off CD then use it.
- The gem issue already mentioned, it is annoying for the default gems to differ from the optimal ones. If this is too much of an issue to change, a button to “Gem Highest DPS” or something would be great to change the item once selected.
- It seems for JC gems it will default to putting them in items if you don’t have the max. But there is nothing preventing you from putting past the max in, nor any information I could find that said when I did have too many. Again with Dragon’s Eyes it is putting agi in when it isn’t ranked the top gem.
- There seems to be something funny with the calculations for valuing stats. In my tank gear that I’m currently wearing on armoury, but with a DPS spec subbed in, it is valuing strength as higher than agility. I’m only at 60% crit so it shouldn’t be any issue with crit capping, so this is confusing.

The first version of the Moonkin Simulator is now available!

There are a couple things missing still, but I wanted to make it available for people to play around with it:

1. No item estimates for moonkins yet, still working on the new estimator code. Item lists appear in alphabetical order for now.
2. Replenishment has not been implemented.

Note: Treants are implemented as a single treant that hits 3x as hard rather than 3 individual treants, so keep that in mind when looking at the summary. This was purely for simplicity/efficiency, and should have no effect on total damage.


I am not nearly as experienced with moonkins as feral druids, so it is quite possible that there are inefficiencies in the default rotation. (Click on the "AUTO" button under the "ROTATION" section to see the logic it is using.) I feel that the simulator is coming out a bit low for DPS, though quick comparisons to combat logs show the various spells doing about the right amount of damage. We welcome feedback as we continue to play with it and finish it off. By the end of the week we should have all the rough edges polished up.

It would be nice if stats and results of simulations are kept when new simulation is started. Like a "previous runs" panel or something.

edit: I noticed that when scrolling down using mouse wheel in a gear / gem selection, it always scrolled 2 pages down, so i had to scroll up again to get to the page i wanted. Mozilla 3.5.7, WinXP.

That is a very nice interface. A few minor things after a look at the Moonkin:

1) I click on Rotation Auto, then on Rotation Custom, then on "5. Wrath". The icons for Eclipse Lunar and Eclipse Solar are reversed (Lunar should be blue, Solar should be yellow).

2) For looking at some sequences, it would be nice to be able to scroll the log less than a full page.

3) It appears from the log that damage and procs from Wrath and Starfire occur when the cast begins. In the game, NG procs when a cast ends. Eclipse procs (and damage occurs) when the missile arrives (instant for SF, but typically some delay for Wrath).

4) In practice, when Eclipse procs, a Moonkin will cast one additional spell (already queued at the server) before switching to the new nuke.

5) In game, when a target gets hit by a large star from Starfall, they also take the splash damage (small star). The small star can crit, and Edit: this may be wrong: that is another chance to proc Nature's Grace.

6) In the log I'm looking at, there were six Wrath crits which were eligible to proc Lunar Eclipse. They all did. That should be only a 60% proc rate. Maybe I was just lucky.

7) The summary should flag that mana was an issue (perhaps say it dropped below 1k during the fight).

8) Mana regen says 176 (it doesn't say what the units are). Over a 1.27s period where nothing happened (a couple of DoT ticks), regen was 45, so I guess regen is 176 MP5. My Replenishment should be 246 MP5, and Improved BoW should be another 109, my gear (I logged out resto) had another 67 MP5, so regen should be at least 422. It looks like Replenishment was ignored.

9) The ability to save the log to my machine would be nice. (some csv format for loading into a spreadsheet).

Keep up the good work!

Upfront, I haven't been able to look through the entire combat log yet, but I just wanted to say that something about the Moonkin sim feels really off to me. It puts me doing ~4.2k DPS, but I don't think I've ever pulled under 6k in a raid, even with bad RNG and a lot of movement/target switching.

I wish I could tell you what the problem was, but Moonkin is my fourth spec that I rarely use, and I honestly don't know what I'm looking for in the combat logs. Perhaps somebody with a bit more experience has seen a flaw in the sim and could point out exactly what is wrong, but I just wanted to throw out that I'm missing about 2k DPS somewhere.

Other than that, I love the sim. It's a great quick and easy way to gauge performance.

Edit: I'm using everything on default; full buffs, auto rotation, random fight duration, and 10,000 tests.

Look at the "M: number" column in the combat log and scroll down. That is mana. For me, Replenishment wasn't working, and the default rotation does not include Innervate.

- As stated in my previous post, I haven't implemented replenishment yet. I'm working on that... have to figure out how to estimate the effects of it, i.e. how often will it be cast on you and not 10 other people instead.

- I'll fix the icons on eclipse.

- I'll work on the timing of procs and things with the end of the cast. This is the first class that has been implemented that has cast times, so it's a little rough around the edges there. Does anyone have a good average missile travel time for wrath?

- I'll see if I can do something to more closely model the spell queue for timing up eclipse switches. I'll put this at lowest priority though, because it's probably not a huge difference in damage done -- in general I consider it "ok" if the simulator comes out a tad high, as it is a theoretical maximum under perfect conditions.

- I'll implement the small stars from starfall, wasn't sure on that one -- thanks for the info.

- I think you did just get lucky on your wrath crits, I'll double check the proc rate on eclipse though.

- I think you might be looking at the wrong number for mana regen. The popup stat window that says "MANA REGEN" is your Mp5 (including BoW). On the main stat window, the REGEN (CASTING) number is your final regen amount (per 5 sec, just like character sheet) which includes regen from intensity, dreamstate, Mp5, etc.

- Yes a way to export the combat log as text would be nice. I'll put that on the to-do list.


And yeah as I said, the numbers feel a tad low for whatever reason. I'll be comparing it to some combat logs when I get a chance, but if you spot a problem in the meantime, let me know.

I did notice something that doesn't seem to add up correctly. I have a base crit chance of ~40%. This seems accurate with Wrath, as it reports ~40% crit. However, Starfire is coming up ~70% crit when it should be ~80%. And my personal raiding experience shows me that my Wrath usually crits ~41% and Starfire usually crits ~82% (average over about 30 parses).

So, maybe your Eclipse proc is only adding 30% instead of 40%? That's all I can think of, but I'm definitely missing about 10% crit on Starfire in there somewhere.

When I cleared all my talents, it showed 28% crit. When I restored all my talents, it showed 28% crit (ok).

Summary display shows Wrath crit at 28.4%. I should be getting 4% from Nature's Majesty. In a real raid I should be getting 3% from Improved Faerie Fire, but I couldn't see how to add Faerie Fire to the rotation. So that is 7% missing crit on Wrath.

During Solar Eclipse, the first several Wraths are not showing the damage bonus:

Solar gained at 25
Next five Wraths do normal damage
Next five Wraths (timestamps 34 through 39) show 40% increased damage.

Solar gained at 1:09
Next seven Wraths do normal damage
Next three (1:20-1:24) show 40% increased damage.

Starfire is harder to analyze, but 50% overall crit (summary display), seems much too low.

Edit: Crit rates:

This was my resto gear, so I had about 9% miss. 32% on my non-misses would have been about 29% of all attempted Wraths. So maybe you're getting Nature's Majesty, and just missing IFF.

Note that spells do NOT suffer the crit% reduction against bosses like physical DPS do.

Good eye Erdluf, it seems that I am unable to keep "lunar" and "solar" straight! A few numbers were swapped between the two eclipse buffs, and it was causing damage values to not get updated immediately when the buff was gained. I'll fix that and post an update sometime tonight hopefully. That should resolve the majority of the discrepancies.

Nature's majesty and imp. FF are correct. I'll probably redo imp. FF though, since it affects every spell you can cast. I'll roll it into your overall spell crit %, and thus it'll be added into the crit % shown in the stats pane.

In regards to crit suppression... I was under the impression that spells do indeed suffer from some amount of crit suppression against a level 83 boss. I have been using the number 3%, though I can't seem to find the source where I got that, so I'm doubtful now...

This thread discusses it to some degree, but doesn't have a definitive number... all it seems to conclude is that "some kind of crit suppression is happening for spells": Crit Suppression discussion

Numbers in that thread say it ranges from 2-4% depending on your hit rating perhaps. Another post in the thread says 2.5% sounds about right. Anyway, not enough data to come to a definitive number...


Another piece of information that may be too circumstantial, but I'll throw it out there: a DK's gargoyle can crit with its attack, but it has a very low crit rate, around 5% from all the combat logs we have checked. People say that the gargoyle has no base crit, and have observed that it does not receive buffs like moonkin aura. So the theory is that it gets crit only from debuffs on the target: winter's chill, and totem of wrath. That totals 8%, but the gargoyle crit rate is consistently lower than that, which indicates a 2-3% crit reduction coming from somewhere... I assumed it to be this "crit suppression".

I posted a quick update with bug fixes:

- Eclipse procs should be working correctly now
- imp. FF crit bonus shows up added into your spell crit on the stat pane now
- Starfall AoE explosions added


The main piece that will take me a little longer to work out is the timing of procs on spells with cast times and travel times and the such. I don't anticipate that will make a significant impact on DPS, but I'd like to get it right regardless.

Interesting... seems pretty inconclusive for spells, but the testing around physical crit suppression is pretty good. I guess I'll remove the spell crit suppression for the time being.

Looking at numbers from team robot: Full raid buffs, no tier gear
SP 3333.5, typical talents, only 2/3 IIS.

Wrath: 3954, or 5535 with Eclipse.
I think you are using 79 damage range: 553-623. At 80, untalented tooltip is 556 to 626.
I think you are using Eclipse as a straight 1.4 multiplier. Other TC tools (simc, wrathcalcs) treat Moonfury, Eclipse, 4t9, and 2t8 as being additive. I'd do that unless I had testing to show otherwise.

SF 7022.
I think you are using 78 damage range 1028 to 1212. At 80, untalented tooltip is 1038 to 1222.

I posted another update. Of note for moonkins:

- damage multipliers for wrath should be correct now

- spells now properly execute at the end of the cast time, not the beginning

- removed spell crit suppression for now, since there isn't enough data to support it


I still didn't model missile travel time for wrath.

With the latest changes, damage went up quite a bit... and wrath is doing a larger percentage of damage than before. I'll do some double-checking of the wrath damage formula, and the new timing of spells with the changes to cast times, but nothing immediately jumped out at me as being incorrect.

I'll try to look at this more soon, it looks neat (looks like Erdluf is making a good start at picking at theorycraft issues). Who's writing the Moonkin model?

Looking at a moonkin summary, it'd be nice to have a few "damage done for one cast" views for things like Starfall and Treants, since in terms of decision making we consider them as a 1GCD cast with 10/30s long effects.

This UI is absolute pleasure to work with.

Flask of the Frost Wyrm with Mixology is 172 spellpower, not 162 (how do you change professions?).

Is the Int stat in the mouseover not including Furor? Looks like total mana is low for the same reason.

Can't figure out what's supposed to be included in the "spell crit" stat. In currently reads 37.6% in my Armory gear (saved on your website). Correct crit chance that's not spell-specific (i.e. including IFF and debuffs on the boss, but not Nature's Majesty or IIS) is 45.13%. I can't figure out a combination of buffs and Furor that accounts for that.

I believe the current default rotation accurately summarizes the standard rotation uses in WrathCalcs (except, as Erdluf said, the fact that there's an unavoidable delay is responding to the Eclipse proc). If this all goes well though, one of its main uses would be to explore much more detailed rules for DoT refreshing.

------

I loaded identical setups in Robot and WrathCalcs. To facilitate comparison, I made the following changes from my Armory setup:
--Disabled Nibelung proc (not modeled in Robot)
--Replaces Lightweave with 23 haste (elimating procs that WC averages in)
--Disabled Bloodlust (same)
--Disabled Demonic Pact (not modeled in Robot)

Spreadsheet attached. DPS outputs are 7765 (WC) and 8510 (Robot). Rather large difference.

Comparing individual spells (for this part I disabled level-based partial resists in WC, since it averages them in, and compared to unresisted hits in Robot). There's no randomness here, since both WC and Robot use mean spell damage values.
--Starfire crit. WC: 17603, Robot: 17669
--Starfire non-crit. WC: 8423, Robot: 8454
--Insect Swarm tick. WC: 1644, Robot: 1650

A quick check here reveals that WC and Robot are differing by around 18.4 total spellpower after all buffs are applied, and that exactly accounts for the difference in spell values. So you're probably computing those correctly. I'll fudge the spellpower difference into the WC computation for now (it obviously doesn't nearly account for the 800 DPS).

--Wrath crit with Eclipse and IS up: WC: 13457.3, Robot: 13453. Sounds like a tiny difference, but since the two models lined up to within 1 point of damage on all other spells, once I matched the spellpower totals, maybe worth looking into.

Ok, found the difference there--WC used 557 to 627 as the Wrath damage range, and you probably used 526 to 626 that Erdluf posted above.

With those two changes in WC, we line up literally within a single point of damage on every spell.

--------------

I know this is a bit disorganized since I just typed while testing. Summary for now:
1) You seem to have the basics correct as far as spell damage computation (talent/bonus stacking) and total haste.
2) The spellpower difference is something we should look into, since 18.4 (which will become 28.4 after fixing Flask of the Frost Wyrm) is greater than the rounding error that either model should have.
3) I can't tell whether you're computing crit rates correctly, since I'm not sure what talents/buffs are intended to be included in the "total crit" display on the front page. The simulation results might suggest that you're omitting Improved Faerie Fire, and also Improved Moonfire as applied to 2T9 DoT crits.
4) Your overall DPS result is still rather high. Once we're sure all the basic computations are right, we'll have to start looking into that.

Overall, I find this very promising. I could honestly it see it being an extremely useful complement to WrathCalcs for Moonkin theorycraft work.