Blood Pact: On the usefulness of SimulationCraft

Every week, WoW Insider brings you Blood Pact for affliction, demonology, and destruction warlocks. This week, Megan O'Neill will tell you how she became the princess warlock of a continent called Pandaria by switching to saur'lock spec.

You know me: I play a warlock spec that starts with a 'D' only over several dozen copies of my cold, dead body. I'm diehard affliction whether the spec is on top or not. But that doesn't mean I don't ever pay attention to the DPS simulations. I just think the vocal playerbase hedges their bets a little too much on what comes out of the BiS heroic simulations.

Fortunately for me, Ghostcrawler also feels players put too much emphasis on SimulationCraft when it comes to spec balance in PvE. Does this mean SimulationCraft is not valuable? No, it just means that players should be more aware of what's going into a particular simulation and how to interpret whatever data comes out.

SimulationCraft 101

I'm not going to write a 101 SimulationCraft guide, because there are already a few out there that are fantastic enough. SimulationCraft itself has a decent starters guide, but I use the wiki mostly to check syntax for the Textual Configuration Interface (TCI) so I can edit the .simc files myself. My personal favorite is Asakawa's on Wowhead, which was written in Cataclysm, but is still very relevant. More advanced coding tricks can be found in guides like Moshne's SimC 101 thread on TotemSpot, and Shaidar, aka Shaidy Advice on Youtube, has some excellent video guides on not only the basics but also reforging and gemming with SimC.

SimC as a tool

If there is only one thing you read about SimulationCraft (properly abbreviated SimC), it should be the formulation versus simulation article, which can also be seen as the precision versus accuracy battle. The real point is that the question shouldn't be "which is most correct?" but "which is most sufficient?" Am I going to do 198,761 DPS every single time? Probably not. Am I going to be in the general area of 166k-245k DPS? If I match all the inputs for that particular simulation, most likely yes.

The usual problem with citing SimC is that players treat a simulation as a tool to produce one DPS number. A simulation is actually a set of many thousand iterations and therefore actually a spread of DPS numbers, not a single one. The number that you see at the top is the average DPS, but the range (this tier anyway) can be as big as plus-or-minus 17k.

The other problem is the old computer coding GIGO philosophy -- Garbage In, Garbage Out. As Ghostcrawler mentioned in the TotemSpot interview, if your simulation model for a particular class or spec is crap, then you'll likely get crap results. Fortunately, SimC has been the typical warlock theorycrafting software for a while -- much like WrathCalcs for moonkin or ShadowCraft for rogues -- so the warlock part is rather reliable.

But also on a smaller scale, if a player doesn't look at what goes into a particular simulation, he might not be aware of certain liberties or mechanics the simulation could be taking advantage of to produce the output it does.

BiS: Best in Stacking?

At first, I wondered why the bar for affliction was still jutting out over demonology and destruction. With the nerf to affliction's preferred Grimoire talent and the buffs to demonology and destruction, I didn't expect affliction to immediately drop down to third place, but I did expect the bars to even out a little more with each other.

Looking at the Stats collapsible of the affliction T15H sim, I saw the 16222 haste rating and thought, "Oh, get outta TOWN, Gul'dan." But seriously, if you gem and reforge for haste, take advantage of every stacking mechanic possible -- including using the patch 5.1 legendary extra weapon socket and the blacksmithing perk sockets to fit in 3 pure haste gems -- and add in not only heroic gear but a handful of heroic thunderforged gear off the heroic-only boss Ra-den, you can hit 16k haste rating this tier.

Why you would want that much, I'm still not sure on. That's 38% haste unbuffed, by the way, 40% if you're in a raid with a moonkin or shadow priest. A single haste proc would probably tip you over the big haste cap of 50% (~21k rating unbuffed), which reduces your GCD to the minimum 1 second.

If you want to look at it from a haste breakpoint for DoT ticks standpoint, 16k while under a Heroism-like effect (30% haste) will net you nearly double the ticks on every DoT: 23 ticks of Agony, 17 ticks of Corruption, and 13 ticks of Unstable Affliction. This is a relevant point since the simulation spends approximately 40% of the fight under either a 20% (troll racial) or 30% haste effect (Bloodlust, Dark Soul, or the legendary meta gem proc), sometimes multiple effects.

Literally, the simulated warlock spent nearly half the fight hardcapped on haste. I could be wrong, but I think that is a level of haste insanity that the developers neither intended nor expect the majority of the raiding population to attain.

So, y'know, let's not use that number as a "realistic" balancing benchmark purely because it came from a well-known, theoretical DPS simulation.

So what is realistic?

The short answer is "it depends." Realistic is a range depending on anything from gear to fight mechanics. But just for fun, let's play with some of the inputs that are easier to change and understand.

I changed the professions and their perks from blacksmithing and engineering to tailoring and enchanting. Why those two professions? They're two most popular professions in normal-mode T14 conquering warlocks, according to World of Wargraphs. I also changed the race from troll to the more popular human, according to RealmPop-US, though World of Wargraphs says the 2/16H Tier 14 warlocks are tied between human and orc.

Later on, I changed out the heroic mode gear for normal mode gear where I could, and forced the hit rating a little higher to compensate just for not really optimizing the newly normal-mode gear levels of stats. Even later on, I ran some actual warlocks off the Armory through the simulation. Here's what I got for average DPS numbers:

• Recall that the T15H BiS simulation resulted in 198k DPS.

• The same BiS gear and action list but with different professions and race had 188k DPS.

• The normal-mode gear with different professions and race had 158k DPS.

• I looked up one of my many warlock Armory bookmarks, so I'm going to pick on the two halves of @RollWarlock here for a second. Running Cyner and Simply as they are -- averaging 515 and 518 ilvl respectively -- both resulted in 138k DPS.

• Running Simply with the legendary meta gem instead of his normal Burning meta resulted in 144k DPS.

• Running myself at 499 average item level resulted in 112k DPS.

My point is not that the BiS sim is necessarily incorrect. I didn't flip toggles such as boss mode -- they're all Patchwerk sims! -- or player skill (they're all done on Elite, which reacts immediately and makes no mistakes). I didn't change up the action list, which is the priority list of spells that the simulated player will cast. You can feed in different character sheet notes, yet have them all act exactly the same way, and still you may get vastly different DPS numbers.

So, the overall DPS number by itself is not actually that helpful for figuring out if the spec's abilities are tuned correctly. Context is definitely needed in judging someone's ability to rock a 'lock.

What if I judge by raid parses?

Logs are better, in my opinion, since they actually represent what has been done in-game (or on the PTR) rather than a theoretical situation. But raid logs aren't the absolute cure to judging DPS, either.

The major problem with logs is that differing strategies and differing player skillsets can skew the performance. Think of Beth'tilac, where your mages and hunters who had spider corner duty did didly-squat for most of the fight with snaring AoEs, meanwhile you with the multidotting were lighting up every hanging spider and drone you possibly could. Similarly, phasing or distances like those on Gara'jal or Sha of Fear can royally screw up the raid parse numbers purely depending on which player logged it.

Then what are these tools useful for?

I'm only a mild theorycrafter at best, but I suggest that SimulationCraft and raid parses complement each other, but not complete each other. You will still need to bring some spec knowledge and scuba gear for log, table, or code diving to read either reasonably well.

Use SimulationCraft to hypothesize on what you could be doing, not what you should be doing. I run Patchwerk on Elite player skill to get an idea of an upper limit on DPS and then HelterSkelter on Average player skill to get a lower limit.

Use raid parses to gauge if you are doing things as correctly as you can. For example, comparing the uptimes of your DoTs to another warlock or to the simulation will do you more good than trying to match the exact numbers. Raid parses are also great for looking at when certain buffs are going off and if you are utilizing those procs correctly.

In short, proper use of tools like SimulationCraft and raid parses means not ever looking at just one piece of data, like a DPS number. So don't get your rankles up trying to judge yourself and others by a simple meter spam, whether theoretically made or dumped straight out of Recount.

Blood Pact is a weekly column detailing DOTs, demons and all the dastardly deeds done by warlocks. We'll coach you in the fine art of staying alive, help pick the best target for Dark Intent, and steer you through tier 13 set bonuses.