Random number generator woes
In World of Warcraft, the random number generator owns your soul. If you mouse over any ability you have on your hotbar, you'll see that every damaging spell or heal has a minimum value and a maximum value. For a level 85 Starfire, the minimum value is 987 and the maximum potential is 1,229. This means that before any other multipliers on the ability are added (spellpower, critical rating, mastery, boss debuffs, etc.), the game makes an initial roll for the base damage of the attack between those two numbers. If a balance druid were to cast two Starfires in succession with identical conditions for both casts, it's almost doubtless that the two casts would hit for different amounts.
To add some more RNG to your RNG, we move from base spell value to critical strike chance. It's a fairly common misconception that a 25% spell critical strike rating means that you will critically hit with your abilities 25% of the time. What 25% spell critical strike rating really means is that every single time you cast, you have a 25% chance on that spell cast to critically hit. This means that you could have a 10-minute parse in which you actually crit 30% of the time or more because you get extremely lucky with your crit rolls.
On top of all of this RNG, add in random procs to the mixture. Affliction warlocks might end up with a lucky string of Nightfall procs, while a fire mage might go for 30+ seconds without a Hot Streak. The extremely lucky survival hunter might see a Lock and Load proc off every Black Arrow one fight and then see LnL proc once a minute the next. A shadow priest might see enough Shadow Orb procs to maintain Empowered Shadow 100% of the time, or might see it fall off when Orbs don't proc for 30 seconds straight.
RNG happens, and RNG is what makes it so that any form of testing on target dummies is inconclusive at best. If you're a frost death knight who wants to test his new plate chestpiece of awesomesauce and you run two 10-minute dummy tests and see a 500 DPS decrease using the new plate chestpiece, that doesn't mean your new cuirass is to blame. You could have seen fewer Rime
procs wearing the new chestpiece. You could have critically hit less than your first attempt. You could have rolled poorly for your damage values. All this stuff combines to make it so that even though the new chestpiece might be a DPS increase overall, the attempt you tested with it saw it fail.
Simulators are set up to try and combat RNG by running long simulation times. EnhSim
, for example, simulates between 1,000 to 5,000 hours of 7-minute fights in order to bleed out the possibility of RNG ruining your DPS values. While it's very possible that you'll see an inflated number of critical hits during your 5-minute dummy test, 5,000 hours of simmed testing will see it even out to be very close to your actual critical strike chance.
DPS tests as a scientific experiment
RNG isn't the only argument in favor of simming your DPS. If you guys can remember 9th grade science with me, think back to your learning about how to construct an experiment. Your experiment in this case is a DPS test. Your dependent variable
is your ending DPS. Your independent variable is what you're changing from one fight to the next; in most cases, this will be a gear upgrade. Your control variables are everything that remains constant from fight to fight.
In your standard dummy DPS test, the problem becomes that you have no actual control over controlled variables. You don't have any way of telling the rude demonology warlock who just started testing on the same dummy as you that his 8% spell damage debuff is influencing your DPS test. You don't have any way of knowing that at 8:54 seconds into your test, you'll accidently misclick something other than Envenom
and your Slice and Dice
falls off. People make mistakes, debuffs get applied, things happen that will cause fight A to be different from fight B.
This is where simulators and spreadsheets come in. When you spreadsheet a gear choice, the only thing that changes within that simulation is the single gear choice. For both test A and test B, your buffs and debuffs remain constant. Your priority system remains constant. Every other piece of gear on you remains constant. You control every variable of the fight so that the only thing that changes is the one thing you want to test, allowing you to see accurately what the benefit of that one change is.
This isn't restricted to gear choices, either. If you want to see the benefit 10% melee haste is for your fury warrior, you just need to change that variable while keeping everything else the same. If you want to know whether your survival hunter should drop Explosive Shot
in favor of Arcane Shot
, you change your priority on Female Dwarf
to reflect only that change. You use spreadsheets to control your control variables so that the only change between two simulations is in the independent variable, and the difference between results in your two fights is your DPS increase or decrease.
My charge to you
I'm not writing this out of any desire to shame people who aren't currently using the simulators and spreadsheets available to their class. I merely want to get the word out that there are legitimate mathematical reasons as to why spreadsheets are so heavily endorsed by the theorycrafting community. These are programs and applications written and designed by some very awesome individuals to help the World of Warcraft
community out, and they're definitely worth checking out!
World of Warcraft: Cataclysm has destroyed Azeroth as we know it; nothing is the same! In WoW Insider's Guide to Cataclysm, you can find out everything you need to know about WoW's third expansion, from leveling up a new goblin or worgen to breaking news and strategies on endgame play.