SimulationCraft: What does it all mean??

It's been a pretty busy week on my blog, the busiest one traffic wise since I started spewing my ramblings onto paper (well electronic paper). My post on SimulationCraft seems to have struck a few chords so in an attempt to aimlessly chase ratings expand on my success I thought I'd have a stab at explaining what the actual results mean in plain English.

So you've read the idiots guide, you've ballsed up at least 5 attempts at simulating your characters DPS, you've gone back and read the guide again, read the official help files a bit, scratched your head a bit, and finally got some figures that look like what you've been reading on the interweb. Brilliant! You've got a page of completely incomprehensible maths, statistics and pretty graphs that'd look more at home on a PowerPoint at that next dull management meeting you were trying to avoid…

There are some figures that (may) make sense to you instantly; there are others that are best when compared to an average, or a target of where you want to be. If you're comparing something you need a baseline, for this example that's you! (or me as the case may be). For the record, I've run 5 different simulations; one with the ultimate tier 11, 25 man gear, with all the raid buffs and trimmings; one with the equivalent 10 man gear and the same buffs; one with my crappy heroic gear with all the same buffs and trimmings; one with (a slightly unrealistic) all the buffs and trimmings but with the target_level override set to 87 (i.e. a heroic boss), you'd NEVER get this amount of buffs in a 5 man as you simply wouldn't have the classes there required (you're pushing it for a 10 man); and then finally, a far more realistic 5 man scenario with just plain old Fort, Int, and MOTW (a reasonable assembly of buffs in my view), oh and replenishment as you'll be keeping that up at all times (WONT YOU?!?) as it's one of your main sources of DPS. The baseline scenario will be the final, and probably worst DPS(it's chugging away as we speak, so we'll see if my prediction shows me up to be the eejit you suspect me to be).

I'm setting the fight length to 350 seconds (nearly 6 minutes), 2 adds, Patchwerk style fight, and assuming you're the best of the best when it comes to mashing keys and given you (well me) an elite skill level – the trick is to only change one variable (or set of variables) at a time so the comparison has some meaning. As homework, I'll leave you to run your own simulation, with your gear, with a realistic skill level and fight scenario, I can't stress highly enough that stat weightings will vary massively depending on your ability (or lack of it), your gear, the number of adds, the type of fight, and a whole host of other variables, what is certain is the figures are a model (a model, by definition is an approximation, and only an approximation) of reality. Go read about statistical modelling in a text book or here on Wikipedia or somewhere else where stats geeks lurk.

Lecture over, and here we are, remember I'm using me as the baseline, the links to the full reports are at the end, I'll reference sections only, for comparison in the main text to keep my ramblings to sub War and Peace word count.

First of all, and the thing you're almost all certainally here for, the scale factors, or stat weighting s as they're sometimes known. You'll notice two values, a scale factor and a normalised value; the latter is the one you want, it simply skews everything so that it's a direct comparison to your best stat (Int) to make it easy(er) to compare one stat to another when you come to chosing your gear. I'll be posting some specific analysis on stat weights in my next blog so if you're still confused check back and I'll endeavour to confuse you more.

Int will always be best, if haste comes in at 0.5 and Crit at 0.25 it means that Int is twice as good as Haste, and Haste is twice as good as Crit (and thus Int is 4 times better than Crit). Put simply, if given those weights, you have a simple choice of a tunic with 100 Int and 150 Haste, you'd always chose the first one (but it's never that simple is it?).

Next you're at the charts, the damage per execute is a telling chart, if interpreted correctly; it's basically saying for the amount of time you spend casting these are the spells which do the most damage – i.e. devouring plague is an instant cast spell so does the most damage per cast.

Next is the damage source pie chart, you'll notice that mind flay is near the top in all, which seems counter intuitive given the last statement. But think about it for a second, mind flay is your filler when all of the other spells are on cooldown or waiting for procs, you're casting it for the most time, so it does lots of DPS, but bang for buck its lower than the others.

The other thing to note in these early graphs is the mana timeline graph, if you're going oom regularly, you're doing something wrong (or something right, i.e. emergency healing which is saving the wipe), I've heard a few Shadow Priests complain about mana issues, this should never be the case with any level of gear.

The most important graph in my opinion is the DPS scaling at the end, this basically says if you add 100(for the sake of argument, you could use any figure) of any given stat, what is the expected DPS increase? You'll notice for me Haste Crit and Mastery all look pretty much the same, crossing once or twice; for the best spec there's a far bigger difference as you follow the graph up with Haste becoming the worst and then the best and then the worst etc. as you go up. This is because of a whole heap of internal metrics (some of which we know about or can approximate, some which Blizzard keep close to their chests).

Notice on the t11 graph, Hit and Spirit have a scale factor up to 0 (i.e. if your gear was worse) and then abruptly stop giving a dps boost – this is the hit cap. Also notice for me, it's pretty much horizontal (at target_level 87) which means I'm over my hit cap for heroics (naughty me, I'll be remedying this immediately).You'll also notice, if you look at the 3rd scenario that my dps drops, even though I've got massive buffs, this is purely because of the hit cap dynamic, and is a good reminder to follow the basic principles that have been with us for all (well almost if you forget the last 3 months) of WOW - get hit capped first!

That's it for the basics, there's loads (and loads and loads) more, have a read, if there's anything you want me to explain (or make up on the spot) leave a comment, or email, tweet, or send a carrier pigeon. I'll be posting an additional blog on the differences of the stat weightings specifically tomorrow.



Report Download

Scenario 1 – best T11 25 man gear, all buffs, raid boss

Scenario 2 – T11 10 man gear, all buffs, raid boss

Scenario 3 – my crappy gear, all buffs, raid boss

Scenario 4 - my crappy gear, all buffs, heroic boss

Scenario 5 - my crappy gear, sensible buffs, heroic boss