Understanding Spreadsheets

Hunter spreadsheet results is a topic that I get a lot of email on, and recent discussions of how ICC loot affects MM vs SV has sparked these questions into overdrive.  I feel like spreadsheets are massively misunderstood, frequently misused, and I’d like to take a moment to explain what they do as well as what they don’t do.

Spreadsheets are a fantastic and well-crafted tool that are shockingly accurate for many calculations. I use hunter spreadsheets regularly myself. However some people discover the spreadsheets and become devoted to them. They spend hours tweaking talents and gear and seeing how it changes the theoretical dps, and then treat the results as gospel. I call them Spreadsheet Disciples.

Spreadsheet Disciples

These spreadsheet disciples generally don’t understand the actual math and assumptions behind the calculations in the spreadsheets. They aren’t theorycrafters, but they think they are. What I mean by this is if there is a discussion about talents or specs, these are the people who will come on and say “No, this is 23dps better, just plug it into the spreadsheet.” And it doesn’t matter what the discussion is, or what kind of logical arguments someone makes, their answer will always be the same: the spreadsheet says so.

I think Blizzard said it best when responding to the increasing wave of Spreadsheet Disciples:

“Not everyone is going to be an expert theorycrafter or mathematics expert, and we’re not expecting you to be one in order to participate in these forum discussions. But also be honest about the fact that linking to someone else’s work doesn’t necessarily turn you into one either.”

Well said, Blizzard.

What Spreadsheets Do

Spreadsheets attempt to model the statistical average dps you would do in a completely stationary fight against a single boss target.

The first step to understanding spreadsheets is to understand that statement. Seriously, think about it for a moment. Now try to list every WotLK boss fight that fits that description. Patchwerk and Koralon are the only ones that come to mind, and a couple of others that come close.

From this theoretical model, spreadsheets go on to extrapolate out certain information, such as the amount of dps that each stat is contributing, the value of different gear, or talent points.

What Spreadsheets Don’t Do

Anything other than that.

When Spreadsheets May Not Be Right

There are three words in the definition of what spreadsheets do that could introduce errors when you use their calculations to predict the actual raid dps differences between abilities or stats: average, stationary, and single. Let’s look at each of these:

• Averaging: spreadsheets generally average abilities with cooldowns – this is very common in theorycrafting and often the only logical way to go about it. However, as we’ve discussed before, skilled hunters intelligently stack their cooldowns, giving a multiplicative effect that’s greater than spreadsheets predict. The benefit adds up to a significant dps gain over the course of a boss fight, and different specs see different levels of benefit, depending on the cooldowns and the amount of control over them.
• Stationary: we all know that in almost every boss fight we’re going to have to be moving. Out of void zones, after adds, away from adds, chasing the boss around, kiting, etc. All hunters, regardless of spec, should constantly be working to lose as little dps as possible while moving – using jump-shots, shoot & scoot, and disengage movement. But that we are going to lose some amount of auto shots and steady shots is just a bare fact. The effect of that loss on stats and specs varies in ways that spreadsheets don’t even try to predict.
• Single Target: increasingly we are seeing additional targets pop up in boss encounters. TotC had exactly zero single target fights.There are only a couple of ICC fights that are wholly single-target.  Skilled hunters can swap targets without missing a beat and understand the effect of their shots on each target and adjust their shots accordingly (such as leading off with arcane shot on non-sundered adds, or saving shots when they know a switch is coming). In addition, the best hunters are able to keep rolling serpent stings on two targets at once – or even more if the situation warrants it.

Hunter Specific Examples

There are many, many examples of abilities that you can either get more than predicted benefit from, or that you can be certain you’ll get less than predicted benefit from. Let’s take a look at a handful:

• Haste: since haste only affects our auto-shot and steady shot, movement completely negates any benefit we’d get from haste, including Imp AotH and raw haste on gear. Realistically, haste is always worth some amount less than spreadsheets tell you.
• Armor Penetration: again, if you’re moving, the benefit from armor pen also reduces.  All three specs leverage armor pen differently — MM a lot, SV less, and BM the most. And the danger of lots of ArP is losing a lot of your stats while on the move. On the other hand if you have an armor pen proc that you can stack with another cooldown – agility, attack power, or haste – you can get a multiplicative benefit from it.
• Rapid Fire: if you have to turn and run just after popping rapid fire, you’ve just lost the strongest hunter cooldown completely. However, since in most fights you can predict when you’ll have gaps where you can stand still, rapid fire is the single strongest multiplicative cooldown you have, that can be stacked with trinket procs, on use trinkets, and boss mechanic effects that increase damage. MM can potentially gain an enormous multiplicative benefit from the ability to use this effect 4-6 times per boss fight – a benefit that spreadsheets do not model.
• Sniper Training: a talent that just clearly is going to have some amount of down time in most boss fights, thus not as good as the spreadsheets model.
• Black Arrow: the dot and lock & load proccing of black arrow are benefits you’ll just about always be able to get. However, a very strong benefit of black arrow is that it increases all your damage done by 6%. If you have to move while black arrow is up and lose auto shots or steady shots or steady shots, you’re now getting less benefit from it than spreadsheets predict. Alas we have very little control over this, since saving it for more than a few seconds also results in a net loss. Similarly the benefit of procs with black arrow are highly subject to randomness.
• Wild Quiver: again, while moving it can’t proc. With such a low proc chance though, the loss is pretty minor.

A Point of Proof

With the new gear being released, many people are now crossing into the point where their gear on average is about equivalent of the previous BiS TotC heroic geared people. And many of them are finding the spreadsheets telling them that SV is a bit better dps. And for some, it may be.

But understand that we can look up the top hunter dps benchmarks in the world and look at each fight from people with similar gear levels, and those show us what kind of dps the best hunters in the best raid groups can do (and before you complain about comparing to the very best – realize the spreadsheets model a hunter even better still). And it’s not just that MM is doing more dps, it’s that MM is doing a lot more dps. Certainly more of a difference than the spreadsheets would show. And you know what, that makes perfect sense once you understand how spreadsheets work.

What Spreadsheets Are Good For and Less Good For

The first and greatest benefits of spreadsheets, in my opinion, is that they do a pretty darned good job of evaluating how good stats are compared to each other. It’s a qualitative analysis so ignore the dps number. One agility is not getting you “1.63 dps,” but it is going to be a bit more than twice as good as 1 attack power. It’s good for comparative purposes. Just realize that your haste and armor pen benefits are going to be a bit smaller than the spreadsheets predict.

And since it’s directly related to stats, the spreadsheets tend to be very good at gear evaluations. However if any bit of gear has a proc or on-use ability, that’s a red flag that the spreadsheet will be averaging it. If it can be intelligently stacked, the benefit will be greater than the spreadsheets show.

In general, the spreadsheets can give you some very good information about talents as well; however, understand that this is where the modeling starts to break down significantly. As we discussed with rapid fire, black arrow, wild quiver, sniper training, rolling serpent stings — when you start to pile them all on top of each other, the discrepancies are not minor, but potentially quite major indeed.

But as long as you understand how the spreadsheets are working, you’ll be able to tell when the spreadsheet results are likely to be wrong. You may not be able to calculate exactly what the dps will be, but you can easily look at sniper training, and say “Oh yeah, my dps will be a bit less than it thinks.” or look at the Death’s Verdict trinket and say “Oh, I can stack this, so it’s even better than the spreadsheet shows.”

In the end, the spreadsheets are remarkable and useful tools. Use them, but don’t worship them.