Archive for the 'Theorycrafting' Category

29
Jul
09

Theorycrafting – Warrior spreadsheet vs Maxdps

Who doesn’t like theorycrafting? Well, honestly, there are quite a few people who don’t enjoy spending hours crunching numbers and formulae. For the majority of the population, these pursuits take a backseat to actually playing the game. Even so, when in an instance or raid, most people like to know what a particular drop will do to their DPS. This is where the DPS calculator comes in, so that us mere mortals can take advantage of the numbercrunching efforts of others.

Enter our contestants of today. In the red corner, Landsouls spreadsheet from the Elitist Jerks forums. In the blue corner, MaxDPS.com.

Landsoul bases his sheet on the discussions and findings of the EJ hive mind; in other words a solid empirical database of knowledge, coupled with a lot of maths and numbercrunching. The result is quite possibly the best model of fury DPS outside Blizzard HQ. The downside is that the sheet isn’t exactly quick or easy to use for a first timer, especially not during a raid.

MaxDPS is the brainchild of Nuuga who has worked on it since 2006. It’s ever evolving and is continuously updated whenever classes are changed or new gear released. When you put in your stats it’ll give you a ranked list of gear based on your choice of instances. In other words, a very quick and easy way to check gear, even during raids. The downside…well, how well does it really model DPS? Is it accurate?

Which naturally enough leads to a comparison test between the two. Now, some people will say “but you’re only comparing two arbitrary computations!”, and yes I am. However, for now I will work on the assumption that Landsouls spreadsheet offers the most accurate modelling of fury DPS and mechanics, even if the numbers are off. For any simulation, there are assumptions and tradeoffs and we will in fact look at those at a later date.

For now, we’ll stick to comparisons between simulations. As a base for the following, I’ve put in the numbers and stats for my own current gear, including the most common buffs we have in my guild. We are currently working our way through Ulduar 10, so the gear is a mixed bunch of Naxx 10, Naxx 25, and Ulduar 10.

Landsouls sheet returns the following:

  • Overall DPS: 2752
  • White damage: 758
  • Heroic Strike: 302
  • Bloodthirst: 646
  • Whirlwind: 473
  • Bloodsurge: 231
  • Deep Wounds: 340

So far so good. Let’s see what MaxDPS gives us:

  • Overall DPS: 2619 (-133)
  • White damage: 535 (-223)
  • Heroic Strike: 443 (141)
  • Bloodthirst: 551 (-95)
  • Whirlwind: 459 (-14)
  • Bloodsurge: 435 (204)
  • Deep Wounds: 195 (-145)

The numbers in parentheses are the differences between the two. It’s pretty clear that the two do not really agree on anything. MaxDPS has significantly higher values for both Heroic Strike and Bloodsurge, but also very much lower values for the white damage and Deep Wounds.

There are several factors that can explain the discrepancies between the results. MaxDPS only takes raw stats for example, so weapon and item procs aren’t likely to be taken into account. Things like the Berserking proc, or the Loatheb’s Shadow on-use, or the Mirror of Truth proc (I could go on and on), will not be modelled by MaxDPS.

This is unfortunately not all. For one thing, we don’t know the exact rotation used by MaxDPS which is really the big unknown. The rotation description is woefully lacking on MaxDPS (it still says “Bloodsurge on BT proc” for example), so there is no joy there. What we’re left with is a largely unknown model which outputs numbers significantly different than a model widely accepted to be as accurate as we can get.

Let’s take an example from real life. I still offhand with the Claymore of Ancient Power. Now, suppose that I replaced that with a second Ironsoul

According to Landsouls spreadsheet, the breakdown would look like this:

  • Overall DPS: 2834
  • White damage: 780
  • Heroic Strike: 325
  • Bloodthirst: 665
  • Whirlwind: 492
  • Bloodsurge: 239
  • Deep Wounds: 332

To put the new data into MaxDPS we need to take the raw stats for the new gear configuration and put them into it. Essentially, we trade Crit Rating for Armor Penetration and Attack Power. And hey presto, MaxDPS spits out the following for us:

  • Overall DPS: 2751
  • White damage:539
  • Heroic Strike: 494
  • Bloodthirst: 569
  • Whirlwind: 481
  • Bloodsurge: 473
  • Deep Wounds: 194

We’re not as interested in the raw numbers as we are in the differences between before and after, so let’s take a look at the differences between the Claymore offhand and the Ironsoul offhand. I’ve denoted gains with a “+” and losses with a “-“. Landsouls spreadsheet is listed to the left of the slash, MaxDPS to the right.

  • Overall DPS: +82 / +132
  • White Damage: +22 / +4
  • Heroic Strike: +23 / +51
  • Bloodthirst: +19 / +18
  • Whirlwind: +19 / +22
  • Bloodsurge: +8 / +38
  • Deep Wounds: -8 / -1

As we would expect, our Deep Wounds damage goes down due to the loss of Crit Rating. And then it starts getting interesting. We see huge boosts in Bloodsurge and Heroic Strike for MaxDPS, but not a very big increase in white damage. Since Bloodsurge is an instant Slam, any increase in damage from it will either come from a)more frequent use or b)an increase in Attack Power. An increase in Attack Power would show a limited DPS gain, and it would also show up on Bloodthirst damage, so we’d expect to see a slight boost to Bloodsurge and a larger boost to Bloodthirst.

One example maketh not the theoretical DPS however. For the second example, let’s take the default set from the spreadsheet (as of 29.07.2009). It’s not entirely the BiS posted in the EJ forums, but it’ss close enough for jazz. To remove any “noise” from the calculations, we’ll furthermore remove everything but Battle Shout (vanilla version) and Rampage. We then get the spreadsheets results:

  • Overall DPS: 3556
  • White damage: 872
  • Heroic Strike: 493
  • Bloodthirst: 866
  • Whirlwind: 588
  • Bloodsurge: 282
  • Deep Wounds: 452

MaxDPS gives us:

  • Overall DPS: 3523 (-23)
  • White damage: 521 (-351)
  • Heroic Strike: 745 (252)
  • Bloodthirst: 714 (-152)
  • Whirlwind: 582 (-6)
  • Bloodsurge: 700 (418)
  • Deep Wounds: 260 (-192)

Again the parentheses show the difference between the two results. While MaxDPS nearly agrees with the sheet on overall DPS, the ability breakdown is quite different. The one exception is Whirlwind, where MaxDPS nearly agrees with the sheet. Of special note is the white damage contribution, which has gone down (!)

So, MaxDPS doesn’t in general agree with Landsoul’s spreadsheet. It still provides a gear list based on slots, with quick overviews of stats and drop place. Is it as updated as the sheet though with regards to gear? And the answer is: pretty much, but not quite. It doesn’t get updated quite as frequently as the spreadsheet, and so it might take a while for very new items to show up.

After all of the above, it’s fair to ask: “Aren’t we comparing apples and oranges here?” And the answer depends wholly on your viewpoint. These are two very different animals. On one hand we have a spreadsheet made to be as accurate as possible, and to include as many of the mechanics as possible, and on the other hand we have a simple-to-use gear interface for showing gear in a ranked fashion. On one hand we have a tool for the numbercrunchers and on the other we have a tool for the person who needs to see whether Gear Piece X is an upgrade, before the Need/Greed timer runs out thankyouverymuch.

If you’re looking to tweak your gear, spec or to see what a bit of kit will do to your DPS, then Landsouls spreadsheet is the better of the two. Likewise, if you’re in a raid/instance and a piece of gear drops that might just be an upgrade, although possibly not, then Landsoul is again the better. MaxDPS most likely cuts some corners with the calculations and ends up some numbers that are…well, let’s call them puzzling. On the other hand, if you look at MaxDPS as a quick and dirty guide to gear, for the raider or dungeoneer who has been caught unawares or unprepared, then it does come into its own. It’s as simple as clicking an icon and looking at the list. The ranking system may be off, but the list is still there.

Verdict: Landsouls spreadsheet snatches this one with its more accurate and believable model of warrior DPS.

Advertisements
26
Jun
09

The new fury rotation

With the latest patch we received a small but noticable change to the fury playstyle. Yes, I’m talking about the reduction in CD on Bloodthirst. Instead of a 5 second cooldown (which doesn’t really fit with a 8 second Whirlwind cooldown), we now have a 4 second cooldown…which happens to fit nicely with Whirlwind, right? Right?

This is where I go: yes and no.

Yes: There is no more constantly delaying BT just to keep the WW cooldown inviolate.

No: Both 4s and 8s fit terribly with the 1.5s GCD.

But this is nothing new you say. Correct, it’s nothing new, but in the past our cooldowns were long enough to leave us space inbetween for other stuff. If we are determined to keep BT and WW on constant cooldown, we have a whooping 1 free GCD per 8 seconds. Read that again and then tell me if you’re dancing with joy. For reference, with the 5s cooldown we had 5 free GCD’s per 16 seconds, as well as 3 BT’s. On the flip side, we have 1 more BT per 16 seconds.

So by keeping to the BT and WW cooldowns we are effectively trading 4 free GCD’s for 1 BT. Hardly a terrific increase in DPS I’d say.

But then, these days we are moving away from the oldschool style fixed rotations and into priority lists. As good or bad as that might be. And whether we like it or not, by keeping our inviolate rotations, we are more or less stifling ourselves. So, let’s look at the 2 main alternatives there are, at least as I see them:

  1. Keep the WW cooldown inviolate, GCD the rest
  2. Fit everything to the GCD

You can of course argue that there are more alternatives, and this is true. There are more ways to do it. However, I have picked the two above, since they are inherently simple and easier to pull off than some others. In other words, you dont need a timer running to pull them off. They are guided by the GCD and the cooldown on WW (which a mod like OmniCC will let you see easily). In the following, I shall refer to them as rotations 1 and 2 respectively.

Rotation 1 (duration: 16s) looks the following:

  • WW at: 0s, 8s
  • BT at: 1.5s, 6s, 11s
  • Free GCD at: 3s, 4.5s, 9.5s, 12.5s, 14s

Rotation 2 (duration: 9s) looks like:

  • WW at: 0s
  • BT at: 1.5s, 6s
  • Free GCD at: 3s, 4.5s, 7.5s

You’ll notice that rotation 2 is 9 seconds long, which means that WW will be off cooldown for 1s each rotation. Shock horror! However, before people start commenting, let’s look at the number of abilities per second:

Rotation 1:

  • WW: 0.125/s (0.125/s)
  • BT: 0.1875/s (0.25/s)
  • Free GCD: 0.3125/s (o.125/s)

Rotation 2:

  • WW: 0.111/s
  • BT: 0.22/s
  • Free GCD: 0.33/s

Numbers in parenthesis are the inviolate rotation numbers, where BT and WW are not delayed. So, we see what is happening. By sacrificing a bit of WW damage, we gain more BT damage and more potential for bonus Slams. We are now faced with the question: which rotation is the better rotation?

The answer is of course: the one with the highest DPS output. Duh.

Instead of trying to justify my flippant answer, I’ll analyse the two rotations a bit further. Essentially, up until 7.5s they are identical. In fact, you could call them merely variants of the same rotation. The main difference is whether we delay the WW for 1s to squeeze in a Slam. But, is it ever justified to violate the inviolate? The answer is of course dependent on what numbers you put out. And this is where your mileage varies, where every person needs to do their own numbers.

It’s perhaps important to note that I’m not saying that WW is no longer the most important ability we have. If your rage says you can either Slam or WW, then you WW. No question. At the other end of the scale, if you are at 100 Rage all the time you are wasting DPS. In such a case, delaying the WW for a second to squeeze in an extra Slam is justified, provided you do not lack rage for subsequent WW’s and BT’s.

08
Apr
09

On theorycrafted DPS versus ingame DPS

Theorycrafting. Reverse engineering the mechanics of WoW, unravelling what Ghostcrawler & co. have “ravelled”. At the heart of it all is a desire to better our game, to progress faster through the content.

For many, however, this is not an interesting pastime. So, enter the gear guides and performance simulators (spreadsheets, applets, webpages, you name it). There is a whole forest full of helpful sites you can go to, and depending on where you go, you’ll end up utterly confused because none of them really agree.

The reason is simple really. DPS output depends on a combination of gear and abilities (also known as playstyle). It is the interaction of these two factors with the mechanics of the game that produce the DPS. We as players control the two former, but not the latter. We can only guess at the exact mechanics, and therefore we end up with models, based on the creators analysis of observed phenomena. Yes, empirism at work.

More often than not, two models will not agree on the value of certain bits of kit, which doesn’t really help the poor raider who is trying to make sense of it all while desperately looking up the epic that dropped just now. The all important, but underlying, question is this:

Which model should I base my gearing choices on? Which model is more precise?

Big topic, as you can probably guess. And one which does not have a final and inherently correct answer. Sorry to disappoint you, there’s no big revel at the end here. There will be lots of my (probably biased) opinions however. I’ll look at the topic in three ways:

  • Comparing various DPS simulators found on the net
  • Comparing simulated DPS to real DPS
  • Discussing the above, with input from all you people

Why the last point? It’s a question of available data material, which forms the basis of any empirical analysis. I might have a lucky day, or a bad day, and I might have a different playing style than others. So, if you have data, information or other material of relevance, bring it forward and let’s take it from there.

In the next installment, I’ll get down to the nitty gritty and compare two DPS calculators: MaxDPS.com and Landsouls spreadsheet from Elitist Jerks.

Note: With 3.1 upcoming, there will be some hefty changes to theorycrafting for warriors, due to the changes to several of our talents. As such, you can argue that analysing current DPS calculation tools is a waste of time. To that I reply: You’re the one who made it all the way to the end of this article.

03
Mar
09

Commanding the Cleavage

Guest post at BigHitBox on 03 march 2009.

I know what you’re thinking. Enjoy those thoughts for a moment. Enjoy them, until I tell you that this is about the -other- cleavage…the one with the capital C, like so: Improved Cleav(ag)e. More to the point, Improved Cleave versus Commanding Presence. Two of those early talents that you have to take on your way through the tree.

Improved Cleave is a highly situational talent, usable only in fights where you have multiple mobs. Apart from trash, this is only true for some current raid bosses. Improved Cleave can furthermore be glyphed to include a third target, making it more of a lesser whirlwind. Like Heroic Strike, it doesn’t consume a GCD, however it does take up an otherwise ragegiving swing. Essentially, it gives you a small bonus to the melee damage you wouldve done anyway, while eating up rage instead of giving it. On the other hand, it does hit an extra mob, which Heroic Strike doesn’t.

In the other corner, we have Commanding Presence. It’s a flat boost to your AP, so it affects basically anything you do. It will also affect your party or raid members, so its utility value is higher than Improved Cleave.

There are four options we can spec our way into:

  1. 5/5 Commanding Presence
  2. 2/5 Commanding Presence, 3/3 Improved Cleave
  3. 3/3 Improved Cleave
  4. None of the above, i.e. you’ve put your talent points elsewhere

The last possibility is mainly for reference, even though it is theoretically possible to avoid both talents, for example by speccing into Improved Demo. Shout, Blood Craze or Unbridled Wrath. Possibility 3 is a similar situation. For raiders, possibilities 1 and 2 will no doubt be the most interesting (unless you’re positively sure you have an Improved Blessing of Might in your raid). Put bluntly, option 2 is for the person who cares less about the wellbeing of their raid, or who is lazy and has another warrior to keep up the better Shout.

Unimproved Battle Shout is 548AP, so a 5/5 spec will net us 685AP and a 2/5 gives us 603AP. Cleave does 222 bonus damage, and a 3/3 Improved Cleave does 488 bonus damage. In the following, I’ll assume an AP of 3500, and an average weapon damage of 653. With those numbers, we get the following:

  1. 1921 damage
  2. 2166 damage
  3. 2022 damage
  4. 1756 damage

So, Improved Cleave outperforms the gains from Commanding Presence in both cases. What can we conclude? Sadly not a whole lot. Taken in isolation, Improved Cleave offers a larger damage increase per talent point than Commanding Presence. Compared to Heroic Strike, Cleave suffers from a highly situational nature. Heroic Strike, if you are facing only one mob, hits harder for less rage used. With 5/5 CP up, a Heroic Strike will deal 2194 damage, so its damge/rage ratio is far better where single mobs are concerned. With 2/5 CP we get 2173 damage, compared to the 2166 for an Improved Cleave. Heroic Strike wins out again.

Saying that Improved Cleave is useless is also wrong though. It does come into its own in multimob situations, especially if Glyphed. For grinding and soloing, and for instance trash, Improved Cleave is definitely a useful tool to have in your arsenal. Simply put, Improved Cleave brings you instant gratification, in some situations, while Commanding Presence brings overall zen and wellbeing. There is no strictly right or strictly wrong way, since it depends entirely on the situations you face.

So in which situation would Improved Cleave give an overall damage increase? That will happen when the bonus damage from Improved Cleave overshadows the overall contribution from the extra 15% bonus to Commanding Presence. A rough estimate would say that unless your overall Cleave damage, in all situations, is about 12-14% of total damage, it is a net DPS loss compared to 5/5 Commanding Presence.

17
Feb
09

Rendering a rotation

In this post, I discussed the feasibility of a Rend rotation. The following is an example of a rotation. It’s by no means the only one, and it may prove less ideal than others, but its my initial idea for a Rend rotation. The assumtion is an unglyphed Rend.

You will note that it’s only 16s long. The reason is, that two full CD’s of WW takes 16 seconds. BT comes off CD after 16.5, but due to the GCD, it cant be used before 17.5s, which is essentially the same situation as at 1.5s.

  • 0s: WW
  • 1.5s: BT
  • 3s: Rend
  • 4.5s: (Slam)
  • 6.5s: BT
  • 8s: WW
  • 9.5s: (Slam)
  • 11.5s: BT
  • 13s: (Slam)
  • 14.5: Free GCD
  • 16s: WW

Rend will be up for 15/16ths of this rotation, following the initial 3s delay. Bloodsurge procs that occur between 14.5s and 15.5s will run out, due to the application of WW, BT and Rend at the beginning of the rotation.

To be viable for inclusion in the rotation, Rend would have to make up for the potentially lost Slam. Since the Rend rage cost is double that of Slam, from a pure damage/rage perspective, it should do double the damage to break even.

Assuming an AP of 3300 and using the Demise (happens to be my current MH), furthermore assuming an unglyphed and non-improved Rend, we get the following values:

Slam: 1728 damage (115.2 damage/rage)
Rend: 1858 damage (61.9 damage/rage)

Not much of a competition really. But the Slam we are losing has only a 20% chance of occurring. Furthermore, we have to subtract the chances of it proccing off the subsequent WW and BT, both of which would negate the proc by overwriting it. Hence, the probability is 12.8%. So over a sufficient number of rotations, the resulting damage is:

Slam: 0.128×1728=221.2 damage
Rend: 1858 damage

This translates into the following DPS increase:

Slam: 13.8
Rend: 116.1

While Slams damage/rage ratio is still better than Rend, it translates into less dps due to the proccing nature of Bloodsurge. It is furthermore only relevant to procs occurring in the 1s window between 14.5s and 15.5s, which basically means it has to proc off a HS which hits at some point after 14.5s. If it hits earlier, there is a theoretical possibility of hitting Slam again at 14.5s without delaying the WW.

The keen observer will not that all of this discussion is based on a 1 second window at the end of a rotation. In practical terms, I think its fair to include any HS between 13s and 16s that hits. Even with flurry and haste, it is unlikely to be more than one however, given the weapons speeds of TG warriors.

The tentative conclusion is, that Rend is viable, from a dps point of view, given latencies and reaction times that support it without delaying the normal rotation. So far so good. The last step is to examine the viability from a rage generation viewpoint. So stay tuned for more rending madness.

16
Feb
09

Will warriors have to relearn their waltzing?

Stance dancing is taking a major turn in patch 3.1. For better or for worse. Instead of the old mechanic where we save up to 25 rage on a stance shift, it’ll now cost us between 10 and 25 rage to shift.  So we’re no longer guaranteed to have any rage following a stance shift.

The critics of the new mechanic, as I see it, are mainly PVP’ers, who rely on having rage to be able to CC opponents. For a PVE person like me, the change is actually a great benefit. Currently, doing anything outside of zerker stance is a big nono for me, the rage loss is simply too great. Which means that things like overpower and rend are never used.

The question is, are the Battle Stance Only abilities worth it? Specifically, is Rend worth the 30 rage and the time it takes to switch back and forth?

The answer to that question will take quite a bit of testing and number crunching. It’ll also involve quite a bit of button mashing to reduce the time lost. Any rending will have to happen while Whirlwind and Bloodthirst are on cooldown. Ideally, it will not result in losing any Bloodsurge procs either. And finally, we dont want to cap rage.

The timing problem is eased somewhat, since stance changes are not on the GCD, so its theoretically possible to stancedance while the GCD runs. However, one has to take the latency into account, which does factor into an extremely tight rotation. And of course, theres the human reaction time to add to that.

Does it sound like a tight squeeze? It is. At most, we’ll have a window of 3.5 seconds where we do not have to hit WW or BT. The CD of WW is 8 seconds (assuming a glyph), and the CD on BT is 5 seconds. So ideally, we can get 2 BT’s off during the CD of WW (1.5s GCD+5s BT CD+1.5s GCD=8s), which leaves 5s-1.5s=3.5s between our BT hits. Plenty of time to get off our Bloodsurge Slam, which triggers the GCD.

There remains 2 seconds in which to Rend. At best. Subtracting the GCD we have 0.5 seconds for human reaction and latency. Remember, this is the best case scenario, where we have the largest “empty” window betweeen our crucial CD’s. Luckily, Rend is a DOT which runs for 15/21 seconds, so it we don’t need to apply it often. The WW/BT rotation is 16 seconds long, so we will lose 1s where the mob doesnt bleed (unless we glyph it of course).

The conclusion is, yes its viable to use Rend. At least when you look at the timing required. Its a tight squeeze, and it requires timing down to the second (as well as a minimal latency), but its theoretically possible.

The above is only a timing discussion, whether its at all possible to fit Rend into a rotation without delaying WW or BT. The next question is, whether it is worth spending the rage on it at all, whether the damage/rage ratio is high enough to warrant it. Finally, its a matter of pure rage generation; do we generate enough rage at all? Using Rend at the behest of WW, BT or Slam is quite simply a no go.

We’ve tackled one issue. But we’re not at all done yet. Until next time.