Regional Flag3.3.3 Worked (sorta) for DKsSource
avatar
Zarko
Target Source
#0 - 2010/04/11 12:03:53 AM
So, one of my pet hobbies around here is representation statistics, simply because I think it's interesting. let me get this out of the way to help avoid trolling later:

1. I don't think that buffs/nerfs should be based off of it.
2. This is a small sample size, but one that I test for statistical significance.
3. Of course blizzard has larger numbers
4. I am bad and need to learn to play

Let's get to the data.

In 3.3, there was a problem with DK representation: notably, that there was less than 10% representation for DKs on two tank fights. The problem got worse when you got outside of the top 100.

For Festergut25 and Sindragosa25, parses in the midrange of guilds, tank representation looked like this:

Warriors: 163 tanks (40.24%)
Paladins: 141 tanks (34.81%)
Druids: 75 tanks (18.52%)
DKs: 26 tanks (6.42%)

3.3.3 came, and brought with it some massive DK buffs. DKs are still thought of as poor tanks (we can see that in these forums sometimes, even if it's not really true outside of Nerubian burrowers), but their numbers have changed.

200 Parses from 3/23 to today on page 1,000 of WoL or later (parses number 24,976 and later)
Warriors: 140 tanks (34.82%)
Paladins: 144 tanks (35.82%)
Druids: 70 tanks (17.41%)
DKs: 48 tanks (11.94%)

I ran a Chi-square on the numbers, and the P value is .037, meaning that this change is overwhelmingly unlikely to be due to random variation. 3.3.3 has been successful in more DKs tanking at a lower level.

For hardmode data, Dks were never in quite as bad a place as they were for lower level parses. However, the differences there are not yet statistically significant. Still, the point remains--DK representation is on the rise.

Most of the gains in representation seem to be coming at the expense of warriors as well, interestingly. Which is probably (although not certainly) working as intended. It's a fairly small amount at the moment. However, the gains mostly came from a decrease in double warrior groups. There were 24 double warrior groups in the earlier samples, and 15 in the later samples. That's 40% of the difference in the warrior numbers. The rest is the slightly higher likelihood of Paladin/DK comps.

avatar
Blue Poster
Target Source
#67 - 2010/04/14 08:46:11 PM
Q u o t e:
I would speculate that this is an effect of perception. DKs felt / were perceived as weak for quite a time during the TOC era, which translates into people moving away from playing their DKs and into something else.


I think perception is huge in these matters, much greater than most players give it credit. WoW players generally really, really hate to be wrong. What I mean is, they don't want to think that they made a bad choice in their character class, their gearing or their talent choice. In WoW it's somehow less embarrassing to fail at something than it is to be caught being wrong.

"You wiped on heroic Saurfang? Yeah he can be hard. You had a paladin healing the raid? What, are you crazy? I don't care if it worked for you. It's just wrong. Ugh, and look at your glyphs." As a result, rather than basing decisions on noticing empirically whether something does or does not make a difference, many players would much rather someone tell them the right or wrong way to do something.

So we see the level 40 mage that I saw in Dalaran last night, spamming trade chat asking whether Fire or Arcane does more damage. (Answer: Dude, it doesn't matter if you're level 40 and probably doesn't matter if you're level 85 unless you're going into a really tough Arena match or talking about heroic raid bosses.)

I think players switch tanks sometimes when they think the forums are telling them that they're using the weakest main tank. "Well, I want to plan for success, so clearly I need a different tank," even when in reality that choice probably affects their chance for success very little. Switching tanks might give them a 3% greater chance of success. Not standing in fires might give them a 30% greater chance of success.

Likewise, I think players sometimes drift off to an alt when their class or spec gets nerfed, just because it sucks to get nerfed and they feel somehow less potent knowing that they were nerfed, even if the effect was imperceptible for them.

Perception.

Q u o t e:
The only logical reason then to bring about tank representation balance (assuming that this should be the goal of Blizzard) is to disproportionately make DK tanking attractive over Warriors and Paladins.


It’s not a goal for us. There are a lot of things that can affect representation. Balance is definitely one, but as I mentioned above so is perception, and then you have tradition, play style (what’s fun to play), utility when not tanking (including healing, but also dps and even PvP) and even art. It’s complicated.

Now if you see a sudden shift in populations, then there might be something happening that you need to pay attention to. Something happened to upset the inertia and drive player behaviors in a certain way. It could be a lot of things. It could be perception. But it could be an actual survival of the fittest deal where raids using one class struggle and those using another class succeed, so gradually you see the latter increase over time. It's hard to separate out all of the various causes for behavior like that, but it's probably good to take notice when they do.

avatar
Blue Poster
Target Source
#96 - 2010/04/15 07:02:33 AM
Q u o t e:
Why isn't it a goal for you? Do you not want the satisfaction of knowing that every class and spec has as much a devoted following as the next?


I found the rest of your post to be pretty ranty, but I think this point is worth elaborating on.

It isn't a goal because of what it does to the game. Let's take the over-discussed example of warrior tanks coming from vanilla through BC and to LK. It should surprise nobody that there were a lot of them, seeing as how they were the only real tank for awhile and how guilds had their established warrior MTs.

So say our goal was 25% representation of all 4 tanking classes. That means we have to clobber warrior tanks really, really hard. We either have to nerf them or make them punishing to play, or we have to somehow buff the other three classes or make them somehow so much more fun to play that players abandon warriors in droves.

Despite what the conspiracy minded among you might think, that's not what happened. It would feel terrible and warrior players would hate it.

Let's take another example. Paladins. There are a lot of them. That might have something to do with the fact that they can fill three totally different roles. Healers can't play a warrior. They can play a paladin. If the class has broader appeal it might not be surprising that they have high numbers. Is the answer to kill off one of the roles or add one to the warrior, or again nerf or buff someone to such an extent that players feel punished for picking the wrong one? Yuck.

Let's take another example. Death knights. They're shiny and new. They start at high level, have a tailor-made start zone and leave it in all blue gear. Do we then reward those players by beating them down until they're only 10% of the population?

Here's a preview. You're about to see a lot of goblins and worgen. Does that mean all of the old races are terrible? Probably it just means they're old and players have been playing them for six years. Is it really logical to strive for perfect racial balance? Can a troll ever hope to compete with a blood elf for popularity? :)

Every class and spec does have a devoted following today. We don't think we need to play the game of shoving representation up and down until it's a ten-way tie. Heck, some players are drawn to certain classes because they're rare (they either like to up their chances of getting an invite, or they just like to be the rare snowflake.) Now if players seem to avoid a class or talent tree like the plague, then maybe something is going on. We don't really see that though.

(To be clear, I'm talking about number of players playing various classes at 80. It's a bigger deal to us when some classes have very low numbers in raids or PvP.)

avatar
Blue Poster
Target Source
#97 - 2010/04/15 07:05:10 AM
Q u o t e:
That was the whole point of the thread. There has been a shift in populations. I'm tracking to see if this shift stabilizes here, or further grows


Yeah, and it's a good attempt and it's nice to see statistics and players not trying to over-interpret data.

None of that means we balance around representation though.

avatar
Blue Poster
Target Source
#98 - 2010/04/15 07:20:01 AM
Q u o t e:
A similar situation is quickly becoming apparent to my raid on LK HM. A paladin with stuns, demo shout and an auto-guardian spirit is just more viable to deal with Shambling Horrors not to mention Righteous Defense and Holy Wrath for later on in the fight. Sure most of that can just be covered by another raid member while a DK tanks, but it's certainly significant enough to make a difference in a fight as difficult as LK HM. Having said that, I'm sure that continuing to increase the ICC buff will ultimately eliminate (read: mask) the real problem which is while the theory of non-uniform tanks is great, you continue to design encounter mechanics that initially favor one over the others.


Encounter design will always favor someone over someone else. It's probably impossible to make things absolutely even across every class (or even every tank class) on encounters that are tuned this tightly. And really, we wouldn't put that kind of constraint on the encounter designers. It's tough enough coming up with mechanics that you guys aren't totally sick of seeing while still giving you enough flexibility to bring whoever you want.

If we thought it was impossible or even really, really difficulty for some classes to tank some encounters, then that would cross the line for us. Death knights coming off of Sarth and going into Ulduar crossed the line because the advantages they had seemed like they were giving them an advantage on every fight. Blocking made too much of a difference on Anub, and we probably won't use that mechanic again. The Lich King hits so freaking hard on 25 heroic mode that a giant health pool is pretty useful. But the three guilds that have killed him now have used pretty different comps and even different tanks, so that seems pretty successful. So few guilds have anything approaching a reasonable shot of being able to handle that fight until the buff stack gets much higher that it seems a little silly to worry about it too much. (Props if you guys think you're close.) Meanwhile, you can do every other encounter in the game with the tank of your choice without being at a major disadvantage. Doesn't sound that bad overall.

avatar
Blue Poster
Target Source
#99 - 2010/04/15 07:23:07 AM
Q u o t e:
It's almost as if rather than just making sound good, creative, progressive game decisions based on years of experience and expert knowledge, you gathered way too much 'empirical' information and somehow became a team of scientists and marketers and forgot that game design is an art, first and foremost.


It's totally an art. I agree with that. We trust our instinct more than anything else when making decisions.

But, and this may be my own personal background showing, you can't be afraid of data either. There's a big difference between designing encounters based on focus groups and knowing how many players have beaten certain encounters.

avatar
Blue Poster
Target Source
#102 - 2010/04/15 07:27:04 AM
Q u o t e:
Damn I wish I had your job, GC. It sounds like so much fun. How about I sit on your shoulder and be your parrot?

It is. I love my job. We still love WoW and we still play it an awful lot. I feel the need to point that out from time to time, because the forums can get so negative and the impression can be that players just yell at us all the time. But we talk to enough players that we know that's not the whole story. We're pretty happy with what we've been able to accomplish and we've got big plans to be able to accomplish more.

avatar
Blue Poster
Target Source
#103 - 2010/04/15 07:28:18 AM
Q u o t e:
Simplifying immensely, there are two available goals:

1) Make every player equally happy, regardless of class played.
2) Make every class equally played, regardless of player happiness.

Unless class played has no effect on player happiness, these goals are not mutually achievable. Unless the classes are completely identical, how would they not have an effect on player happiness?

Given that, why would it ever be a good idea to choose the second goal over the first?


Demanding players choose equal numbers of each class is not reasonable.


Totally agree. Number one is emphatically more important.

Q u o t e:
The goal is not to make everyone happy, the goal is to make the choices equally attractive. This may very well mean that the people at the top need to be made miserable until such time as their ownership of tank equity is near equivalent to the others.


Sorry, I just really disagree with that. The goal is to make the game fun. Things like "the goal is not to make everyone happy" and "people at the top need to be made miserable" don't sound like a fun game. :(

The goal is not to make you feel really awesome about your current class. The goal is to get you to love WoW so that you're still here playing with us a year or five years from now.

avatar
Blue Poster
Target Source
#107 - 2010/04/15 07:39:11 AM
Q u o t e:
Surely representation does mean something when trying to balance out the classes. I understand that the overall representation might not be the the best statistic, but when looking back, it feels like representation was the main driver to some of the nerfs/buffs. Sartharion and General Vezax made Death Knight tanks seem invaluable, and when looking at their representation, surely that was something that mattered too.


It does, but it's *a* data point. The impression you get from some of these tin foil hat representation arguments is that the Blizzard computer spits out current percentages, and we buff or nerf until we get the next numbers and see if things look closer.

At the time of Sarth, everyone we knew was switching to DK tanks. Our own guilds were switching to DK tanks. The strategy guides said "Get a DK MT." The boards all discussed how DKs could tank the content. The thought leaders and theorycrafters that we know all were saying the same thing. Our tests seemed to show a big advantage. The community's numbers seemed to show a big advantage. It just felt like pretty consistent evidence. We honestly haven't seen anything like that since. The block tank on Anub'arak burrowers was probably the next closest thing (and not that close), but even in that case you just needed a tank with block while your MT could be anything, so it was only those groups with a druid + DK combo that seemed like they were really struggling, and druids and DKs had already had lots of moments to shine.

avatar
Blue Poster
Target Source
#108 - 2010/04/15 07:41:04 AM
Q u o t e:
Honestly, I think the real problem with Death Knight tanks right now is that their niche just seems irrelevant. How many "magic using" bosses are there in ICC or ToC that a large health pool or other cooldown won't get you through?


That's why we gave up on that niche (as magic tanks) long before we shipped LK. :)

That was the original design, but that was under a model where we assumed guilds would bring Sunwell-like raid comps and just swap in the DK tank for certain fights. As we got farther into development we kind of realized how whacked that whole philosophy was and the whole "bring the player" era was born.

avatar
Blue Poster
Target Source
#109 - 2010/04/15 07:44:24 AM
Q u o t e:
From Shigeru Miyamoto "I try to let the players understand something very simple as the basic background and then I challenge them so they will like to challenge many different things to become creative."


He's a very smart man. I've met him twice I think. One time at E3, in the ES days, we sent a bottle of wine over to his table. It was a pretty epic moment. I can't think of a game developer that doesn't just love him. We all owe him a lot.

avatar
Blue Poster
Target Source
#115 - 2010/04/15 08:20:26 PM
Q u o t e:
The difference is that guilds did Sarth 3D a short while later with other tanks. The same cannot be said about HANUB. (So if we really want to nitpick, it's in fact much worse)

I think we're quibbling here, but nobody was very scared about druids tanking Sarth. Some did tank him no doubt, but all of the hoopla was about how DKs trivialized Sarth and were then about to trivialize half of the Ulduar fights.

I'd also argue that you still needed a very good MT on Anub but you could use almost anyone on Sarth as OTs.

Regardless, it didn't feel like the same magnitude to us. I'm not sure we'd do the Anub fight that way again if we had to do it over. We weren't setting out to make an encounter that showcased the power of block. It's just the way the debuff worked and we thought in that case that we should reward players for cleverly figuring that out rather than punishing them for upsetting tank balance.

The biggest difference was even the players most frustrated about their inability to tank Anub adds weren't saying how block was going to make warriors and paladins the uber Icecrown tank. They knew the benefit was limited to that fight. The DK cooldowns on the other hand were something that were going to be good on almost every boss fight that did periodic massive damage (which, hey, most Ulduar bosses did).

And running concurrently through all of this is the "bear paradox" wherein you have the tank who lots of players believe is overpowered but that few groups actually use. :)

avatar
Blue Poster
Target Source
#118 - 2010/04/15 08:32:10 PM
Q u o t e:
A single big nerf will shift perception far more than a series of smaller nerfs.


I'm not sure I'd agree with that. I used to. But we also found that in some cases, it felt crappier to the players to see a small nerf in every single patch coming down. I think some would have preferred to just have us pull the band aid off all at once. A spec that feels like it's just going to lose ground over long periods of time doesn't tend to attract new players. :)

But I also agree that changing the cooldown to Will of the Necropolis managed to change the perception of DKs from understated to overpowered single-handedly (even though it still doesn't fire all that much).

avatar
Blue Poster
Target Source
#122 - 2010/04/15 08:42:28 PM
Q u o t e:
But you could've also at the same time fixed many of the quality of life issues that DK tanks faced to not make the population plummet so much. Do you regret that at all?


Eh, that's a big topic and I think they shut off the generators on this remote yet fantastic island I am currently on very soon.

I always regret when players who loved a class end up feeling bitter over it. However, I also think a salient point is that there was a lot of early gushing from players about how they loved the DK rune system or they liked being the cooldown tank or they liked doing melee and spell damage or they liked the talent trees. However, what it turned out they actually liked was being overpowered. The class lost a lot of its luster for the fair weather DKs when that was no longer the case. I think the class design benefited from retaining the DKs who actually liked the DK class. Some of the changes (which I realize are controversial), such as DW Frost and Blood tanking and even the rune mechanic change came out of feedback from those players who stuck with it. So in some sense, we wouldn't be as excited about our Cataclysm changes for the DK if they hadn't been nerfed, because some of their issues were just being masked.

avatar
Blue Poster
Target Source
#123 - 2010/04/15 08:47:47 PM
Q u o t e:
I imagine that one conclusion you already reached is that Bears are boring to play, which is the reason for giving them more buttons to push in Cataclysm.


I've had this discussion a lot with players in this forum, so I'm not super excited about rehashing the whole thing again, but one of the things I said then is something I still believe: No matter how horribly unfun a class is to play (and I don't really think that's true of bears, but stick with me), if it had a massive advantage tanking, a lot of guilds would still be using them.

I've never bought the argument that bears have been OP for some time, but nobody wants to play them. Given the level at which the community obsesses over tank balance, every guild would have a bear MT if they were that good. They might be cursing us while doing it, but they'd do it.

We know (and so we have to be very careful about it) that players will do monumentally unfun or even self-destructive things if it conveys even a small advantage.

avatar
Blue Poster
Target Source
#172 - 2010/04/16 11:00:17 PM
Q u o t e:
Without knowing goals, it's impossible to really evaluate performance. People often bring up "bring the player, not the class", but I'm not sure that was ever really the goal for WotLK for tanking.


The main goal was to let a group use any of the four tanking classes (properly geared, etc.) as a MT and for as many OTs as they needed. Despite many of the things we said, I think there was still a perception going into LK among many players that warriors were still the real tank and the other three were the niche tanks or what you used when you couldn't find a warrior. I think most players are convinced now that whatever tanks they bring they'll probably do okay (assuming they are equipped to handle the encounters in general).

There will still be players who blame any mistake on their part to it being the class and not them. Likewise, there are players just on the edge of extremely difficult content for whom a 1% difference in tank performance might push them over the edge. That is by far a minority of raiders though.

We'd prefer to see groups not bring two of the same tank class, but for the most part they don't do that, and we'd be unlikely to come down hard on groups that did.

Overall, our philosophy is that the challenge of beating an encounter should be what you do *inside* the dungeon. The only considerations relegated to before you zone in should be things like making sure you have good gear and talent specs and appropriate consumables. The challenge of beating the encounter should NOT be building the perfect raid comp and it certainly shouldn't be the decision you made months or years ago when you first rolled up that character. Raiding should be about strategy and tactics on the battlefield, not logistics long before the battle starts.

Q u o t e:
It is no coincidence when the EH numbers are in plain sight, not obscured, and not even up for debate. The EH imbalance was approaching 20% including Ardent Defender. THe overwhelming bulk of posters wanted nerfs to Ardent Defender. I was one of a very small few who saw that if AD were removed altogether that the EH imbalance would remain, and that the increasing amount of stamina on gear in future tiers would exacerbate the situation.


Here is the standard GC caveat that EH is a concept used by the community and not something that shows up on your character sheet. It is better used to help you determine how to gear your character than it is as an indicator of how awesome a particular class is. When used strictly as health and mitigation, it can be a helpful construct. The numbers get less realistic, in our opinion, when you try to model cooldowns as part of EH. Part of the conceit (for want of a better word, though it sounds like too much of a pejorative) of modeling EH is that avoidance is useless as a tanking stat since it doesn't factor into the calculation. Avoidance is, by far, not useless. It's not dependable, which is a big difference.

avatar
Blue Poster
Target Source
#188 - 2010/04/17 02:57:25 AM
Q u o t e:
But you consistently don't address the fact that even taking into account cooldowns and avoidance, some tanks consistently perform worse than others.


But that is rarely discussed on these forums. What is usually discussed is some variant of "I tried to come with a number for tankiness and it says that my class is terrible compared to another guy" or "This one day I tried to tank and it didn't go so well and therefore it must be my class." I'm exaggerating a little bit, but often the arguments are theoretical (i.e. on paper) or anecdotal (i.e. the evidence may be true but doesn't warrant a broader conclusion). Empirical evidence is quite hard to gather on things like tanking, because it's very difficult to determine the precise point of success or failure on an encounter. I get that that creates a big burden on players to "prove" anything tanking related, but it also means you should be very careful discussing theories or opinions as self-evident facts.

Actual evidence of "consistent performance problems" is something we very much pay attention to. Unfortunately, it is very hard to document.

Q u o t e:
A fair interpretation would allow for "favor someone over someone else" to mean "non-trivially favor someone over someone else".


There's no measurement for that though. "Non-trivial" is very much open to interpretation. This isn't a statistical test where you can apply generally accepted rules for significance or not. What more typically happens is a tank has trouble on a boss, looks at forum discussion, and concludes that if he or she was a different class that they would have won (or alternatively, they won barely and would have won cleanly with a different tank). No doubt this is true... sometimes. Many other times there are other factors that had a far bigger effect on whether you succeeded (or how cleanly you succeeded).

Q u o t e:
Blizzard works very hard to ensure that all there public releases are put forward with a great deal of care and editing. Ghostcrawler takes the time to post on these forums out of his own free time. I'd rather he be given more latitude by ourselves as the public to allow him to write and converse colloquially so that he can spend less time crafting / editing and more to time conversing. We get more from understanding the developers' point of views and they get more in the ability to help foster a discussion which brings them the information they are seeking out of these forums.

If this was a press release you'd be right. Since it's a community forum . . . give the man a break.


Yeah, this is worth remembering. I post mostly on my own time. The more care I take with each individual post, the fewer you're going to get overall. I don't just spew text all over the page, but I don't have the time to perfectly wordsmith every line. I exaggerate sometimes like most posters do, but when I do, I'm usually trying to get across an idea in shorthand so I can move on to something else.