In the beginning, there were numbers:
# of Class per Top 20 World Teams
2v2
Warrior: 5
Paladin: 2
Hunter: 1
Rogue: 2
Priest: 6
Shaman: 2
Mage: 1
Warlock: 12
Druid: 9
3v3
Warrior: 15
Paladin: 10
Hunter: 0
Rogue: 3
Priest: 9
Shaman: 5
Mage: 5
Warlock: 10
Druid: 4
5v5
Warrior: 24
Paladin: 25
Hunter: 6
Rogue: 0
Priest: 21
Shaman: 18
Mage: 18
Warlock: 8
Druid: 5
Overall (Weighted, average = (#2v2/2)+(#3v3/3)+etc)
Warrior: 12.30
Warlock: 10.93
Priest: 10.20
Paladin: 9.33
Druid: 6.83
Shaman: 6.27
Mage: 5.77
Rogue: 2.00
Hunter: 1.70
Spec Diversity (% of top 20 overall each class off spec)
Warrior: 20% (80% 35/23 variant)
Warlock: 40% (60% UA variant)
Paladin: 10% (90% Holy)
Priest: 25% (75% 28/33)
Druid: 45% (55% Resto)
Shaman: 30% (70% Elemental)
Mage: 5% (95 % Water Elemental, 17/0/44 itself being 55%)
Rogue: 45% (55% Combat)
Hunter: 20% (80% Marksmanship)
Repeated for Top 50 Teams (As Per Request)
2v2
Warrior: 11
Paladin: 9
Hunter: 1
Rogue: 6
Priest: 14
Shaman: 6
Mage: 4
Warlock: 29
Druid: 18
3v3
Warrior: 27
Paladin: 22
Hunter: 0
Rogue: 13
Priest: 20
Shaman: 10
Mage: 12
Warlock: 25
Druid: 11
5v5
Warrior: 54
Paladin: 57
Hunter: 12
Rogue: 4
Priest: 47
Shaman: 44
Mage: 40
Warlock: 23
Druid: 8
Overall (Weighted, average = (#2v2/2)+(#3v3/3)+etc)
Warlock: 27.43
Warrior: 25.30
Priest: 23.07
Paladin: 23.23
Shaman: 15.13
Druid: 14.27
Mage: 14
Rogue: 8.13
Hunter: 2.9
I think the second to last table is the most interesting; it gives a rough idea of how the respective classes are performing overall in arena. Ideally and mostly simply, these should all converge to somewhere near 7.25% per class. The last table compares the general talent spec singularity between the top 20 players of each class -- I was definitely surprised, but one should take the numbers in this table particularly as being potentially biased due to interpretation of "same spec" and extremely small sample size. Other errors of course arise from players being listed with not their current arena specs etc.
I've been reading through a lot of fantasy class changes, balance suggestions, and ways to improve WoW PvP on the large blogs this week and have found myself in strangely significant disagreement with some of the points being made. I'll let the numbers talk for me while I collect my thoughts into something more cogent to say. (And read Harry Potter ^^)
** Updated with Top 50 Tables ** (and after another 200 pages of harry potter, maybe I'll add some more)
Subscribe to:
Post Comments (Atom)
52 comments:
Bit surprised that Mages take the 3rd spot from the bottom, but the Rogue total just makes me sad. :\
Oh and - first etc.
Interesting. Not what I expected.
So according to this warlocks are ten times more popular than hunters on high rated teams?
I would have really thought mages would fair better, why do you think we do so poorly as a class on these lists?
harry potter is gay radi
Snape killed mages.
I think its because mages have to compete with warlocks for their spots
Christ, hunters need help, and they need it badly!
Hell, just had to comment on the warlock situation in 2s.
It's getting beyond ridicilous, seriously. You pop up the first armory page of your
battlegroup(top20 teams). What do you see? Warlock + random, or better yet, warlock + healer.
You either team up with a lock, roll one, or die. That's the way it is folks.
Nothing too surprising on these lists. Cool info though thx^^
3 dps classes at the bottom? no way
Would you be willing to post a list of potential changes similar to what Dahis put up on his site? I\'m curious to hear your thoughts on how to address this situation.
My (admittedly terrible) 3v3 team (Rogue, Shaman, Warlock) seems get totally wrecked if there is a BM Hunter on the other team. Granted, we're all in PvE specs, and only play 10 games a week, but I'm surprised at the complete lack of Hunters in the 3v3 bracket.
Is it because those Hunters that do amazing against people in Blues, drop off a cliff when the Gear and Skill increases?
Deadzone, pillars/LoS, and not many ways of a Hunter escaping Melee unless they're BM spec(18sec blow your load then you're screwed spec) are some of the reasons why Hunters aren't represented much in top 20 imo.
Tea,
We always seem to fight them in Lorderon (or however it's spelled), and they are always BM it seems. There strat seems to be go full out for the Warlock (who doesn't have great PvP gear, very low on STA and Resiliance for my liking). If the Shaman doesn't get his NS + HW off on him, he's dead before the Beastial Wrath is over. Then it's 2v3, and the fact he's BM makes no difference because eventually (but not always), they'll win the numbers game.
I think part of the problem is we did very little PvP in Season 1, and pretty much everyone we are fighting even as low as 1550s have lots of PvP (Arena and Honor) gear. That really surpised me.
This blog is great, thank you for so much good Informations.
Wex
Raddy, the numbers make my head asplode.
And about the warlocks... it's pretty much a skill less class. Dot shit, run around, fear, etc, repeat. That's why Cody plays one.
hey raddy i would like to ask if im playing in a warrior, hunter, holy priest, holy paladin and mage combination, would it be viable for me to go 17/44 or 7/54 instead of a 33/28 variant. i know the strategy revolves around chain sheep and cc and freeing up my priest and hunter to drain their mana but that being said i feel that the burst damage from 33/28 is too low given the high resilience everyone is running. if i do go water elemental spec would you shed some light on how i might play this (water elemental) spec in a hunter team?
Just finished Harry potter; it's awesome.
The problem with WOW players and statistics isn't that the stats they collect are wrong - it's the interpretations.
Blizzard have stated that they balance classes around 5v5 (whether or not that's a good way to go about balance is a whole arguement in itself).
But in any case if you want to compare classes you should be doing it seperately with each bracket. By combining the stats for all three brackets what you're basically saying is that a class is equal across the three brackets (which is obviously wrong). Look at warlocks, in 2v2 they're massively over represented but in the other brackets they're not so. Because of this, it skews their power in the other 2 brackets (3s and 5s) giving the misinterpretation that locks are all powerful in 2s, 3s AND 5s when in fact it's just in 2s.
And then of course the amount of players that play each class should be taken into account. It's no point doing these stats if (as an exxageration) there's only one hunter played in the whole WOW universe.
The problem is people with no experience in stats (and perhaps maths as well) making grossly statistically incorrect interpretations. People study years on stats, don't think that common sense will get you thru.
- K
Just wondering...how do these stats deal with substitute players? Could a class be double counted?
K,
While I do agree with you, it's not like he just gives the last table without showing how the numbers are derived. You can see the class makeup per bracket, and I assume Rad just trusts that people can make their own educated opinion about them.
Also, I do agree with you about populations, as we all know Druids are vastly underrepresented in the general population, so it might explain their numbers. But really, it doesn't take a math wizard to see the 2v2 numbers and the over representation of Warlocks and think something is amuk at the Circle K. ;)
Well yeah i never disagreed with the data (nor am i necessarily disagreeing with your view on locks in 2v2), it's the interpretation of it and the other variables (errors) which no one seems to take into account. eg:
1. class population
2. analyzing each bracket independently (if a class is good in 2v2 it doesn't transend that it will be good in 3v3 etc)
3. a third correlation (eg are wars only good with paladins?)
4. stability (are these numbers stable over an arena season? or do they fluctuate dramatically)
5. derivation of sample (the top 20/50 is relative and independent of the BG ie. 2300 rating in BG1 might not equal a 2300 rating in BG2)
The way in which the numbers are 'weighted' is also a bit weird (even though the whole notion of combining the date across 3 brackets is flawed anyway). I'm not certain, but i think it's a (poor) attempt to normalize the data. The way in which they're weighted gives far too much emphasis (and power) on the 2v2 bracket.
I'm not going to pretend that i'm the be all and know all about stats. But i can safely say that unless you've got some kind of solid education in analyzing stats any 'educated opinion' will most probably be incorrect.
And no rogues in the top 20 5v5 *cries*
- K
A more thorough analysis is certainly needed if you want to really trust the stats for sure. I merely provided some numbers for entertainment and to point out a few of the extreme abnormalities. (Bottom 2 vs top 2)
I believe honestly that you can't balance 5v5 without balancing the smaller arenas first, and I partially feel (as Ming does) that this hinges on balancing 1v1 gameplay. At the same time, balancing 2v2 and expecting it to scale up to 5v5 is simply not going to happen.
I will say that small changes are often all that is needed to change the relative balance -- huge class overhauls are just going to tip the scales in the other direction.
Hunters and mages are similar in the sense that if arena was a big open field, I don't think anyone would complain about their viability. Line of sight punishes both classes pretty severely. As for rogues, all they bring to 5s is damage, and the abundance of CC and snares make their damage really not better than the other DPS classes. Imp kidney is a great and expose is nice in theory, but as it stands an elemental shaman is a better option as far as damage goes for most teams as he brings more overall damage through totems and heroism and greater versatility. True that rogue can really shut someone down and that is really the strength of rogues in the smaller brackets, but other classes CC options are generally just as strong.
rad, are you going to being recording any duels this test server?
I sadly haven't really had much time to be on test and won't really have much time today to be on either. I'm not really sure the changes warrant another dueling video, and there's no way without a long test build I'd be able to get the opponents from the previous vids. I keep meaning to start frapsing 5s and 3s, but we've never really been able to play more than 2-3 games at a time for a while now.
Okay, enough /emo. Short answer, probably not. ^^
How is it that you time to post on here but not have time to play? =p
Yeah, really surprised about mages. When you say weighted ayerage, you mean weighted according to class population? If so, where did you get your class population data?
Your method for weighted averages across all brackets is horrid. It makes a single player on a 2v2 team (lowest point value) 2.5 times more valuable than a player on a 5v5 team (highest point value).
If anything success in the larger bracket should be weighted to favor larger team compositions instead of smaller ones.
There is no 'right' way to weigh the arena brackets for averages and I see nothing wrong with how it is done here.
You're right, there is no correct way to weigh an average. Weighting is effectively arbitrary and is determined by the person running the numbers.
However there are ways to weight averages so the information is actually relevant and useful and there are ways to weigh data so it's misleading. For example I'm going to weigh the same top 50 results based on the class's position in the alphabet.
Druid: 9.25
Hunter: 1.625
Mage: 4.308
Paladin: 5.5
Priest: 5.0625
Rogue: 1.278
Shaman: 3.158
Warlock: 3.348
Warrior: 4
Since there is no 'right' way to weigh averages this data is equally relevant. I'm sure we can agree that this method of weighing these values is ridiculous and the constant generated by this method tells us nothing about class balance.
What I was getting at is weighing the participation against class population, the total number of players within each bracket or points awarded per bracket tells you more about what's happening across all arenas than using a weight method that makes 2v2 2.5 times as "valuable" as 5v5.
Totally agree with the person posted above me. Another example, for the top 20 I've weighted the following classes according to how blizzard values the bracket (70% for 2s, 80% for 3s and 100% for 5s) and combined them. Once again I don't think combining brackets is a good idea but as an example anyway:
Warlock 24.4
Warrior 39.5
Mage 22.7
Rogue 3.8
- K
I would also suggest adjusting for class population; data can come from 70s on warcraftrealms.com. I might post my own analysis of this later on. -Lumah
Warrior: (11*0.7 + 27*0.8 + 54) = 83.3
Paladin: (9*0.7 + 22*0.8 + 57) = 80.9
Hunter: (0.7 + 0 + 12) = 12.7
Rogue (6*0.7 + 13 + 4) = 21.2
Priest (14*0.7 + 20*0.8 + 47) = 72.8
Shaman (6*0.7 + 8 + 44) = 56.2
Mage (4*0.7 + 12*0.8 + 40) = 52.4
Warlock (29*0.7 + 25*0.8 + 23) = 63.3
Druid (18*0.7 + 11*0.8 + 8) = 29.4
Adjusted for population (by using diyision, smaller populations are buffed and larger ones are nerfed slightly, factoring in the fact that if there are twice the warriors running around than shaman, and there are, there are likely to be twice as many warriors on top teams if the classes are perfectly balanced. Data taken from all leyel 70s spidered by warcraftrealms on all seryers):
Warrior: [83.3 / (275414/1908979)]=577.4
Paladin: (80.9 / 166951/1908979)=925.0
Hunter: (12.7 / 249450/1908979)=97.2
Rogue: (21.2 / 223278/1908979)=181.3
Priest: (72.8 / 214201/1908979)=648.8
Shaman: (56.2 / 123299/1908979)=870.1
Mage: (52.4 / 266905/1908979)=374.8
Warlock: (63.3 / 207717/1908979)=581.7
Druid: (29.4 / 181764/1908979)=308.7
Total: exactly 4565
So class rank percentage:
Paladin: 20.3%
Shaman: 19.1%
Priest: 14.2%
Warlock: 12.7%
Warrior: 12.6%
Mage: 8.2%
Druid: 6.8%
Rogue: 4.0%
Hunter: 2.1%
As there are 9 classes, if the classes were perfectly balanced then after adjusting for population each rank should equal 11.1%.
Feel free to post this Raddy.
-Lumah
Lumah, thanks for that info, I was looking for that.
Nice calculations bertrand/lumah. From his/her analysis you can see the correlations (team success v class) are far different to those that were originally posted.
Though as an aside, a correlation doesn't show a causation (paladins might be the most positively correlate of team success but it doesn't mean that having paladins is the cause of team success)
- K
Personally I think some people are reading too much into Raddy's weighted values, wanting it to provide some universal truth rather the interesting overview that I am sure was the intention.
Ultimately each person is going to weight the values differently, for example I would rather see mage perform well in 2v2 and 3v3 (than in 5v5) where it is far more exciting to play, less about CC & OOM and more about CC & pew pew. As such I would actually give the 2v2 value a HIGHER weighting rather than 70% of its value as numerous posters above have done.
Likewise in 5v5 the limitations of a class can be more readily offset by its 4 partners, and this maybe gives an unfair indication of that classes power/popularity. e.g. initially mages were far more popular because of water, but the addition of water to arenas has had an effect on mage popularity disproportionate to the change in class power.
Likewise if rogues had a nice group buff (say 5% to dodge?) then their popularity would probably go up even though their 1v1 power had not changed much.
What the figures give us is an interesting overview of class popularity at the top in each arena bracket. Allowing for population is definitely worthwhile but the weightings are up to the individual as the 70%/80% by blizzard is purely a points-based % not a class value based % etc.
If you read the post about weighting according to alphabetical order you'll come to realise that weighting/normalising can be done however you like but unless you can rationalise it, it'll be meaningless and or people will think there is an intended bias (and if you've heard about scientific experiments in the past you'll know how much of a problem bias can play on interpreting data)
Also if you read previous posts none of them (the latter posts anyway, unlike the first few comments) are saying that the data provides a universal truth (as you put it), but rather it is subjective and relative to how you interpret and manipulate the data. There is no essential cookie cutter way to go about analyzing these things but there should be a certain accountability and rationale behind it.
Sure all this number crunching might be a bit much for some people but i don't see any point in interpreting/normalizing data unless you do it with proper reasoning (unless you intend to create a bias to prove some point).
but imo, none of this data means anything definitive to me. There's too many variables (for starters there's 8 variables in the 8 other classes, not to mention races) to accurately interpret let alone attempt to solve class balance thru data analysis. For example, are there no rogues in 5v5 (top 20) because a) rogues needs buffs b) warriors need nerfs c) war/pally synergy needs a nerf d) wars 'steal' all the pally's blessings e) rogues are all played by 12 year olds (jk, i play a rogue anyway) f) other classes are too good at snaring g) all rogues read WOM and didn't bother trying for 5v5 teams h) all of the above?
It doesn't take a brain to see that rogues aren't prevelant in 5v5 arenas and that warlocks are very prevelant in 2v2 (according to the sample) but as for why/ranking classes (their pvp power)/interpretations/solutions it'll take an extensive analysis of data (accurate) or theorycrafting (subjective) to get even a fraction of an answer.
- K
Looks like 2.3 is the end of stopcasting macros, or the vital need for them:
http://blue.cardplace.com/newcache/us/388011140.htm
LOS isn't more of a problem for hunters than it is any other dps class.
The problem is a combination of: poor damage against armored targets, no synergetic damage abilities (ex: mortal strike, wound poison), and limited CC.
To strengthen hunters, one of these areas needs a significant buff.
Heat seeking bullets that go around corners?
Hunters get hit with LoS harder than other ranged classes.
They often can't use instants when they do get LoS when playing ring around the pillar because they're too close to actually use a ranged attack. If the pillars were bigger and more square hunters wouldn't have such a problem with them.
Some good analysis here in these comments, except for one thing, which makes me gag every time I see it:
Overall class populations are NOT at all relevant to class performance when that is measured using the top 20/50 teams. That's such a small, elite sample, that it measures true class power.
Whether or not Jimmy The 8 Year Old Boy rolled a hunter cause he loves his kitty cat pet is irrelevant to those playing at the high end. If hunters were a brutally effective class, elite players would play them, and teams with them would dominate. Jimmy The 8 Year Old Boy and his kitty cat pet has no bearing whatsoever on class effectiveness in 5v5s.
If you guys believe that overall class population matters, then ask yourself why of all NBA players, 25% of them aren't Chinese, and another 25% Indian.
Also, you guys would be surprised about druids. Everyone makes claims about how they are "VASTLY UNDERREPRESENTED." This may shock you, but they are not. They're a bit under average, but it's really not out of the ordinary at all.
According to www.warcraftrealms.com, at the time of this post, druids are 10% of the overall population at level 70. With 9 classes, you'd expect 11% if everything were perfect. I think you'll all agree that this does not count as "VASTLY UNDERREPRESENTED."
Oh, and as for weighting, I can see 2 ways of doing it. Here they are, with the justification as to why you'd want to do it that way:
Method 1 - Blizzard Weightings
The justification for doing it this way is because you're measuring the PURCHASING POWER of each class. It measures how much loot each class can generate by playing in the arenas.
This isn't a bad justification, because it's a roleplaying game, where character development via loot is a gigantic portion of the game.
In this scheme, 2v2s is worth 70%, 3v3s is worth 80%, and 5v5s are worth 100%.
Method 2 - Weighting Done By Team Size.
The justification for doing it this way is to normalize things by player. This seems like the most raw type of "class power" measurement. This assumes that you're in it all for the fun and glory, and don't really care about the purchasing power of your class.
In this scheme, 2v2s would be worth 40%, 3v3s would be worth 60%, and 5v5s would be worth 100%.
Really great article. Not just for what Radikal wrote, but also for what some of the guys (K, Lumah, etc.) posted on their comments.
However, if anything is left to be said, it's that it has been statistically proven that statistics are worthless. ;)
- Kraderif
I don't think the stats are useless at all. Rogues being underpowered b/c warriors are oyerpowered, or for other reasons mentioned, will show in the stats. And I do think you haye to adjust for population; the more people are playing a certain class the greater (linearly) the chances are that they will end up in that elite group. The high schools with the top sports teams are disproportionately those that are simply big magnet schools. Why? Because with so many students the chances of you putting together a star team are much greater.
It's true that there is one thing that isn't being measured directly: the push for class diyersity. Eyen if a class is particularly powerful, the game is balanced in such a way that haying more then one of them tends to be an exponentially weaker choice. This does haye an oyerall balancing effect though, which means that the numbers shown might be eyen less balanced if a 1 on 1 arena were eyer to be made.
(response to anonymous 6 posts up)
"Hunters get hit with LoS harder than other ranged classes.
They often can't use instants when they do get LoS when playing ring around the pillar because they're too close to actually use a ranged attack. If the pillars were bigger and more square hunters wouldn't have such a problem with them."
Hunters also have an easier time catching people who pole kite.
@ Hamchook
Yes, hunters can catch up to people, that's not the problem. The problem is what does a hunter do to someone they've caught? Use raptor strike?
All their damage is balanced around them using their ranged attacks. As it stands they can't do that at the range you'll catch someone. If they changed the shape of the pillars it would help to fix that.
You can wing clip, step out of the deadzone, and get a few shots off. Also, its easy to force somebody who's pole kiting into a trap. Compared to warriors, I'd think hunters are better off in this situation.
Do you play a hunter or are you just speaking in theory?
Same anonymous as before Hamchook.
I don't play a hunter, I have one on my team though so I've seen how it plays out enough times.
Wingclip-> step away -> shoot
only works well if improved wingclip procs. By the time you've moved back they've gotten far enough around the pillar to be out of LoS again. Once you're close enough to keep LoS on somebody pole dancing they're within the dead zone. Forcing with traps isn't always easy, in an arena like blade's edge they can chose to eat 1 autoshot while switching to another pillar then the trap is on CD.
OTOH a warrior in the same position as a hunter ready to wingclip can hamstring then start unloading big white swings and any special he has rage for. He can dog the other player's steps and keep pounding away.
Pillar shape and quantity is one of the main reasons hunters have such a low representation in arenas across all brackets.
Wrong about wing clip proccing. There's enough time to get off a volley of shots if you position yourself correctly.
And no shit that a warrior standing on top of a pole kiter is in a better position, that's not pole kiting. The situation I'm speaking of is when the chaser is on the opposite side of the pole as the runner. What does a hunter do? Drop a frost trap, run in one direction and force the runner into it. Another way? Cheetah > wing clip > step back > shoot. Warriors don't have any abilities to catch up like that.
LOS is not one of the significant reasons for the lack of hunters.
Post a Comment