Armory Class Representation Numbers

[Tyveris]

Well Kalgan posted his numbers, so I figured I would post the numbers from our database. Here is what population is included:

1) This only counts a player once even if they are on two 2200+ teams.
2) The player must have a winning record on the team to prevent team selling members from counting.
3) Players are counted from all across the world.

Legend:
First Number - Total number of players in that class.
Second Number - Percentage of total for the bracket/cutoff.

I'll leave you to make your own interpretations of the numbers.

AJ Numbers: Players on 2200+ Teams

Class 2v2 3v3 5v5
Druid 729 (24.5%) 502 (15.9%) 285 (7.5%)
Hunter 177 (6.0%) 189 (6.0%) 276 (7.2%)
Mage 109 (3.7%) 317 (10.0%) 435 (11.4%)
Paladin 112 (3.8%) 170 (5.4%) 516 (13.5%)
Priest: 298 (10.0%) 433 (13.7%) 604 (15.8%)
Rogue: 456 (15.4%) 604 (19.1%) 315 (8.2%)
Shaman: 100 (3.4%) 125 (4.0%) 396 (10.4%)
Warlock: 400 (13.5%) 312 (9.9%) 462 (12.1%)
Warrior: 589 (19.8%) 506 (16.0%) 535 (14.0%)
AJ Numbers: Players on 1850+ Teams

Class 2v2 3v3 5v5
Druid 7322 (17.0%) 4841 (12.3%) 2209 (7.0%)
Hunter 2674 (6.2%) 2522 (6.4%) 2237 (7.1%)
Mage 2424 (5.6%) 3997 (10.1%) 3434 (10.9%)
Paladin 2669 (6.2%) 3439 (8.7%) 4595 (14.6%)
Priest 5173 (12.0%) 5375 (13.6%) 4644 (14.7%)
Rogue 6903 (16.1%) 6733 (17.0%) 2551 (8.1%)
Shaman 2102 (4.9%) 2624 (6.6%) 3483 (11.1%)
Warlock 5934 (13.8%) 3998 (10.1%) 3381 (10.7%)
Warrior 7772 (18.1%) 5963 (15.1%) 4966 (15.8%)
Kalgan's Percentages: 2200+ where 100% is expected (Believed to be normalized based on class population)
http://forums.worldofwarcraft.com/thread.html?topicId=5103773379&pageNo=4&sid=1#74

Class 2v2 3v3 5v5

Druid 276.0% 184.0% 80.5%
Hunter 43.0% 50.2% 43.0%
Mage 8.7% 96.0% 96.0%
Paladin 19.7% 29.5% 147.4%
Priest 113.3% 164.8% 185.4%
Shaman 37.8% 50.4% 138.6%
Rogue 144.2% 175.1% 61.8%
Warlock 149.2% 93.2% 111.9%
Warrior 130.4% 90.7% 79.3%Kalgan's Percentages: 1850+ where 100% is expected (Believed to be normalized based on class population)
http://forums.worldofwarcraft.com/thread.html?topicId=5103773379&sid=1&pageNo=13#254

Class 2v2 3v3 5v5

Druid 184.0% 138.0% 92.0%
Hunter 50.2% 50.2% 50.2%
Mage 61.1% 87.3% 87.3%
Paladin 68.8% 88.4% 137.6%
Priest 133.9% 133.9% 154.5%
Shaman 75.6% 88.2% 138.6%
Rogue 154.5% 154.5% 82.4%
Warlock 121.2% 102.5% 102.5%
Warrior 90.7% 85.0% 85.0%

[Vexo]

post what his #'s were again, i dont wanna go through a million pages to find it

[Tyveris]

Updated the original post.

[Kaligo]

Wow you nub tyveris :) Add up the %'s for 1850+ 5's teams. It only comes to 80%.

[Tyveris]

Wow you nub tyveris :) Add up the %'s for 1850+ 5's teams. It only comes to 80%.
Thanks, fixed. Accidentally was dividing by the 3v3 total in my script :(

[Escabar]

he did it to hide the fact that warriors dominate all brackets with druids LOL and that casters get the shaft

[blackwater]

Have you seen this site they do a worldwide breakdown
http://www.realmhistory.net/arena-statistics/class-breakdown.html

[Tyveris]

Have you seen this site they do a worldwide breakdown
http://www.realmhistory.net/arena-statistics/class-breakdown.html
And those numbers appear to have a much smaller sample size. I'll trust my data as I gathered it and posted the method used to obtain it.

[Escabar]

regardless both yours tyv and realmhistory are very similar.

they show that

warriors are #1 or 2 in all brackets
Druids are #1/2 in 2v2 and 3v3 and quite high in 5v5
rogues sit up damn high around 3-4 depending on bracket

then come in slight diffs to bracket. in 2s/3s its locks/mages next, where as in 5s its priest/pallies next then shamans then lock then mage

[Tyveris]

Updated with Expected being Percentage away from 11.1% (1/9)

[Anarexea]

just checking, but kalgan's numbers divide by the population of a class as well, right?

was that an intentional difference between these stats and kalgans? or is tyveris saying that it doesnt make sense to normalize by class population?

just making sure...

[Tyveris]

I'm just posting numbers from data gathered from the armory. I did normalization based on number of classes, not entire population. I just did that because in a "balanced world", every class would be represented equally.

A few people asked for some numbers, so I figured I could post them. It looks like there is a discrepancy in balance of some classes compared to the figures he posted.

[Agonizer]

There should be a disclaimer saying that you're using a different method to get those numbers and that you can't compare your charts and his charts side by side for inaccuracies. As it is, it's quite misleading.

Kalgan specifically stated that his expected percentage was based off class population, meaning you'd have to divide the numbers in the 2nd column by ~16% if it was warriors, 16% for rogues, 7% for druids, etc...

You're dividing all of them by 11.1%

Not saying which way the numbers should be interpreted, just pointing out the inconsistency.

[Tyveris]

I just added some information to how his were calculated, but in reality, no one knows how he normalized them.

There have been a few posts comparing them to wow census data and its still not even close. He magically tweaked the numbers somehow which is why I posted the raw data so people can compare them. the Expected %s were added by request via PM just so they could see how similarly they lined up to Kalgans.

[Augnon]

Glad that u made ur own calculations.I was looking at kalagans data and there is no way in hell warlocks outnumbers warrs in 2v2

[Agonizer]

I just added some information to how his were calculated, but in reality, no one knows how he normalized them.

There have been a few posts comparing them to wow census data and its still not even close. He magically tweaked the numbers somehow which is why I posted the raw data so people can compare them. the Expected %s were added by request via PM just so they could see how similarly they lined up to Kalgans.
If you're not trying to replicate Kalgan's numbers and claim that nobody knows how Kalgan got his, then you shouldn't line up yours and his side by side.

If you're trying to replicate Kalgan's numbers, you should realize that Kalgan said "normalized for class popularity." Your numbers are normalized for equal popularity. That is a big difference.

Your explanation for Kalgan's numbers: "Kalgan's Percentages: 2200+ where 100% is expected (normalized based on total population)" is quite wrong.

[Tyveris]

If you're not trying to replicate Kalgan's numbers and claim that nobody knows how Kalgan got his, then you shouldn't line up yours and his side by side.

If you're trying to replicate Kalgan's numbers, you should realize that Kalgan said "normalized for class popularity." Your numbers are normalized for equal popularity. That is a big difference.

Your explanation for Kalgan's numbers: "Kalgan's Percentages: 2200+ where 100% is expected (normalized based on total population)" is quite wrong.
Kalgan's numbers can't be replicated so trying to replicate them is pointless. I added his numbers so people can look at all the data available considering class balance and make up their own opinions.

Why is that quite wrong? I mean total class population, I'll rephrase it as that sounds more clear.

[Daseman]

kalgan clearly stated that his stats are normalized by actual population, your numbers are normalized based on the assumption that all classes have an equal number of players - thus making these charts even more meaningless than kalgans...

[Tyveris]

kalgan clearly stated that his stats are normalized by actual population, your numbers are normalized based on the assumption that all classes have an equal number of players - thus making these charts even more meaningless than kalgans...
I give you the exact numbers of all classes. Then I show what % that is, then I normalize them. That gives you every type of information I can give you.

I could just as easily remove those percentages if that somehow makes it more useful. /shrug

[blackfoot]

Kalgan's numbers can't be replicated so trying to replicate them is pointless. I added his numbers so people can look at all the data available considering class balance and make up their own opinions.

Why is that quite wrong? I mean total class population, I'll rephrase it as that sounds more clear.

rather than divide each by 11.11%, since the populations are not equal I would be more interested if you divide it by the actual % of players of that class, because it would be more relevent.

For example, imagine there are 100 players, lets say 20 are warriors and 80 are druids

lets say 1/2 of each of those classes don't even arena, while the rest do and have 1700+ rating.

If there are only 10 warriors doing arena and 5 of them get to 2200+ that would mean warriors had a great chance to be high level. Furthermore if you had 40 druids and 10 of them got to 2200, that would mean druids don't have as good as a chance to hit 2200+ as a warrior, since only 1/4 druids made it that far, while 1/2 the warriors did.

If you divided both the 5 warriors and 10 druids by the 11.11% then you would be thinking, oh man druids are way overrepresented and thus OP, which would be false.

Since so many players play warriors, you must account for this.

[Tyveris]

I took the normalization out so that no one will think I'm trying to mislead anyone. I'm just posting the raw stats so that people have other numbers to look at and ti was requested quite a few times in the other thread about Kalgan's numbers.

As for normalizing the top end based on the total population, does that even make sense? There were quite a few good arguments in the other thread of why that should not be done. People reroll classes in order to be competitive.

Anyways, its up to you to infer what you want from either set of stats, I just figure the more information people have, the better it is.

[Agonizer]

I give you the exact numbers of all classes. Then I show what % that is, then I normalize them. That gives you every type of information I can give you.

I could just as easily remove those percentages if that somehow makes it more useful. /shrug
Let's just say that if you add to the top of your post a paragraph like this:

Our numbers are normalized based on equal representation (11.1% of every class), Kalgan's numbers are normalized based on class popularity (17% warriors, 12% priests, 9% druids, etc...)

It would clear up some confusions and greatly reduce the number of illogical deductions like these ones:
Glad that u made ur own calculations.I was looking at kalagans data and there is no way in hell warlocks outnumbers warrs in 2v2
he did it to hide the fact that warriors dominate all brackets with druids LOL and that casters get the shaft

[Agonizer]

Sorry to double post, I didn't want to clog up my last post :(
As for normalizing the top end based on the total population, does that even make sense? There were quite a few good arguments in the other thread of why that should not be done. People reroll classes in order to be competitive.
My opinion is that we should use equal class representation for the very top end of arena (like you did), while for lower end mark (1850+, for example) we should take class popularity into account.

[Augnon]

What do u not see were locks supposedly out number warrs in 2v2???


Class 2v2 3v3 5v5

Druid 276.0% 184.0% 80.5%
Hunter 43.0% 50.2% 43.0%
Mage 8.7% 96.0% 96.0%
Paladin 19.7% 29.5% 147.4%
Priest 113.3% 164.8% 185.4%
Shaman 37.8% 50.4% 138.6%
Rogue 144.2% 175.1% 61.8%
Warlock 149.2% 93.2% 111.9%
Warrior 130.4% 90.7% 79.3%

[Tyveris]

Easier thing to do is just remove the normalization because I'm really not trying to imply anything and no matter how I word it, people will think I am. I just have access to a database filled with information that I figured inquisitive people much like yourself would love to use and crunch.

Am I personally surprised at some of the numbers? Sure.
Do I think Kalgan is lying? Of course not.

[Agonizer]

What do u not see were locks supposedly out number warrs in 2v2???
Our raw data for 2v2 is:

Warlock: 13.5%
Warrior: 19.8%

Let's say that of all the arena players, 9% are warlocks, 15% are warriors.

13.5 / 9 = 150%
19.8 / 15 = 132%

[Tyveris]

Sorry to double post, I didn't want to clog up my last post :(

My opinion is that we should use equal class representation for the very top end of arena (like you did), while for lower end mark (1850+, for example) we should take class popularity into account.
Yeah, I would agree, but I'll leave the main post as so that people don't make assumptions about what I'm trying to imply. We can certainly try those methods to see what it comes to.

[blackfoot]

Kalgan's Percentages: 2200+ where 100% is expected (Believed to be normalized based on class population)
http://forums.worldofwarcraft.com/thread.html?topicId=5103773379&pageNo=4&sid=1#74


Class 5v5

Druid 80.5%
Hunter 43.0%
Mage 96.0%
Paladin 147.4%
Priest 185.4%
Shaman 138.6%
Rogue 61.8%
Warlock 111.9%
Warrior 79.3%Regarding 5's
If anyone is interested I decided to normalize the numbers myself by taking the actual number of players 1850+ and finding what percent of those got to 2200+ and then finding out what that number is compared to all the rest of the people who got 2200+.

Doing so I got these numbers, where 100 percent should be where its balanced.

Druid 95%
Hunter 99%
Mage 96.3%
Paladin 108.9%
Priest 93.6%
Shaman 107.1%
Rogue 99%
Warlock 89.1%
Warrior 112.5%

You'll notice that the variations aren't that large and that would be because you can imagine that whatever the game balance is that allows a certain class to get to 1850, it will help them get to 2200+ too.

[Tigs]

Best way express the numbers would probably be to find the total number of players in the bracket and divide it by 9. You now have the amount of players who should be playing if all classes were equally represented. Now, take the actual number and divide it by this number, then subtract 1 and multiply by 100. If you didn't follow that, it's:

((Real / Expected) - 1) * 100%.

For example, if the expected value for 2v2 is 500 players (I.E. 3500 Total / 9 Classes), and Mages have a 100 player representation, the value given for them would be -80%, or 80% below norm. A perfect representation would have a 0% number.

Call it the "How fucked am I" rating. If you're 0, you're not fucked at all. If you're below or above, you get to bitch / get bitched at, respectively =p.

I have a test in a few hours. When I get back I'll take care of this, unless someone (<3) does it for me.

To go even further, take the lowest value you get below zero and scale that to -100%. Scale the other negative numbers to that, and you've now got the "most fucked" class at -100% and the other classes numbers are how fucked they are compared to them. Do the same for the positive classes, to 100%. Congratulations, Warriors are the definition of overpowered, and Warlocks are only 50% as overpowered! (EXAMPLE ONLY).

I can't believe I actually learned something in Economics.

[Grimgore]

I'm just posting numbers from data gathered from the armory. I did normalization based on number of classes, not entire population. I just did that because in a "balanced world", every class would be represented equally

Well, you have to normalise in some way. If x class outnumbers y class it is pretty damn obvious there will by more of x class at high ratings.
You're saying people reroll based on arenas, in my experience this would only account for a very small %. Many people play warrior because they want a badass looking character. And regardless if they do think that warriors are OP in 5v5 the normalisation takes this into account.

Currently there is no way of accuratly normalising as you have to take to many factors into account for a correct normalisation (gear, specc, setup population etc). However this does not mean that not normalising is the way to go. Take hunters for example, ridiculously overpowered in all brackets (like warriors) if playing with the correct setup, yet the numbers indicate them to be underpowered.

[Escabar]

the fact is kalgans method of doing it shows how many of each class PvP. that is all

so you cant take kalgans #s and compare them class to class.

all when looking at kalgans normalization #s

all you see is that of the lock population a bigger % of the lock population pvp then the % of warriors pvp, this does not state that locks out # warriors in pvp. (as the warrior population is about 1.5x the size of the lock)

hence Kalgans #s are purely retarded.

all its good for is seeing what classes are being rolled for what

[Exist]

Thank you for posting the RAW data Tyveris. These numbers are definitely in line with everyone's arena experience.

[erich]

I calculated for the player population based on the numbers provided by both. Basically I divided Tyveris's #s by Kalgan's #s.

I assumed that Kalgan's magical 100% representation means for every class, the same % is expected to be 1850+/2.2k+ in arena. eg. if there are 10 1850+ warriors and 100 warriors in wow, then there should be 5 1850+ priests if there are 50 priests in wow.

5's data
Class 2200 1850
Druid 8.83 7.53
Hunter 16.02 13.98
Mage 11.31 12.34
Paladin 8.73 10.47
Priest 8.13 9.43
Rogue 12.72 9.71
Shaman 7.13 7.88
Warlock 10.30 10.34
Warrior 16.83 18.32

You can read the table as, from 1850+ 5s data, in a pool of 100 wow players
7.53% are druids
13.98% are hunters
12.34% are mages
etc.


From the wow census site, it seems population is pretty evenly spread among lv 70 players. Shamans are the low at 8%. Warriors/Mages/Hunters are the high at 13%. This is a big difference from the above table in which there are 17/18% warriors. Either there at 17/18 warriors out of every 100 wow players or Kalgan is biased towards warriors, expectin them to do better in 5s than other classes. (Remember I am not talking about the actual representation, I am discussing Kalgan's expectations for representation).


These numbers are reflected in data from other brackets.

3s.
%(out of 100)
Class 2200 1850
Druid 8.26 8.93
Hunter 11.39 12.80
Mage 9.99 11.66
Paladin 17.44 9.91
Priest 7.95 10.22
Rogue 10.44 11.10
Shaman 7.51 7.58
Warlock 10.13 9.93
Warrior 16.88 17.87

2s
%(out of 100)
Class 2200 1850
Druid 6.50 8.85
Hunter 10.14 11.84
Mage 30.85 8.82
Paladin 14.00 8.63
Priest 6.48 8.59
Rogue 7.79 20.30
Shaman 6.51 3.03
Warlock 6.60 10.89
Warrior 11.12 19.05

The 2200 2v2 data should probably be disregarded, mages are unusually high, and it skews all the other numbers. Perhaps because of a time difference there are a lot more mages in high ranked teams.