*I posted this over on Dio's, but I'm interested in getting feedback from people here who don't post over there.

Seems like a lot of people have been asking what kinds of classes are good against everyone lately. I think the idea of sticking to a type solely for pk-worthiness is boring. Another thing that bothered me was perception versus numbers because a lot of what people tell others to play seems to be based on opinion. The mantra of "well, it didn't work for me, so don't bother trying it" seems to have become dogma to a lot of folk, but I argue that some of the best characters have chosen race/class combos with an outside of the box perspective. So, I took the data available in the pk stats section (which are all attached to PBFs) and calculated averages based on the midrange players*. My assumption is that PK ratio is going to represent the best overall players to the worst as far as skill and effort put into a character, therefore the middle of the range is going to represent the averages the best. I've included the math and my sample size below the tables.

1 - 10 Scales Ranking (green best, yellow worst)


Raw Numbers (green best, yellow worst)


*Technical part:

To get the figures above, I sorted the class selection by "Ratio", greatest to least. If there were less than 100 PBFs for a class, I took a sample size of 10 from the middle and excluded characters created before the gank-o-meter started tracking group sizes. If there were greater than 100 PBFs, I took 10% of the total number and calculated from the middle.

Totals, Averages and sample range:
Anti-Paladin (74: 32-41)
Assassin (148: 67 - 81)
Bard (123: 56-67)
Berserker (35: 14-16, 19-21, 23, 25-27)
Conjurer (55: 23-32)
Druid (56: 25-34)
Healer (99: 45-46, 48-50, 53-57)
Invoker (104: 48-52, 54-56, 58, 60)
Necro (61: 27-36)
Paladin (135: 62-63, 65-73, 75, 77)
Ranger (86: 39-48)
Shaman (100: 46-54, 56)
Shifter (147: 67-71, 73-79, 81-83)
Thief (144: 66-79)
Transmuter (84: 38-46, 48)
Warrior (647: 291-355)

To get a 1-10 scale, I took the range of a given category (lowest number and greatest number), the average of all the numbers in the category, plotted points at 1, 5, and 10, and then calculated a quadratic regression curve through the points. So, the best for any category is 10.0 and the worst is 1.0, with the average of the numbers being equal to 5. The formulae for each is below.

Formulas:
pk/hr = (-7.309*(x^2)) + 29.7659x + 0.2193
pk Win = (-0.0002552115*(x^2)) + 0.1394929873x + 0.2678560592
pk Loss = (0.00655*(x^2)) - 0.64924x + 16.74914
avg kill group size (solo) = (11.8187*(x^2)) - 55.6521x + 66.4775
avg kill group size (support) = (-6.43076*(x^2)) + 34.69306x -36.2598
avg death group size = (55.8171987*(x^2)) - 193.0751788x + 167.603349
mob deaths = (0.017361*(x^2)) - 1.185173x + 18.996944
hours = (0.0001249*(x^2)) - 0.0031453x -1.1426302
Popularity = (-0.0000522*(x^2)) + 0.0502771x -0.6958059

gr

Graph 1 is a 10 point scale ranking of each class in each category.
Graph 2 is the raw data from each class.

All of the numbers were calculated using the PK Stats on Dio's. Instead of using all of the data, I sorted it on PK-ratio (percentage of pk wins divided by total wins/losses), and took 10% of the total PBFs from the middle. For example, if there were 100 PBFs for warriors, I averaged from #s 46 to 55.

My theory is that the average player is going to be the most representative of what to expect from playing a particular class.

Based on the chart, if I wanted a character who could solo pk, win more fights than I lose, explore, and survive against a gank squad, I'd choose to play a shaman. The shaman ranks better than average in pk/hr (6.4), total kills (6.3) and pk ratio scores (6.9), the best score in group size per kill (10.0), and a low amount of mob deaths score (8.9, meaning fewer mob deaths). Simple enough?

People have pointed out that there isn't enough breakdown of the data, especially when related to alignment, but the chart should give enough of an idea based on the averages already calculated.


- Baron

I can't figure out what mid-range is exactly...

Also, AP data is horribly skewed by the fact that only those APs that get a decent weapon going in the midranges (and therefore kill a lot of people and die very little) are likely to go on to hero and buy a PBF. The other data is similarly skewed (selection bias), but not as badly.

It would be better if some of our IMMs could generate data like this based on all characters to get a better sense of certain things (and to publish to various purposes), but that probably requires a level of coding skill and time not currently available.

Especially because I don't have a way to access any other numbers. However, one way to interpret the data is to see it as people's best effort. So, even a newer player would publish a PBF because they want to see how far along they've come along. I think the data would be more confusing if you took everyone's stats and put it into a table like this. For instance, I've had characters I've leveled to hero in 50 hours for the sole purpose of exploring the class, an area, to see if I could win a role contest, or on a bet. I wouldn't want those PBFs because they'd represent me not trying very had to play the best way I know how. I do see your argument and I think it's valid, but I don't think there's a way to present the data in a way that's the most representative of everyone's true ability.

At least with the PBF data, even if someone gets to the midranks and level-sits, they're testing the limits of the class. Feels like a better stress-test to me and I'd like to have those numbers in the table instead of Joe Schmo ranking a throwaway shifter to test the latest new form.

- Baron

I think you need to break it down by alignment to make sense of what's going on. I don't think it's an accident that evil-only classes have the highest pk ratios. It'd be interesting to see what evil warrior does vs evil conjurer vs antipaladin vs nec.