DPS stats : Is his really bigger than yours?

[misterP]

hello everyone, this is some copy/pasta of a thread i made on the EU forums. Enjoy.

Hi, i was a bored programmer. So I made a program that extracts stats from WWS, and applies some simple algorithms to them. Thing is i'm not really too knowledgable on stats.. Is anyone here good with stats that feels like running me through the maths of making something useful out of my data?

For those that care, it takes the top patchwerk WWS reports. it could do a different boss but patch seemed like the best to go with.

NOTE : If you dont already know, the rogue number on this is high because of a bug. Rogues aren't that OP :P

This is a readout of a short run i did - i'm restricted with how many samples i can take, because wws likes to take ages to respond to requests.

Class | average dps | number of samples | standard deviation
DEATHKNIGHT | 3829 | 264 | 688.8
DRUID | 4313 | 413 | 1342
HUNTER | 4576 | 110 | 940
MAGE | 4515 | 404 | 726
PALADIN | 4036 | 423 | 1180
PRIEST | 4503 | 501 | 1401
WARRIOR | 4279 | 239 | 1116
WARLOCK | 4201 | 383 | 646
SHAMAN | 4514 | 509 | 1931
ROGUE | 4893 | 399 | 1263

shamans are interesting - the large standard deviation suggests one spec is OP, the other is UP - ele and enh anyone?!

UPDATE #2

I decided to add some visualisation to these results - these are some graphs i knocked together in a few minutes. They give a better idea of the distribution of these numbers.

http://img357.imageshack.us/my.php?image=dpstj2.png

OMG! ... not a virus. honest!

EDIT : Title edited for dramatic appeal

[killaklown]

the point of this is? Half the crap on wws are private servers and faked stats. Should remove all Anonymous reports, then recalculate.

[misterP]

wow, 50% are faked? well i won't even bother explaining the complexity involved in outright faking a report, but certainly a private server could be annoying..

except not even nearly 50% are private servers.

85% of statistics are made up. Mine are the 15%..

EDIT : it does ignore anonymous reports..

EDIT2 : I recognise that these results aren't perfect, some results may be fixed. But lets be honest, it's going to be a small proportion of people that chose to manually edit their WWS report to increase their personal dps or whatever. Larger sample sizes (500 per class here, which isn't bad) will help to avoid that.

[killaklown]

EDIT : it does ignore anonymous reports..

Then ignore my other post

[Deleted]

Rewrite your programm and try to filter some specs.

Some of those results simply can't be true.

Take a look at a warrior. Our Titan Grip guy does a very high amount of damage. But a warrior can also be a protection one and he does something like 2k dps. Right by now your reports/statistics/algorythmics aren't valid anymore.

[misterP]

filtering by spec is difficult - wws doesn't record spec, so i'd need to write the program to look up each individual character on armory. which is difficult.

My program filters out dps below 2.5k to avoid tanks, check out the graph for more info. Some warriors do do very high damage - check the graph.

[Offhand]

So statistically, most DK's suck. This matches my empirical evidence quite accurately.

Could you set it to exclude HaT and see where rogues were?

[Jynx]

filtering by spec is difficult - wws doesn't record spec, so i'd need to write the program to look up each individual character on armory. which is difficult.

My program filters out dps below 2.5k to avoid tanks, check out the graph for more info. Some warriors do do very high damage - check the graph.

Filter by ability usage. It's easy to tell which is which then.

Tanks do over 2.5k on Patchwerk pretty routinely. Either filter by ability, or drop everyone below 3k (which amusingly will catch some elemental shaman. Poor elemental shaman.... hence why ability filtering is better.)

Hell, the protadin in a friend's guild did 3.6k tanking on Patch.


Also if you aren't already making sure it's just 25-mans, you should do so.

10-man results are way too skewed because you'll never come even close to buffing everyone in the raid, it will favor melee or casters (and probably melee, since tanks bring most of the melee buffs they need, whereas casters need odd offspecs like eleshm and moonkin.)

[misterP]

yeah i did think about filtering by ability usage - that'd make it very easy to catch HAT abuse and such as well..

but it's just so slow. I might buy a WWS account to fix the slowness problems. Damned $9 a month :P

yeah i'll give that a go when i'm back off holiday.. unless another programmer steals my idea and does it better. Then i'll just be pissed off :P

Oh and yes, it's only 25 mans.

[Fad]

I think wowmeteronline.com already has something like it

http://www.wowmeteronline.com/rank/classrank/6

[Annaconda]

yeah i did think about filtering by ability usage - that'd make it very easy to catch HAT abuse and such as well..

but it's just so slow. I might buy a WWS account to fix the slowness problems. Damned $9 a month :P

yeah i'll give that a go when i'm back off holiday.. unless another programmer steals my idea and does it better. Then i'll just be pissed off :P

Oh and yes, it's only 25 mans.

I like your data. You should really make a site which gives these info and put some addon to filter some info if people dont want it.
And to be honost i would really like some DPS comparing for tank aswel. Just to see if your Damage(threath) right

[Accalon]

I think wowmeteronline.com already has something like it

http://www.wowmeteronline.com/rank/classrank/6

Is off on a few fights....

[Fad]

Is off on a few fights....

remember this is the average of the class in each boss not the top.

top is here

http://www.wowmeteronline.com/rank/clazz/dps/all/6

[Fad]

I like your data. You should really make a site which gives these info and put some addon to filter some info if people dont want it.
And to be honost i would really like some DPS comparing for tank aswel. Just to see if your Damage(threath) right

wowmeteronline has a list for top TPS for each boss

http://www.wowmeteronline.com/rank/clazz/tps/all/6

I think if u ask them to make a list for Average maybe they will make it. the guys there almost make every feature request anyone ask for.

[Dragonfiregum1]

This is kinda funny, considering most of the WWS are fake for top like 25ish of every fight.
You can tell by the fight length and the amount of cd's and things like procs.
Useless data.

[Mixon]

that data seems to report a relativly close grouping of most dps classes....wonder where the evidence for the hunter nerfs are comming from....earlier on when our dps jumped drastically b4 lvling out and the other classes catching up and often passing us by a little?

and in the list of top dps-ers there are no hunters...i would have expected at least one and i also am not sure i believe thoes numbers they seem way to high.

at any rate our dk's are doing alot more than that

[Sebid]

What's with the 34k mage tps @ Razuvious

[killaklown]

I think wowmeteronline.com already has something like it

http://www.wowmeteronline.com/rank/classrank/6

thats just crap, lol..

Notice how the bottom are always hybrids? its because they dont exclude healers or tanks from the dps, which brings the class's dps down.

[Fad]

What's with the 34k mage tps @ Razuvious

what?. where?. cant see a mage in that list :P

thats just crap, lol..

Notice how the bottom are always hybrids? its because they dont exclude healers or tanks from the dps, which brings the class's dps down.

agree it can use more work to get the hybrids right.

This is kinda funny, considering most of the WWS are fake for top like 25ish of every fight.
You can tell by the fight length and the amount of cd's and things like procs.
Useless data.

to state that most off wowmeteronline or even wws logs r fake is an overstatement. can u give us some examples plz. and tell is why is it fake. and yes there r some fakes but very few and are founded easily. I think it can be useful data.

[Razulath]

yes, and this http://www.wowmeteronline.com/combat...3314#damageout
is so true also. Fakes are everywhere. 20k mauls in bearform, FF critting for 17k, 4k lacrate dot.

Stop comparing to top dps, 100% chance they are fake