I have been wondering with all these WWS reports and Armory data gathering, that would it be possible to scan all information from lossendil website and create graphs of class DPS outputs for different bosses, or even universal graphs like what is average damage of class X attack Y over Z samples (Maybe even split with combat durations). All that kind of little details what would bend into reasonable averages due sample size being so big.

Im not much of an coder so I cant say if this could be possible, but it would sure give pretty interesting information about class performances. Would this kind of information help our fanatic theorycrafters?

Yes it would =)

Write a parser that sends a ?Search=bossname, get results, parse the results (narrow it down to top100 maybe?), request results from each link from parse, parse those results. You'd need a paid for account though cause a free one just takes forever to load one resultpage

At least for me, seeing each classes average dps on it's own isn't that interesting. It would be cool to pair it up with an armory crawler though. See how dps would break down for (just an example);

Rogues with:
*T5 ilevels
* Various specs
* Pala buffs
* Druid buffs
* Enchantment shaman contribution to damage

T4 vs T5 vs T6 ilevles etc.

It would be higly academic and not really all that useful, but I'm a sucker for usless statistics like this .

On a more useful note, one could probably figure out, statistically, what specs benefit other classes the most, when or if there's a "breakpoint of usefull" etc.

One could also probably proove quite a lot of the more uncertain theorycraft experiments and possibly have enough data to figure out things uncertain but still pretty much belived to be figured out anyway (possibly) for stuff like 1roll vs 2roll etc.

The amount of data could easily become quite overwhelming though.

Biggest problem I see with it is, your data parsing engine would need to understand group composition, which isn't available through combat logs uploaded to wowwebstats.

i.e. Two different raids, both with an arms warrior in. In one raid, he's in the MT group, to give the MT battleshout, but gains no group buffs himself. The other raid has their arms warrior in a group with an enhancement shammy, feral druid and BM hunter. The latter is going to do WAY more damage.

I think grouping boss kills by raid dps would be a good way to create tiers for statistic comparison. I would love to be able to punch in my raid's total dps on a boss kill and see aggregate statistics for all raids that have dps in the same ballpark.

It's easier to learn what I'm doing wrong by looking at a shadowpriest in a guild near my guild's gear level. Looking at what players are doing in a BT guild can help too, but then it's figuring out how much is gear and how much is skill/playstyle.

Would it really matter? Lets say there was 1000 bossfights, and we would show for example top 100 of each class that had more than 90% appearance in the combat? Im sure there are plenty of ways for getting rid of bad samples.

There are many good reasons why it should be done as an "official" feature.

As for group composition, I'd think you should be able to interpolate quite a bit of it simply by looking at what buffs each player has had access too, don't you think ?

If you go to the WWS forums, you'll see that he's beta testing WWS 2.0 at the moment. It includes a mod that you install, and it will track more than just combat log things, such as group composition, loot recieved, etc...

I have yet to try it out on a raid, but it will be running for Sunday's run.

Here's the link to the beta forums. Wow Web Stats :: View Forum - WWS Client V2 beta test