Good job summarizing paragraph #3.

Agreed. The fault in the logic here is the class and spec population distribution, which varies between both factions and servers, not to mention the assumption that players who spend talent points in those particular trees actually use those spells or invest talent points to obtain them in the first place.

The other thing to consider is that Ret Paladins and Shadow Priests dominate the rankings at TalentChic. Hunters, as a class with 3 DPS specs (and valuable buffs in each tree) are going to focus more on perceived utility in deciding what to be.

While the analysis is solid napkin math, it's not much more than that.

Well egg on my face for missing that line by reading too fast.

I still maintain though that it's too large of an assumption to be making for this argument. I'm not asking the author to do this, but there probably is enough information out there available to the public to allow someone to make a sufficiently through analysis of the character population to determine how likely it is to have at least one character with replenishment.

That's probably more math than anyone would care to do and I wouldn't blame them. In the end the simplest solution (which people do) is to say, "LFM - DPS that has replenishment buff, for Naxx".

Also, is that assuming that each tree is a spec, so 10 classes times 3 trees per class equals 30 classes? Because mages seem to have four, frost, fire, frostfire, and arcane. and it's my understanding that DKs have specs for deep into each of the trees as well as a duel wield spec which involves a certain degree of hybridization to grab the tastier dw talents. Those are the only classes I am currently raiding with, so maybe there are a few trees that are unable to support a spec, or maybe there are well over 30 total specs.

While the article pays lip service to the fact that the entire article is a waste of time, it does still go on to post itself. :)

Hence the comment "Good job summarizing paragraph #3." is a little rough and short sighted.

This article is complete rubbish because the assumptions in it are rubbish. It even says so, although not strongly enough.

What is worse is that there are websites that allow you top use the REAL CLASS DISTRIBUTION and not phoney and exceptionally naive numbers to calculate.

If I was not the busy person I am, I would bother to run the numbers ith this data.

For all servers: http://www.warcraftrealms.com/census.php?serverid=-1&factionid=-1&minlevel=10&maxlevel=80&servertypeid=2

All there in bar graphs no less. Yes it is a self submit census, but there is no bias in that the /who they use to take the census has no opt-out for those online. I am sure you could find niggles and small holes in the sampling method that would increase the error by 5% or so.
Does not matter thought because it is 1000% better than 1/30?!?