Back at the time of TBC, every single hunter was a BM hunter, raiding was as easy as “send the pet and auto-attack” to top DPS charts and the Hunter class grew to be one of the most hated classes in the whole game.
With the class balances actions that have been taken with WotlK and the Dev’s idea that players should be rewarded for the complexity of the game style they have decided to play, BM was hit hard with the nerf stick, while SV was boosted beyond even MM to become “THE” raiding spec. With time this has changed and MM was rebalanced again over SV making them both “viable raiding spec”
Unfortunately BM was forgotten. Very few hunters kept playing it, and major raiding guilds refused to let BM Hunter seriously raid. This is still the case in min/max guilds.
There has been lately a lot of talk about BM viability in raid and true difference between BM and MM in terms of DPS. But no real serious analysis. The theorycraft is showing with the latest EJ DPS spreadsheet (.93i) about 10% difference between the two specs using BiS objects. But despite this, reputable and well known people in the Hunter community such as Frostheim of WHU, keep saying and advertising self fulfilling excercises and test to show a larger distance between the two specs, always in favour of MM.
Now i am an hardcore BM raiding hunter. I did play all the specs in the past, both as a rogue in Vanilla and as an hunter in TBC and beyond. But since the nerf i have decided to stick with BM to make it viable and explore all possible alternatives that would make it viable.
The difference i see when i play BM and MM is just about 7-8% at 251 iLevel. the theorycraft say with MM i am at 10274 DPS and with BM at 9564 DPS confirming the feeling i get from raiding constantly of just 7% difference (which goes down to 5% with the 3.3.3 soon to be introduced changes).
But those are theory and feeleings.
What i want to do here is prove them statistically. By using the correct scientifical statistical methods.
I want to set up the stage for this by defining equip, spec, gemmings, rotations, pet etc. then start with staging individual tries to validate the theorycraft.
I will do this by defining the minimum sample size required to have a confidence level of 90% with a confidence interval of 5% (which means i want to be 90% certain that the DPS i will calculate as average are correct within an interval of +/-5%)
After that will run as many test as the minimum sample size , calculate the average, the standard deviation, plot it in a gaussian curve. Each one of the above for both spec.
After that we will be finally able to compare the two gaussian curves and really define what is the gap between the two specs.
It is not going to be quick, nor easy becasue the sample size to obtain such level of significant data is not small. In fact, with a margin of error of 5% and a confidence level of 90% we will require about 116 experiments (assuming a total population of 200).
Which means i will have to run 200 scenarios and pick 116 random from the data set to run the statistic …
Quite a large undertaking … let’s see if, how and when it happens 
