Let's do a little math on server stats

Makes absolutely no sense that there's 126 US servers in Swtor, but Guild Wars is launching with 25.

No wonder every servery is a ghost town (mins The Fatman).

Yeah, that seems unreasonably high.

A nonsequitur statement.

GW server farms have absolutely nothing to do with a discussion about SWTOR server averages, etc. etc.

And only about 15-20% of the servers are "ghost towns" on SWTOR, so you put a heap of hyperbole into your nonsequitur as well it seems.

suuuuure......

current status:

http://www.swtor.com/server-status


Worldwide Stats:

Full: 0
Heavy: 2 Servers
Standard: 42 Servers
Light: 175 Servers

Oh yeah, "Primetime" is the word I´m waiting for.. let´s do this again every 2 hours and recount.

This is the main problem. Not that there isn't anyone paying for and playing the game (there is) just that these people are spread too far apart across too many servers, and with no cross server anything you are restricted to your own server to get anything done.

It's not really a BS thing, it's another issue of the average vs the standard deviation. An average of 900 concurrent players across 3 servers (for simplicities sake) can be acheived in various ways.

3 servers at 900 players standard deviation 0
__
1 server at 900, 1 at 450, and 1 at 1350
Which gives a standard deviation of 300, so the corrected average would only include values from a 600 player minium to a 1200 player maximum.
So the 450 example and the 1350 example would be abnormally low, and abnormally high. So they would be excluded from your averaging and subsequent calculations.

Since you haven't put forward your breakout numbers for the servers, your scenario fails at the basic checks for statistical rigor.
In my opinion after going back over things, this appears to be a form of the 'the game is dieing' troll with some bad math thrown in to make it look good.
So shenanigans, shenanigans I say. I cast dispersions on the method, form, and substance of your argument.

Right. And don't forget that each region has its own prime time, or you'll end up with yet another "garbage in, garbage out" calculation. Unless, of course, your goal is to demonstrate that off-peak, fewer players play than on-peak.

You don't throw out data just because it is outside one standard deviation of the average. If you have a normal distribution (i.e. your classical bell curve), then 32% of your observations will fall outside +/– one standard deviation.

http://en.wikipedia.org/wiki/68-95-99.7_rule

That´s the meaning of "average", for every population number which is below average, there is most likely one which is above average, that is the principle. Take the population numbers together and divide them through server numbers. Then compare to real life situations, check server status, log in at "primetime" and see how many are actually there on light/standard/heavy servers.

So, why exactly are you replying then? What is your point instead of putting funny and popular forum compatible words like "shenanigans" and "troll" in your post?
Fix the calculation if you think it´s wrong, or play somewhere else. Prove where the 1.7M subscribers are and how long they are playing on average, is it really that hard?

Rejecting the data outside the standard deviation band is a method to normalize the data, because the sample is non bell curve, and in the case of fatman or the other couple very high pop servers could show numbers an order of magnitude higher then other servers. It is an accepted method to exclude data like that but acknowledge it in the analysis.

Sorry about that understated it, for a sampling like server pops the sample isn't going to be normal for various reasons including local peak times. Any sample is going to have bad data, and the standard deviation gives you the acceptable observation band. The calculations realistically need to be done one of two ways, on only the items in the standard deviation band, and acknowledging the rejected data and why it was rejected. Or 2 sets one with the out of scope data and one without.

Rejecting the entries outside the standard deviation range for the server population calculation lets you clean up the data to show the trending better. This will allow you to compensate for the insane swing that the very large population servers like fatman would do to pull up the concurrent login numbers.