EVE General Discussion

Ack! Math! My only weakness! How did you know?!

Personaly, I would just use multi-linear regression and hope for the best. Maybe toss a few variables like, I dunno, number of people online versus the population of Amarr at a given hour. But then I'd have to check collinearity to make sure both variables are independent and...

Gosh, I really hate Math! I don't need Sine Waves!

MacGyver and a Minmatar sit trapped in a small room.

MacGyver pulls out some chewing gum,
a LED,
wires,
a knife,
two packs of nails,
a condom,
old news paper,
hla hankerchief,
a PowerPuff Girls comic book,
the EVE Second Decade Collectors Edition ...

... and a stick of wood.



Suddenly, he looks desperate. He checks all his pockets,
pants, jacket, everywhere...



"OMG I LOST THE..." he starts, as the minmatar signals him to silence ...










..... ssssssssssssssshhhhhhh we're free soon .....











... and pulls out a huge roll of duct-tape.

Nope. The sum of two periodic functions is still periodic. You have players eating, sleeping, working and playing eve in Australia. You have players eating, sleeping, working and playing eve in Britain. They both tend to play eve in their prime-time. If you track the activity level of any sufficiently populated system in eve, they will all be periodic with the same pattern.

I read some of the posts in his blog. Just seems like dressing up simple concepts with math to make the subject matter seem less trivial than it actually is. Previous post was: if you look at activity x in region y and activity x in region z, if they're similar you'll see similar behaviors in those two regions. No ****.

The most recent post is: here's a periodic function. Population levels in a given system are periodic. Yay! His previous post is: Here are the jumps in Jita. Here is the Fourier Transform of the jumps in jita. And look, the same pattern is present!

Here's an useful example of signal processing. Cold war, 70's: satellite takes a picture of a forest in east germany, and drops a canister of microfilm back to earth which is picked up by a c130. The film has a picture of a forest, seemingly empty of anything but trees. Use Fourier optics to convolve the film with another film with an image of a tank. Boom, every single soviet tank on the original microfilm lights up like a flare.

In general, fourier analysis is used as a way to simplify complicated math. Say you want to see how similar two complicated signals are (this can be data points, or as in the soviet tank example, an image). You could do a direct convolution, but for two arbitrary signals this would be computationally intensive. Or, you could convolve each arbitrary signal with an exceedingly simple function (this is a fourier transform) and multiply the two signals together in fourier space (this step is simple multiplication), then you can look at the spectra directly or convolve the product-signal with an exceedingly simple function again (this is an inverse fourier transform).

The point is, you take a complicated signal or image, and extract useful information that wasn't readily available beforehand (e.g. soviet tanks in what appeared to be an empty forest). In that blog I see already simple patterns dressed up with math, no useful insight over the original data is provided.

The entire blog just kinda rubs me the wrong way, dressing up exceedingly simple things with math to seem "ardently intellectual" as the blogger describes himself. Seems pretentious, maybe I'm judging too harshly though. vOv

Actually that's where I started, but I got annoyed that I needed all these discrete variables to account for time. I wanted a better way, hence where the sine wave stuff came in. Ideally I want a formula to convert all of the time data into a linear function so I can only use one variable to capture all of the time-based variance. Something of a pipe dream, but even with the messive amounts of data I have, I don't have enough to properly analyze the game on all the dimensions I want to.