Has anyone ever done Fourier Analysis on the stock market?

[Trigonal Planar]

Title pretty much says it all....I know that even for really noisy signals (ie, astronomical data) Fourier Analysis can often reveal a repetetive pattern. Has anyone ever applied FA to stock market data? What was the outcome?

[Joey P]

Probably not what you looking for.

[Mathochist]

Originally Posted by Joey P

Probably not what you looking for.

(a) What little math there is in there is awful.

(b) Cohen was supposedly a number theorist, not a harmonic analyst. Yes, there's analytic number theory, but he was really concerned throughout the examples with natural numbers.

[Trigonal Planar]

Errmm, no, not quite....good movie though.

[spingears]

Originally Posted by Trigonal Planar

Title pretty much says it all....I know that even for really noisy signals (ie, astronomical data) Fourier Analysis can often reveal a repetetive pattern. Has anyone ever applied FA to stock market data? What was the outcome?

Don't know but suspect some work has been done.
There is plenty of data to support a Doctoral Thesis in mathematics or Business.
Modern PC's should make it a snap to work on various stocks and their relationship to each other, market averages, etc. etc.

[ultrafilter]

As easy as it would be to do, it's very hard to believe that no one has. The only reason why they wouldn't publish is that they can't find anything interesting.

[Squink]

Google turns up over 6,000 thits for Fourier "stock market"
Lots of snake oil, and debunking:
Stock Market Pseudoscience

the technical system designer doesn’t even know how to quantify the variables underlying the market. Those who have tried to apply Fourier analysis to markets have found this out to their cost.

Also a few like this:
Multifractal Spectral Analysis of the 1987 Stock Market Crash

[astro]

See The Discrete Fourier Transform Illusion

[Shalmanese]

I rememeber quite a while back SciAm had an article about the fractal nature of the stock market (something about a random walk down wall street?).

Anyway, I suspect wavelets would be more appropriate to stock market data since it doesn't throw away time domain info and I can't imagine theres no time related trends in stock market analysis.

[ianzin]

Yes, there are people who have tried this. I used to work for a company that even sold its 'analysis' of trends in the currency markets to willing punters.

There are also lots of snake oil sales creeps who claim to offer hot tips based on advanced analysis of one kind or another, often alloyed with (take your pick) 'expert' insider knowledge, access to privileged or restricted-but-legal information, meta-analysis of top tipsters and recommendations, faster delivery of crucial information and so on.

It's all rubbish and it doesn't work. Here's all you need to know: if someone were able to produce any kind of forecast pertaining to the markets which made it possible to achieve any kind of special financial profit, then they wouldn't need customers. They wouldn't need to sell their 'information' or 'tips' or 'advice' or 'analysis' to anyone, and they wouldn't need all the attendant hassle (taking out advertising space, paying for it, submitting advertising copy, processing orders and so on). They could just sit at home all day and make money.

By special I mean more than the average Joe can make from a perfectly normal portfolio of investment products obtained from the regulated market - brokers, banks, investment houses and so on. These can provide whatever mix of investments Joe wants, selecting from the normal spectrum of high risk/potential high gain at one end and low risk/low potential gain at the other.