Totally true. Who would forecast future direction/intensity of the wind using historical directions/intensities of the wind ?
Cybernetic Analysis for Stocks and Futures : Cutting-Edge DSP Technology to Improve Your Trading (Wiley Trading) by John F. Ehlers (Hardcover - March 19, 2004) Rocket Science for Traders: Digital Signal Processing Applications by John F. Ehlers, John F. Ehlers (Hardcover - July 20, 2001)
I've never seen any beginning yet of an independent verification of what is being peddled here as 'rocket science' for markets. Except of course for the bought commercial 'performance' blurb. If you happen to have anything SOLID & CREDIBLE on this, please let us know. I put this in the same category as the Kalman market-scientism, i.e. a step lower than our Jack and our Profilogic goodies.
Don't get me started on Ehlers. He uses the double-sided Fourier transform when for a causal time process he should be using the single-sided LaPlace. These geniuses all fail the first step: conformance to assumptions and existence conditions. The only thing that I have found that is worth a shit is trivial things like fast estimators on price and derivatives. There is NOTHING hidden in price. BUT focus on pure price prevents perception of what is REALLY important: market action.
another case of 'if i haven't found it, it doesn't exist' ...natural response, bash the professionals
Mathematically, I AM a professional, thank you. And I make the odd dollar here and there trading. Enough to be able to distinguish traders from non-traders here.
Haha. I pity the person who tries to study this stuff in hopes of a market application. John M., in response to your comment about it being stupid to treat the market as a stream of numbers to find a "signal". While I know what you mean I will say that if the numbers are going up then I'll favor a buy "signal" and vice versa for numbers moving down. Also, with this "wind" analogy - how would you apply support and resistance? A change in wind direction? Whatever. Anyway, the kalman filter is very useful in video processing/sensing applications and mainly has applications in detecting texture from "noise". It is a useful tool in pattern recognition when you have an array of multiple signals such as in a video frame. Hence adaptive neural networks that can easily sense motion. All this BS about time series and discrete transforms is silly, there is nothing discrete about price action because time is not linear. You can have a 1 million buyers at one price and only 100 at another. If you want to sample a non-linear system and then curve-fit (as I believe this paper is trying to do) then you should stay in engineering/math. Mike
Mike. The real test of your number crunching is to know when price is going up only to go right back down immediately, as in a volume blowoff, a floor rotation, or a "false" breakout. Mathematical analogies to physical systems can't do that because they don't admit infinite amplitude derivatives. It's like trying to use mathematics to predict when a woman will spin on her heel and stalk away. Mike.