What is the proof that some new application is better than the alternatives?
After the Berlin Wall fell, demarcating the end of the Cold War, the financial world saw an influx of rocket scientists. Former engineers, physicists, and NASA scientists flooded Wall Street, the hedge fund world, and the managed futures world. I suspect they helped spur many evolutions we’ve witnessed in financial technology.
In the mid 1990s I met many of the biggest CTAs around at the time. I recall two of them, both former Cold War engineers, referred a lot to the mathematics of chaos theory. Needing to edify myself on the subject before meeting them, I read everything I could find on the topic, which at the time included writings by Mandelbrot, Gleick, and something ancient I found at the Harold Washington Library by Edwin Hurst. The most relevant writing pertaining to trading, however, were books by Edgar Peters. Interesting stuff – Hurst exponent, persistence and anti-persistence – it opened my eyes to some of the pitfalls of treating everything as if it had a Gaussian distribution, as Taleb rails about so often.
One of the chaos CTAs was enthusiastic about my having read up and was happy to discuss how interesting it all was. As I had already concluded, he conceded that Hurst exponents and related math were more useful to find market conditions where trend following might be profitable than to identify precise entry and exit points. In other words, chaos math didn’t make you any more money than the traditional trend following metrics did. The other chaos CTA stood fast, insisting he actually was using chaos mathematics in his system and it was the basis for his trade entry and exit points. Taking him at his word, I then asked how was this an advantage to his clients. Why did he have such a high correlation with and returns indistinguishable from all the other trend followers? If using a Hurst exponent to find persistence is no better than using less interesting technicals, how were his clients any better off?
Plenty of time has passed since then – per Moore’s Law, microprocessor speed is over 1,000 times faster than it was. Bloomberg has offered the function KAOS since the late 1990s, and these days there certainly are a lot of technical trading systems you can buy purporting to include indicators based on chaos math. It wouldn’t surprise me if some hedge funds or HFTs employ chaos math in their algorithms – in fact it would surprise me if none did.
Whether you’re vetting a new trading signal, a risk model, or an attribution system, don’t get seduced by flash. Look for the substance. Insist on the proof that the result is better than your alternatives, or that it has no alternative.