Howard Bandy, Foundations Of Trading by Brenda Jubin
Howard B. Bandy is best known as the AmiBroker guru. As a result, traders who used other platforms sometimes ignored his earlier work. But as Foundations of Trading: Developing Profitable Trading Systems Using Scientific Techniques makes clear, they do that at their peril. There’s so much twaddle passing for proven fact in the trading community that it’s refreshing to read a sober discussion of just how hard it is to become a profitable systematic trader. And to get a detailed account of the elements that must be integral to any successful trading system.
The text of Foundations of Trading is only 160 pages long, but it is packed with insights. Bandy, a former professor of mathematics and computer science, seems to approach writing as he would a computer program—his prose is precise, succinct, efficient.
This year has been a record-breaking year for initial public offerings with companies going public via SPAC mergers, direct listings and standard IPOS. At Techlive this week, Jack Cassel of Nasdaq and A.J. Murphy of Standard Industries joined Willem Marx of The Wall Street Journal and Barron's Group to talk about companies and trends in Read More
Foundations of Trading: Developing Profitable Trading Systems using Scientific Techniques by Dr Howard B Bandy
Two seemingly intractable problems confront any trading system developer. First, he must find a weak signal amidst all the noise. Second, he is subject to the constraints of stationarity. For a trading system to be profitable, “the distribution of signals must be stationary over the combined length of the in-sample and out-of-sample period.” The problem is that financial time series data are notoriously non-stationary. So the longer the holding period, the more likely it is that “conditions change, stationarity is lost, and profitability drops.”
Trading system development is a straightforward, if demanding, process. It involves “an extensive in-sample data mining and (hopefully) learning process” in which the developer “generates many alternative trading systems (ATS), each based on a specific model-data combination; evaluates each ATS, giving each a score computed by the objective function; and ranks the ATSs, preferring those with high scores, selecting one.” This is followed by “limited (ideally one time only) testing of the selected system using out-of-sample data to validate that the system has learned.” The objective function that Bandy uses is CAR25 (a lower limit for 75% of the distribution), “the credible value of expected equity growth associated with the risk-normalized forecast of a trading system.”
Foundations of Trading includes a lengthy chapter on risk in which Bandy explains how to quantify one’s personal level of risk tolerance. That is, to define the point at which the drawdown of the system is such that, when exceeded, the trader accepts that “the system is broken and must be taken offline.” In the process he introduces the notion of safe-f, the maximum position size for the next trade.
Here I have shared only a couple of points from Bandy’s book. It contains ever so much more—the distinction between impulse and state signals, traditional system development platforms vs. machine learning, and I could go on and on. It’s a book that every trader should read and re-read. And, even then, recognize that, as Bandy concludes his chapter on trading system development, “There is always a best. It is not always tradable. Even when it is, it may not be good enough.”