Multi-Dimensional Equity Curve Analysis

It is hard to fault our tendency to think in only a few dimensions at one time. Our minds are designed for survival, and hence we rely on heuristics, experience and linear-thinking. But its important to think about these things in the stock market because it impacts our chances of survival in the brave new world of modern computing and artificial intelligence. While you look at a chart of Apple Computer, the computer algorithms at Goldman Sachs are  simultaneously analyzing  Apple, QQQQ, IBM, Microsoft, option prices, the US dollar, interest rates and a host of other inter-relationships. Their goal is to determine relative value, or to predict changes in the stock price. So while you are predicting a breakout or a breakdown by looking at historical patterns, they are looking in multiple dimensions at the “invisible” factors that affect Apple prices. Clearly they are going to be more accurate than you at predicting the future price of Apple,  no matter how good you are at reading charts. The market is a highly complex and interconnected web, and to look at the price of the underlying is to operate in only 2 dimensions.

This line of thinking affects strategies as well. It is by no means limited to complex real-time intermarket analysis —beyond the scope of even sophisticated traders. In fact, simply integrating multiple time frames in a comprehensive sequence will greatly assist in classification accuracy. This approach is well-documented by Brian Shannon of Alphatrends, and his book is highly recommended as both a theoretical and practical framework: http://www.alphatrends.net/.  Basically, to properly predict direction you are better off aligning yourself with more than one time frame, and also giving added weight to the long-term time frame. This general concept affects equity curve analysis in unique and interesting ways.

When we are trying to determine whether or not to turn strategies on or off such as daily follow-through, we are only looking at one time-frame. Our assumption is that we can “time” daily-follow through with a reasonable degree of accuracy for the purposes of creating an adaptive approach.  However, much like the examples mentioned above, there is reason to believe that looking at more things simultaneously will improve our ability to make predictions.  In this case we can take a multiple time-frame approach and look at the spectrum of mean-reverting strategies, and ideally consider intermediate and long-term time frames as well. Furthermore, we can look at the factors that affect mean-reversion such as volatility and trendiness. From there we can use an even more layered approach by considering markets other than the S&P500, and perhaps even look at the individual sector or stock level.  This is the type of approach that is far less prone to random error, and data-snooping, and is much more “robust.” An effect such as mean-reversion is broad-based and driven by strong common factors that many of us many not currently understand, but nonetheless exist. Detecting such shifts in the tectonic plates of the markets requires a multi-dimensional approach to equity curve analysis.