Inside the Black Box: The Simple Truth About Quantitative Trading

3.5. BLENDING ALPHA MODELS

Each of the decisions a quant makes in defining a trading strategy is an important driver of its behavior. But there is another extremely important set of choices the quant must make in constructing a trading strategy. Specifically, the quant is not limited to choosing just one approach to a given alpha model. Instead, he is equally free to choose to employ multiple types of alpha models. Even the method used to combine these alpha models is an arena rich with possibilities. The most sophisticated and successful quants tend to utilize several kinds of alpha strategies, including trend and reversion, and various kinds of fundamental approaches across a variety of time horizons, trade structures, instruments, and geographies. Such quants benefit from alpha diversification in exactly the same way that diversification is helpful in so many other aspects of financial life.

Blending or mixing alpha signals has many analogues in discretionary trading (and decision making) in general. Imagine a mutual fund portfolio manager who has two analysts covering XOM. One analyst, focused on fundamental value in the classic Graham and Dodd sense, expects XOM to rise by 50 percent over the next year. The other analyst, taking a momentum approach, thinks XOM is likely to be flat over the next year. What is the net expectation the portfolio manager should have of the price of XOM, given the two analysts' predictions? This is the core problem that is addressed by blending ...