Mar

6

We do not use moving average systems*, but we do use approaches (e.g. sector rotation) that have lookback periods. One of the first things we did was to check all possible lookback periods. That research was not to find an optimal one, but to look for "sweet spots" and to hopefully identify the reason. We did find sweet spots for all parameters, and not just the lookback periods. The lookback sweet spots tended to cluster around very identifiable periods, like monthly, bi-monthly, quarterly, etc. Through an interesting approach we identified some of them as related to options exercises, and others as related to earnings releases. We subsequently theorized that since a large portion of the market followers are fundamentals-based and specifically earnings-based, it is quite logical for a momentum/velocity/acceleration-based ranking system to be successful with a quarterly lookback period. That is, what appears to be a technical system (i.e. price-based) is in reality a fundamental system.

One wonders why recent success would have any predictive capability. That's a fair question, and one we continually agonize over. However, let me use a sports analogy. (I hate to use sports analogies, as it usually alienates many women, although my wife would disagree.) Last years NCAA men's basketball champion was Florida. Forgetting their current performance, Florida was highly ranked at the beginning of the season. Why? Well because they had many returning players. Past successful expertise doesn't stop just because of an arbitrary end of the season. The same is true in investing. Long ago it was discovered that one of the reasons for mutual funds having a second year of good returns is that the stocks held in the first year continued to be held.

The above approach is fine, but requires repeated testing using continually-updated out-of-sample data. What we prefer is identifying a particular characteristic of the markets that is reflective of current conditions and using that characteristic to define and alter your lookback period. This makes your lookback period adaptive, and effectively negates the need for out-of-sample data.

* If forced to come up with a long-short strategy and in need of the short side system, we have found the best approach to be moving-average based. It won't make money either long or short, but will go nowhere and reduce the margins for the long positions. However, I don't think that's an endorsement of moving average strategies.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search