Historical Research

Patton Fund Management, Inc.’s research and back-testing of its investment strategies sets it apart from most other long/short strategies. Testing of daily data over more than four decades demonstrates that Patton’s long/short strategies would have produced consistently positive, market-beating returns over the period with reduced downside risk.

Back-testing is the process of saying “(1) had I been utilizing these investment disciplines (2) at this particular time (3) under these types of market conditions (4) with these stocks as candidates for my portfolio, (5) what would have been the results.” Clearly, only purely quantitative, rule-based strategies that are devoid of human discretion, such as Patton’s long/short strategies, can be tested in this manner.

Patton’s Testing Process

The testing process requires resources, knowledge, and time. Successful testing that produces reliable results requires all three.

Data Resources - Accurate, comprehensive, unbiased data are the backbone of the back-testing process. The University of Chicago’s CRSP U.S. Stock Database is arguably the best source of historical data on U.S. stocks. This database contains daily data, including price, volume, market cap, industry, and dozens more, on virtually every stock from July 1962.

The Standard & Poor’s Compustat database contains hundreds of data items on virtually every U.S. stock. The Standard & Poor’s data are updated daily. Patton merged these two data sources creating a continuous stream of data on all U.S. stocks.

Technology Resources - Data, unorganized, are of little value. Patton designed and created a proprietary, sophisticated computer software system to manipulate the billions of bits of data. This software system is specifically designed for both back-testing and for execution of the disciplines on a daily basis in the long/short strategies. This software system allowed Patton to run hundreds of scenarios spanning the entire four decades of data to confirm that the investment disciplines produced successful results regardless of minor modifications to the exact formula.

The same model that was used for back-testing is also the system that is used for implementation going forward. Using this same system allows Patton to implement the exact same investment disciplines in the long/short strategies as those that were tested.

Patton spent more than four years conducting research. The combination of the best available data and a sophisticated computer software system were the tools used to rigorously test, refine, and ultimately, clearly and explicitly define the investment disciplines implemented today.

Key Principles to Successful Back-Testing

There are five key principles to successfully back-test an investment strategy. A description of each follows.

1) Maintain a Static Discipline - When back-testing any investment discipline, at no time should decisions be made utilizing the benefit of hindsight. There is often a temptation to believe market behavior is predictable when doing research and to alter investment disciplines as the conditions change. This temptation was avoided in Patton’s research.

The investment disciplines utilized by Patton are both quantitative and static. It is only disciplines such as this that can be reliably back-tested.

2) Test a Long Period of Time - One of the most common mistakes made by investors is to base future expectations on results from short periods of time. Many investors during the late ‘90s thought stock prices would continue higher by 25% and more in the future. Many of these same investors thought stock prices would suffer indefinitely during the bear market that followed. At both times expectations were based on recent short-term periods of time and were flawed.

The same problems occur in the research process. Often back-testing is performed on short periods of time resulting in false expectations of future behavior and performance. The investment disciplines of the long/short strategies have been back-tested over more than four decades of data from 1963 – 2008. Several market cycles and a vast variety of conditions occurred during this time.

3) Eliminate Survivor Bias - This challenge is only overcome with comprehensive and quality data. Many data sources used when conducting research only include data on those stocks that are being actively traded today. Those stocks that no longer trade for a variety of reasons (ex: bankruptcy, mergers and acquisitions, leveraged buyouts) are not included in many databases. Research utilizing such data produces false expectations of future behavior and performance, a problem Patton avoided.

Patton’s research utilized the best, most comprehensive and accurate data sources available. The University of Chicago’s CRSP U.S. Stock Database was the primary data source for this research and includes complete data on virtually all stocks that have traded on the U.S. exchanges. Survivor bias was eliminated in Patton’s research.

4) Avoid Data Mining – Data mining may involve searching the data for a formula that provides a “best fit” and then using the data to justify the formula, or it may involve selectively searching data until support for a particular formula is detected. In contrast, the goal of Patton’s research was to select investment disciplines (i.e., a formula) that were likely to generate superior performance going forward based on the principles of Behavioral Finance.

Patton did NOT develop the long/short strategies through data mining.

5) Ability to Implement - Research and investment disciplines that cannot be implemented in real time with real money are of no value. Patton’s quality assurance process demonstrates the proven implementation of the investment disciplines with real money. This process includes a comparison of actual results to the performance of the model. Important to this process is that an estimate of all fees and expected transaction slippage has been incorporated in the model. This quality assurance process shows only a marginal difference since its implementation. This real money experience clearly demonstrates that Patton’s static and quantitative investment disciplines can be implemented efficiently in real time with real money.