Historical Simulation: The Software Gap

By: Louis B. Mendelsohn
President, Market Technologies

Over the past several years there has been a dramatic increase in the number of investors utilizing microcomputers, concurrent with a proliferation of software designed to assist in the investment decision-making process. As technological advancements occur almost daily in the computer industry, a “software gap” — due in large part to the lengthy lead time necessary for the development of application software — has become evident. Software increasingly fails to take full advantage of the architecture and capabilities of microcomputers. This gap has particularly serious implications in the investment field, with respect to software directed toward market analysis (often referred to as decision support software) where inadequate software design has a direct impact on “bottom line” trading performance.

Tool Box vs. Black Box
There are currently two major approaches to decision support software. The first, and by far the most widely accepted among stock market traders, is traditional technical analysis software. Designed as a “tool box” for individuals attempting to analyze market trends, technical analysis software is extremely flexible and offers traders a smorgasbord of technical studies from which to choose. The primary function of this type of software is to provide extensive charting and graphic capabilities, thereby reducing the burden of manually tracking and displaying this information.

There are currently two major approaches to decision support software. The first, and by far the most widely accepted among stock market traders, is traditional technical analysis software. Designed as a “tool box” for individuals attempting to analyze market trends, technical analysis software is extremely flexible and offers traders a smorgasbord of technical studies from which to choose. The primary function of this type of software is to provide extensive charting and graphic capabilities, thereby reducing the burden of manually tracking and displaying this information.

The second approach, initially adopted by futures traders in the late 1970’s, generates actual buy and sell trading signals through the use of secret “black box” logic locked into the software by the developer. Referred to as system software, this genre is presently gaining increased popularity among stock market traders.

The Need for Historical Testing
Perhaps the most serious manifestation of the software gap affecting decision support software is the unavailability of highly sophisticated and valid historical testing and simulation programs. As a result, individual investors today are no more able to formulate trading plans and evaluate their performance than in the past. With few exceptions, traders must still rely on evaluation methods such as “paper trading.” This involves manually applying a trading plan to real data spanning a relatively short time period. After superficial determination of the plan’s performance, it is implemented in real-time trading.

Perhaps the most serious manifestation of the software gap affecting decision support software is the unavailability of highly sophisticated and valid historical testing and simulation programs. As a result, individual investors today are no more able to formulate trading plans and evaluate their performance than in the past. With few exceptions, traders must still rely on evaluation methods such as “paper trading.” This involves manually applying a trading plan to real data spanning a relatively short time period. After superficial determination of the plan’s performance, it is implemented in real-time trading.

This alternative to extensive computerized historical testing is an anachronism. The process of manually calculating technical indicators is extremely burdensome and imprecise, even for relatively simple trading plans. The trader is not able to adequately evaluate many important factors such as maximum draw down, the number and dollar value of consecutive losing trades, and other measures of risk that determine the plan’s overall performance. Also, with manual testing (and even using limited computer-assisted testing) it is not feasible to vary technical indicator values and ascertain the effect on the plan’s performance. Under these circumstances, trying to determine the plan’s profitability and acceptability on a portfolio of trading vehicles becomes impossible. Finally, the plan often appears profitable when manually evaluated over a very limited time span, but fails to produce an acceptable profitability/risk tradeoff during extended periods of real-time trading.

A second and more dangerous alternative to computerized testing involves the formulation of a trading plan and its testing through actual trading. If the plan is profitable, the trader continues to adhere to it. Otherwise, he takes his losses, formulates a new plan and begins the process all over again. This hopscotching phenomenon is alarmingly prevalent and is largely responsible for the losses that many traders incur.

The formulation, testing and implementation of trading plans, through use of sophisticated software, is totally feasible within the framework of existing microcomputer technology. Historical testing capabilities could allow the investor to apply an exact set of indicators to past price data. Buy and sell transactions would be simulated, and the resultant gain or loss from each trade recorded. At the end of the simulation, the program would generate a comprehensive report outlining exactly what would have happened to the trader’s positions and account equity had he been trading in real-time over that period, using that exact combination of indicators.

The major benefit of computerized historical simulation and testing is the ability to perform optimization. Through the simulation of multiple combinations of technical indicator values, optimization allows the formulation and fine-tuning of trading plans that are unique for each individual investment vehicle and customized to each investor’s trading objectives and risk propensity. As the market undergoes structural and fundamental changes over time, the resultant trading plans are easily re-optimized to reflect these changes (as opposed to remaining static and becoming outdated in relation to prevailing market conditions). Subtle refinements to trading, such as timing of executions and sensitivity of the indicators to retracements, could also be incorporated into these historical simulations.

The inclusion of sophisticated historical simulators as an integral part of decision support software would allow traders to conduct extensive evaluations on the effectiveness of trading plans prior to risking capital in real-time trading and could therefore dramatically improve individual trading performance.

Conclusion
Until investors recognize and acknowledge the importance of historical simulation and software developers accept their responsibility to provide such capabilities, this software gap will continue to severely limit the role of computers in the investment decision-making process, preventing traders from fully benefiting financially from available technology.

Louis Mendelsohn is president of Market Technologies in Wesley Chapel, Florida. Mr. Mendelsohn is a recognized expert on futures trading software and a widely-published author of articles on technical analysis.