Testing Money-Management Strategies

We covered the standard method of testing a system in Chapter 22, “System Design and Testing,” but now we are looking at complete portfolio systems that include, hopefully, many subsystems and many issues traded using each subsystem with different objectives that satisfy the requirement for diversification. Of course, we should already have tested each subsystem by itself with a diversified number of issues and have confidence in its results alone. By merging it into a portfolio system, however, we now need a method to test the entire model. We will see that the model often will produce better results than any of the separate subsystems do individually. To test the model, we can use the same methodology as used in Chapter 22, or we can use what is called a Monte Carlo simulation.

Because the testing system in Chapter 22 is plagued with the risk of curve-fitting and the inability to determine the success or failure of the system under varying circumstances, we need a testing method that looks at a multiple set of possibilities and tells us how well the rules, variables, and parameters in the combined subsystems handle change. The Monte Carlo simulation is one of the better and more often used testing methods of portfolio risk.

Without discussing the mathematics behind a simulation (because it can become complicated), we instead just outline what it does. A simple simulation can be performed on a spreadsheet, with considerable work, and a more sophisticated software program called Equity Monaco is available from www.tickquest.com for free. More sophisticated software is available elsewhere for a fee. More information on the mathematics of Monte Carlo simulations can be seen at www.montecarlosimulations.org.

As we will see later in this chapter, with the martingale betting system, a system may be profitable, but the bettor must be able to withstand a long series of losses (a large drawdown). This means that the trading system results, even if spectacular from the optimization tests, could be just the result of chance or luck.

In the martingale betting system, the system is profitable only as long as the bettor is able to withstand a long series of losses, a large drawdown. To measure whether the system is valuable in all circumstances and what the odds of failure from a series of losses might be, especially because the trader usually has limited capital, the trader (or investor) needs to test for as many different circumstances as possible. The Monte Carlo simulation does not use the rules, variables, or parameters in the original trading system. It uses only the actual trades, entry and exit, and the profit or loss from each. It looks at the sequence of these trades from various angles to see the probability of a series of losses leading to ruin in the system. It, thus, is a money- management test rather than a system test, although obviously the system determines the trades. If the money- management test fails by showing a high probability of ruin, the original system must be discarded, adjusted to improve these results, or other safeguards installed to prevent such a possible disaster.

The simulation takes the original trade data, profits, and losses and scrambles them in a random manner. This is done many times, usually at least 100 times and better if 1,000 or 2,000 times. An equity curve is then created for each scrambled sequence of trades. The results from each equity curve are then assembled to give the results and related to a normal distribution curve. The simulation is testing to see if the system is random and by how much. The less the system is random, the more likely it will be profitable with minimum risk of failure.

Source: Kirkpatrick II Charles D., Dahlquist Julie R. (2015), Technical Analysis: The Complete Resource for Financial Market Technicians, FT Press; 3rd edition.