Monitoring Systems and Portfolios

All systems and portfolios must be monitored for changes in behavior. The obvious change is when the system is running a series of losses larger than usual. There are methods of monitoring the system that will warn of changes, however, before any substantial loss is incurred.

Bryant (2001) suggests the following methods. Calculate the average profit factor over a moving number of trades—say the last 20 trades—just as in an oscillator such as the stochastic. Plot this calculation and then run a moving average through the plot and watch the behavior of the profit factor window to its moving average. The profit factor should always be above 1.0, and minor oscillations should be ignored. Any drift lower, however, is a warning of something potentially wrong. The recent calculation of a profit factor window should be compared with the entire history of profit factors with a t-test to see if the deviation from the moving average is significant.

Perform a test run to see if the strings of wins and losses are within a normal distribution—that is, if they are random or not. If no dependency is present, smaller size positions should be traded after a win and larger after a loss. If the test shows a positive dependency, the run streak is significant and positions should be reduced until the last trade is a win, at which point position size can be increased.

The equity curve of the system should be checked periodically. One method is to sum the profits and losses over a specified number of trades—say 30—and plot this figure in time. The sum should remain positive, or equity momentum is declining. Another common method of watching the equity curve is to calculate a moving average of equity. Breaking of the moving average is not necessarily a signal for action but a warning. If a forward line is calculated or a trend line plotted and broken, action likely should be taken, as the system performance is deteriorating for some reason. Often when these signals of danger arise, the position size in the system is reduced until evidence of recovery is seen or the problem is resolved.

The sum or average of the percent of winning trades over a specified number of trades will tell if there is a change in runs. A z-test can test whether the differences in proportions or percentages are significant and worth investigating further.

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