Avoiding drawdown breakdowns
When starting to trade a new strategy, many traders wonder, “How will I know if and when the strategy is broken?” The fear, completely understandable, is that the trader made some sort of mistake during development, and that without a proper feedback mechanism in place, he will not notice the system is broken until it is too late.
Many traders use a gauge based on historical hypothetical drawdown to determine their quitting point and, unfortunately, many times they end up trading a bad system for too long. Thankfully, there are alternative methods to determine objectively when to stop trading a strategy. This article examines two possible methods and compares them to a standard maximum drawdown approach.
A trading cliché that most of us have heard is, “Your worst drawdown is always in the future.” For most trading strategies, that is absolutely true. Many traders anticipate this, and stop trading when the live trading maximum drawdown exceeds some multiple of the historical maximum drawdown. For example, if a backtest shows a maximum $10,000 drawdown, the trader might decide to stop trading at 150% of this level ($15,000 maximum drawdown), or 200% of the historical value ($20,000).
In either case, the trader succeeds in limiting system losses, hopefully before the account balance is wiped out. That obviously is key. The adage “cut your losers short” applies not only to trades, but also to trading strategies. But before the strategy is stopped, the trader has to endure a painful drawdown. Is there a good alternative, with hopefully a smaller amount of financial pain?
The fields of statistics and statistical process control offers two alternative quitting points. One is a control chart and the other is a cumulative history chart.
In the world of manufacturing, process control is key. For any machine or process to produce high-quality, consistent parts, the machine must be in control. This will ensure that any parts produced are within specification. To monitor the process, quality control engineers typically use a statistical process control (SPC) chart. This chart plots a key characteristic or dimension of the part in question, and alerts the machine operator to any potential issues (see “Staying in control,” below).
The goal of using a control chart in manufacturing is to identify problems with the machine before defective parts are produced.
The following are example rules for alerting the operator to potential “out-of-control” problems:
- One point above/below the three standard deviation control limit (denoted by square in the figure)
- Two of three successive points above/below the two standard deviation line (denoted by circle in the figure)
- Four of five consecutive points above/below the one standard deviation line (denoted by triangle in the figure)
- Eight points in a row above/below the average line (denoted by diamond in the figure)
Once an out-of-control problem is identified, the machine operator likely would take some action, such as adjusting settings on the machine, performing maintenance on the machine, etc.
In a similar manner to manufacturing, a trading system can be considered a machine. Market data is the raw material going into the machine, the strategy rules compose the machine itself and trade results come out of the machine. The trade results are considered the finished part. So, for trading purposes the part dimensions shown in “Staying in control” just can be replaced with individual trade results. When a strategy is thought of this way, a process control chart can become a useful tool in seeing if the machine (that is, the trading system) is broken.