Mechanical Trading Systems

Chapter progress:

A mechanical trading system is one in which every single decision is made for you by a computer program that has been designed to generate buy/sell signals. Over long periods of time, mechanical trading systems outperform human judgment, chiefly because human judgment includes impulse and emotion. By removing the day-to-day swings of emotion — including greed and fear but also pride, anger and sense of self-worth — the computer program has a distinct edge.

Overriding Signals

Having said that, there is no computer as savvy as the human brain. Indicators capture trader behavior with arithmetic, and by definition indicators always lag — but the human brain can detect and imagine human behavior more directly, and also more quickly. The predictive value of indicators is limited, whereas humans can often predict others’ behavior with fair accuracy. For example, if a leading economic indicator like US payrolls comes in double the consensus forecast, every human trader knows instantly how the market will react. Even the fanciest mechanical system will have a delay until after some of the price evidence of the reaction has had time to get recorded and included in the indicators. This is the rationale for overriding the signals of mechanical trading systems, and it is not a bad one — when used sparingly and only in situations where human interpretation is clearly superior. Otherwise, the rule holds that over long periods of time, mechanical trading systems almost always outperform human judgment because of the destructive force of emotion.

It is important to think through these issues because almost every Forex trader today uses a mechanical trading system to some degree. Most mechanical systems that have been correctly designed and executed will work to generate higher gains than trading willy-nilly on observation and instinct. The problem is not that mechanical systems do not work — it is that traders cannot resist overriding them. When the trader overriding a system has deep knowledge, lots of experience, and good instincts, a good mechanical system can be enhanced to deliver even better results. When the trader doing the overriding is not experienced and worse, vulnerable to chatter, overriding the mechanical system results in losing outcomes.

This is one of the reasons why trading coaches say that trading is a journey of self-discovery. If you really believe in the science behind technical analysis, you do not override your mechanical system. If you override to the point of getting bad outcomes, it means you have a discipline problem.

Which Indicators?

You can buy a mechanical trading system — hundreds are available — or you can build your own. Choosing indicators is not hard. You start at one end of a list of indicators provided by your software or platform, and look at them on a chart one by one. If the indicator aids your eye to see patterns and flows, it is a good indicator for you. Run through the list until you get two or more (but not more than ten) that you believe will guide you well. Then put them all on the same chart and “audit” the resulting trades to see how many were winners and how many were losers.

Building a mechanical system is a lot easier than completing a trading system, mostly because as you keep learning about technical analysis, you will always find some technique that is new to you that seems like an ideal complement to the indicators you already have. Sometimes this is true but more often, you find that as with all indicators, some will contradict the others. For example, the parabolic SAR and MACD work together quite well to generate buy/sell signals at nearly the same time, but both are lagging, and what do you do when faster indicators like RSI or the stochastic oscillator are generating the opposite signal? If you have a preference for trend-following, you will stick to the parabolic plus MACD. If your strategic style is to trade breakouts, you will place more weight on the RSI and stochastic.

The point is that you need to weight your indicators according to the timeframe you are trading and to your main trading strategy and style. If your trading style is to wait for an entry that is very unlikely to be wrong, you will miss the first part of a new move, but you will not fall victim to false breakouts and whipsaws. Your main style is trend-following. If your style is to snatch every opportunity, you need to resign yourself to a high ratio of losing trades — but also the occasional home run. This style likely makes you a swing trader. This difference in style harks back to your original trading plan. Which strategy is your main strategy? You cannot be a trend-follower and an opportunistic breakout trader at the same time. You must choose between them — or have two systems and divide your capital stake between them. If sometimes you heed trend indicators and sometimes you heed breakout indicators, you cannot trust the system’s hypothetical track record – the one you yourself created using backtests.

Jumping from one style to another while pretending to be following a single system is very common. It shows a lack of focus and inability to stick to the methodology already chosen, and thus it reveals the rule of emotion. In the case of the trend mode vs. the breakout mode, you chose trend when you are fearful (and probably just took a loss). When you are in breakout mode, you are in the grip of greed — after all, a breakout is an opportunity to make a profit. You are likely to lean toward following breakouts if you just had a fat gain (or maybe had a big loss and feel desperate about recovering it).

Choosing indicators is an interactive process that seesaws back and forth between the trading style you think you are starting out with and the features and benefits of the indicators in real time. Sometimes it is a work in progress that is never completed.

Combining Indicators

Nearly every system will have trend-identifiers (such as moving averages, support and resistance lines, etc.) and nearly every system will also have breakout and pending breakout indicators, like channels, patterns, and momentum indicators. How to use both types of indicators over a series of trades is the job of honest backtesting and serious discipline. Analysts always advise that you choose more than one indicator for your system and that no indicator should be highly correlated with another, since what you are looking for is confirmation. When a second indicator agrees with the first, the probability is now much higher that the new signal is correct. The confirmation principle is well-established, but combining competing technical concepts is not well-defined or described anywhere. Technical writers just say “use what suits your risk profile” without describing how. This is annoying and frustrating — the experts back away just when you are at a critical point — but it is a necessary cop-out. The only technique for choosing your own set of indicators is to backtest each of them separately and then together.

Backtesting

Honest back-testing is critical to selecting indicators, whether you buy or invent a mechanical trading system. You need a sufficient number of observations to draw a reliable deduction, and that means at least 30 and probably more instances of a specific indicator’s buy/sell signals. You need to examine the performance of every indicator and then, to make matters hideously more complicated, you have to examine the performance of indicators combined. You may really appreciate the characteristics of a particular indicator but find that it does not merge well with others. This is a particular fault of average directional movement (ADX) and its many cousins, for example. 

The hardest part of buying or building a trading system — after answering the override issue — is making your money management rules fit your indicators. Indicators almost always have embedded buy/sell signals. Some do not, like bands and channels, but even the basic moving average crossover has, by definition, an embedded buy/sell signal. You buy when the spot price or a short-term moving average crosses above a longer one, and sell when it falls below. Anytime you have a changeable parameter like the number of days in a moving average (or any other indicator), you will be tempted to tweak the parameters to force the outcome you want. The problem with this procedure, named overfitting or curve-fitting, is that it may have been the optimum parameter for the period you are studying but will probably not be the best parameter for upcoming conditions.

If you love the MACD but using the standard parameters gives you entries that are too late for your trading style, you must decide between changing your trading style and abandoning MACD. A breakout trader will hardly ever get confirmation from MACD in the first — or second, or third — period after the breakout. If you still love MACD for its reliability in Forex, the breakout trader may consult it as a sanity check. Okay, you are taking the breakout and fading the trend, but your stop will be tight and your target will be modest. You will be prepared to reverse back to the primary trend on a dime. This is uncomfortable, but if your nerves can take it, it may work.

The key point is that your trading style — trend-following vs. breakout trading in this example — is not something you discover by soul-searching. You discover it by working with indicators and back-testing them. You may think at the onset that you are a trend-follower because you like the conservative sound of it, but discover you like more risk and should really be a swing trader. Or you set out to be a swing trader — such a wonderful term! — but discover you prefer less risk and trend-following suits you better.

Quiz:

1. The biggest problem in choosing or building a mechanical trading system is

2. You should always have more than one indicator to get

3. Which one is more correct?

Done