Automated Futures System development is a stepped process involving theorizing a conceptual hypothesis to see what can be exploited quantitatively. The details involve strategy identification, trade pattern recognition, tools that can measure or identify those patterns, strategy development and finally testing and weighing against your hypothesis.
When looking to develop an automated trading strategy, the first step is to identify what you want to exploit, or patterns which you have identified to be true in a discretionary sense. Most trading strategies are developed around an implied truth that is identified by human intuition such as noticing price behavior during certain events. An example could be that you notice price will decline or retrace if the price becomes overbought for five consecutive periods and volume average declines over those five periods. Once the price declines as intended, if it moves below a 10 period moving average you will take profits. With that hypothesis, you have noticed through your visual analysis of the chart that this is a general truth and something you would like to exploit.
Now that you have a hypothesis and general strategy idea, the next step is to think of which tools can be used in conjunction with the strategy that can identify the values necessary to automate the trade idea. For this example it could be something such as a stochastic, RSI (Relative Strength Index) or MACD (Moving Average Convergence Divergence) that can identify overbought conditions and a volume average that can measure the relative decline of overall market activity and lower volumes. A simple moving average would be used as your profit target. These will be your logical tools that form a function of identification for your automated strategy.
The next step is developing your system using logical operators and the tools identified that can measure the factors within your idea to form the strategy itself that can operate trade functions without human interaction. If you are unable to code the strategy yourself in the language of your platform of choice, many custom strategy development companies exist that can help out.
Once your strategy is complete, it comes down to testing. Testing of a strategy is the most important step in the process of developing a successful strategy. Once the system is run on the instrument/s of choice, you need to pay close attention to how your hypothesis is reacting to the tools being used. Is the strategy consistent? Is the representation of theory accurately being reflected? Are the values being used on your indicators optimal? Is slippage and commission being factored? And most of all, is it profitable?
At this stage, you may have a good sense as to whether or not your strategy will be successful. If it shows profitable, but only slightly, what can be done to improve the strategies ability to capture more? Perhaps the strategy exits prematurely and you notice that the price consistently declines more once you exit, giving up further profits. This future development crosses into the optimization of a strategy, this is a very fine line that can drastically improve a strategy or force it to post unrealistic expectations and performance moving forward.
While there are no black and white rules to strategy optimization, here are some rules of thumb to keep in mind to avoid overfitting. Overfitting is optimizing your system parameters and functions to perform best case trading scenarios during your testing period.
• Is your strategy relatively consistent with minor parameter changes? If a value of 10 works well on a certain indicator/parameter, but a value of 9 or 11 does not, It will likely not be profitable moving forward and is fit to the data. A parameter value refers to the input you place in your indicators and/or price action calculations. An example could be a 20 day Moving Average, in this case the 20 is the value.
• Does your strategy work on other instruments? While you may not intend to trade on different instruments, it is wise to see if your strategy is capable of being profitable across an array of markets and market conditions.
• Are there any performance outliers that greatly boost the overall systems profitability? Circumstances that occurred in the past may not and likely will not replicate themselves moving forward, performance outliers especially.
• Is your equity curve stable? A good strategy should perform with relative consistency over years of testing data. An unstable curve suggests that high drawdowns can occur in the future.
The most difficult part of strategy development is conceptualizing something to exploit. Noticing exploitable pattern takes a lot of experience analyzing different charts and patterns. Thousands of strategies exist and many of them do not work or only work with curve fit historical data. The markets are ever-changing and the landscape in which a strategy once worked may not work again in the future. Consistency of performance and ample testing data can provide more confidence to the overall success of your automated trading strategy looking forward.
There is a substantial risk of loss in futures trading. Past performance is not necessarily indicative of future results.
VFM 29671
