This article is the second in a series of three articles providing some tips on building successful trading models and systems. This article will discuss: simplicity vs. complexity, the importance of backtesting, and putting your strategy into action.
The general overview you should take towards building your trading models is to keep things simple. The less variables, indicators, or conditions you impose, the more flexible your trading models will be. Many traders make the mistake of adding more and more variables to their trading models, believing that it makes their models better. But the irony is that strategies that have many variables, and are very complex, tend to give poor trading signals, especially as the underlying market conditions change. We know the market is like the ocean, constantly moving and changing. If your models are not flexible, they will break when the market changes. Keeping things simple gives trading models more flexibility, it makes them more robust, and better able to handle the ever-changing market.
The next step is to backtest your trading strategies and models. There are many ways to backtest, but here are some things to keep in mind. First, do some small, random, sample backtesting, Monte Carlo style, and look at your results. If they look positive, then expand the backtesting as much as possible. It is very important to test your trading models across a variety of time periods, market conditions, and market directions. You need to make sure that your trading strategy is robust; it must be able to handle up, down, and sideways markets, and low, middle, and high volatility situations.
Sample size will also matter when you backtest your strategy. In general, the larger your sample size, the more confidence you can have in the results. Never trade a strategy that is based on small sample sizes. One rule of thumb that can be used, statistically, is that the sample size should contain at least 30 observations in order to have some confidence in the results of the study. My general feeling is that 30 observations is still small. Another concept I have come across is that it may be useful to have 30 observations for each variable in your trading model. This can still be misleading; ultimately you must use your own judgement about sample sizes and testing in order to have confidence trading your models. For example, even though you have a large sample size, if it is only from one time period, or from one year, then you probably should not have too much confidence in those results. It is not only the size of the sample that matters, but the variety of periods it is drawn from. Keep these ideas in mind when backtesting your strategies.
Once you feel comfortable with your backtesting you can begin to paper trade. My feeling towards paper trading is that it is ok to do, but only for a little while. Do not spend too much time paper trading. Paper trading results can be misleading because you may be misinterpreting entry and exit prices due to slippage, etc. It is much better to trade with real dollars. My suggestion with real dollar trading is to initially trade small position sizes. You do not need to begin trading a new strategy with maximum dollars and maximum position size. The reason you want to trade real dollars is because when there is money on the line you will focus better and learn quicker. In addition, having real money in a trade teaches you about your own personality and what you are like when you trade. In general, the information you will pick up trading with money is better than the information you will gather from paper trading.
In closing, by applying some of the ideas presented in this article you will significantly increase your chances of building successful and winning trading models and systems. Keep an eye out for the last article in this series.