Most people buys and runs a given strategy just looking at fantastic equity curves presented by the vendor, expecting a holy grail that will make them rich in no time. If you are reading my blog, hopefully you already know that earning money using automated trading systems don't work this way.
The main steps before trading a strategy live
- Identify if the strategy or Expert Advisor meets the basic requirements: This step will prevent you loosing your time as most of the commercially available Expert Advisors are not able to pass this step. Some things that quickly invalidate an EA for long term profitability are running on a low timeframe, not trading all signals, requiring tick data to get a good backtest, not surviving to a 12 year backtest, having a lot of indicators, trading before the current bar is not closed, a straight up incredibly pretty equity curve on backtests, ... I will go deeper on this in subsequent articles.
- Accurate backtesting: An accurate backtest needs to be done in a specific backtesting platform without network connection to the broker and good data. It requires you to control the spread too (the difference between Bid and Ask prices, which is part of the trading costs).
- Robustness analysis: On this step we analyze the parameter space to identify curve fitting issues. Needless to say that strategy parameters has to be available to analyze the optimizations made. On this step we have to make a basic Walkforward Analysis, which can be done very easily in Metatrader without any additional tools.
- Statistical analysis: If we know that the strategy is robust and that the backtest is accurate, then we deeply analize the statistical behavior of the Expert Advisor.
- Trading decissions: On this step we decide the amount of money to commit, the risk level, whether it will run in a portfolio or alone, we will document important information such as statistical boundaries of the strategy and what are the conditions to stop trading the strategy.
- Test on demo for some trades: To avoid technical problems that could not be present on backtests it would be wise to run the strategy on a demo account for 15-20 trades. Coding problems such as not reacting correctly to "trade context busy" or "requotes" or any other technical problem should raise on this test.
- Perform a backtest of the tested period: A backtest of the 15-20 trades we took on demo might give us more information about the reliability of the backtests.
Minimum requirements to make money with a trading strategy
- A deep understanding about the strategy, so you will not ask yourself things like closing trades manually on weekends, filtering news and the rest of questions that arises as a consequence of not understanding what you are trading.
- All trading decisions already taken such as risk level, when to stop trading, ... so in the middle of a drawdown period you will not loose faith in the strategy. You will just look at it from the distance from the pure statistical point of view knowing where you are in every moment.
- There are no obvious technical problems, although there could be other not so obvious. If you are the one who programmed the strategy, there are some good practices such as unit and functional testing that can reduce a lot the errors in the code.