Previously we might have learned how to backtesting trading a strategy using a simulator, this time we will talk about how to read the backtest results of a strategy that has been run before.
Testing a system or forex trading strategy, or Expert Advisor (EA), commonly known as a "trading robot", can be consuming a very long time. But fortunately there is a tool called the Strategy Tester in MetaTrader, so we can test a trading system by using data recording price movements from the previous few years. Thus we do not need to test a trading system in real time. This testing method is called "backtesting", which can speed up the process of testing a trading system significantly.
Here we recommend testing with the backtesting method before testing in real time or commonly called forward testing. Of course, after backtesting is executed, we still need forward testing in the real market, but the purpose of backtesting is to provide initial information about a trading system that we will test and then - if the results are satisfactory - we will more confidence to use it.
What needs to be remembered is that to get quality backtesting results, we need a record of price movements that are of good quality to avoid overfitting. Overfitting is an "error" caused by limited data used. Therefore, to do backtesting we should have a record of price movement data at least three or four years before
After backtesting, we must analyze the results of the test so that we know whether the trading system or EA is what we want or not. There are several parameters that we must see and understand from the results of the backtest.
Now let's try to look at the following example of the backtest results:
The following are some key parameters and what we can know from the trading system.
Total net profit
This is the total result of all transactions carried out by the trading system that we are testing. A winning strategy should generate profits, not losses. This parameter tells us what can happen if you use this system in accordance with the rules attached to it. Remember, this is not a guarantee that in the future this system will work exactly as the test results. However, at least this parameter can show how much profit we can possibly get by using this system.
Maximum drawdown shows the biggest loss that might occur. This number should not be too large. Of course, big or small is relative, depending on how much your capital is and how much your risk tolerance is. For example, if our capital is $ 10,000 and our risk tolerance is only 20% of the capital, then the maximum drawdown should be no more than $ 2,000. Simply put, the smaller the maximum drawdown, the better.
If we pay attention, the report has provided a more detailed description of Maximal Drawdown. In plain view, without us having to look for more details, Maximal Drawdown looks clearly when we finish backtesting EA.
When the maximal drawdown value starts to show the biggest limit intolerance provided by the user in accordance with expectations, then EA is feasible to be used in real transactions. Vice versa, if the maximal drawdown is still too large and not in accordance with the limits provided by the user, then EA needs to be reviewed until EA really matches the user's expectations.
Maximal Drawdown can be seen not only on the report during the test. If you saving the report, the results can be read on the graph image in the save report. This review is considered important because maximal drawdown is always the cause of the level of risk which if ignored may cause your capital to run out is not left.
Maximal Drawdown Advantages or Weaknesses
From the description above, we have got an idea of the meaning of maximal drawdown in EA. well, here we will see the function besides reading the level of risk, what are the advantages and disadvantages of this Maximum Drawdown.
The advantages of Maximal Drawdown are as follows:
a. Able to know the risk level early
b. Can be categorized as an indicator of the level of loss
c. Can see the Equity spent by EA
d. Help prepare the right capital for EA
The disadvantages of Maximal Drawdown are as follows:
a. Can not be used to find a drawdown in the future
b. Only able to estimate the current loss limit
c. Will change if there is a greater Maximal Drawdown value
Maximal drawdown is closely related to the risk limit problem that EA does. This limit is used as an EA as a reference in handling how good is EA to overcome price movements to completion. At the same time, the user will observe how much capital is in accordance with EA if Maximal Drawdown experiences the greatest risk to protect the balance from losses.
This is the total transaction made. We can analyze whether this system is aggressive or not by comparing this number with the time period we use.
For example, if the total transaction that occurs is 200 transactions in three months, it means that this trading system generates an average of three transactions per day (assuming that there are twenty working days in a month). That is, this strategy is not too aggressive and may be suitable for intraday trading strategies. If the numbers are getting smaller, then it might be more suitable for medium-term or maybe long-term trading strategies.
This parameter shows the number of transactions that generate profits and what percentage of the portion of all transactions made. Simply put, this parameter shows the win ratio of the system we are testing. The bigger the number, the better.
As mentioned above, backtesting should be followed by forward testing in the real market. It is necessary to validate all the results we get from backtesting, it is also important to get clear evidence that we have managed to avoid overfitting during the backtesting process.
The backtest method can help us to speed up the testing process, but make sure you have a record of price movements that are of reasonably good quality.
After reading the results of the backtesting and the results giving confidence to the trading strategy or trading system used during the backtest, the next step is to do a forward test, this is also important because what is produced during the backtest does not guarantee the results will be the same as using real accounts.