Disadvantages of Expert Advisors
Following on from my advantages of expert advisors post here are the disadvantages of trying use expert advisors for automated forex trading.
Curve fitting
Because it is easy to backtest expert advisors many times with different parameters it is very easy for them to suffer from curve fitting.
Curve fitting means that the EAs are highly optimised for the specific period of time for which they have been tested against, but may turn out to be loss making if run on actual live data.
A good expert advisor will not need to be micro-optimised in order to make a profit. There may well be certain values that perform better than others, but a range of values around the optimal ones should also be profitable to prove that the expert advisor isn’t just working because of some very specific circumstances that are unlikely to occur again in future.
As well as curve fitting the parameters it is very tempting to add new code branches into the expert advisor to filter out bad trades and allow though good trades. Again this is likely to produce an expert advisor which is highly optimised only for the historic data on which you are testing it.
Curve fitting over-optimisation is a very easy trap to fall into when writing expert advisors. It is a waste of time, as you won’t end up with a profitable expert advisor. Spend that time on writing quality expert advisors which can handle whatever data you throw at them.
Unrealistic backtests
Many pieces of trading software that allow you to backtest EAs do not allow you do it in a way that realistically simulates the real market. Doing so would be very difficult because there is a lot more information you need to know other than the price. E.g. you need to know about the spread, execution delay, slippage, and liquidity.
Backtesting data does not tend to give you access to this kind of information. In addition the granularity of backtest data is often much less fine than what you would get if you were trading live.
Most backtesting data is only available for 1 minute and above timeframes. This means that you miss out on the tick-level information that could be essential, especially if your trading strategy works on shorter timeframes.
Many backtesting software packages compensate for this by ‘filling in the gaps’, and create simulated ticks. This can be dangerous, as the algorithm used may well not generate ticks in a way that is realistic enough for your strategy.
It is quite possible (in fact easy) to create an expert advisor that works brilliantly on historical data using a backtest but consistently loses when used for real.
Difficult to write
Translating a working strategy from your brain into an expert advisor can be very difficult.
When you trade manually you can interpret huge amounts of data and news in a way which can be very difficult for a computer program to do. Interpreting simple facts such as the value of an indicator is easy for an EA to do. However interpreting something such as the current economic climate or the latest news that has just been released is extremely difficult.
Getting EAs to recognise complex chart shapes and patterns can be very difficult. It is all to easy to program the EA to recognise specific patterns that occurred in the past rather than getting it to understand the meaning of the chart patterns in the way that your brain will.
Expert advisor bugs
All software beyond trivial code snippets will contain bugs. For most pieces of software the consequences won’t be too bad. Maybe the program will crash, get stuck, or if you are unlucky perhaps you will lose some work.
Expert advisors for forex and CFDs are somewhat different as they have the power to make decisions that affect your bank balance! If your EA has a bug then it could make the wrong decision resulting in a loss of money. If the bug is bad then maybe you’ll lose a lot of money!
Would you trust a piece of computer software to buy and sell on your behalf? Even if you have done large amounts of testing how can you be sure that your EA won’t trip up when it is left to run by itself?
As EAs are often left to trade by themselves, any bug or defect in the code could be very costly. You might not spot that anything is wrong until you find your account has been emptied of all its money.
Think carefully before you let an EA lose on your live account. Make sure it is well tested, and keep a careful eye on what it is doing. To avoid it losing too much in the event that it doesn’t work as expected make its trade sizes small to start with and don’t keep to much money in your account.
Conclusions
Before you spend time creating an expert advisor make sure you don’t fall into any of the problems listed above. Doing so could cost you serious money.
