I guarantee that you are not the first trader to consider the idea of reverse engineering an expert advisor. The idea pops up most frequently among traders that don’t want to pay for a commercial expert advisor. Alternatively, they want to use the same strategy when another trader that doesn’t want to share the idea. The motivations for reverse engineering remind me of decompiling EAs, making me leery of the idea.
Reverse engineering a strategy only stands a chance when some parameters are known about the strategy. If, for example, you knew that the strategy involves MACD and moving average crossovers, it at least provides a reasonable starting point. The programmer could write software which combines every possible combination of two moving averages of various types with every known type of MACD. The programmer could then make guess about which types of signals might result in a buy or sell decision. If the guesses are not very good, then the outcome of the reverse engineering attempt is certain failure.
Then you have the problem of guessing on which chart to base the decisions. Many traders use standard charts like the M1 and M15, but many others use less common options like an M3. If the trader uses multiple charts like an M2 and M10, the resulting trade history would clutter together. Good luck trying to pry apart the different series.
Making things worse is if the trader uses a chart that doesn’t depend on time at all. Tick charts are the most common, although you occasionally see less orthodox options like Point and Figure charts and Kagi charts. Time is irrelevant. You wouldn’t have any idea on which charts to run the test.
This approach of making somewhat intelligent guesses while throwing mud at the wall is called a brute force attack. You literally designate every unique possible combination, then see which one opens the metaphorical safe. Some results will bear no resemblance at all to the actual results. If you get extraordinarily lucky and/or have fantastic intelligence, then you might find a close replica of the strategy.
It would be possible to study the correlation of the tested values with the values in the supplied account statement. You would ideally want to find data clusters with similar variable settings that do not dramatically alter the correlation between the reverse engineered strategy and the actual account statement.
If you don’t know very much about the strategy, or if what you think you know turns out to be wrong, then you stand no chance at all of reverse engineering the expert advisor. For all that you know, the EA that you thought used RSI might turn out to run on phases of the moon (yes, there are real strategies that do that) or that make decisions based on coin flips.
Most people assume that they know more than they really do. I would discourage all but the most fool hearty or stubborn from making the attempt, unless you had a very good reason for doing so.
