Has anyone ever taken all the historical data of a pair and put it into a database, then written a program to go through it all with a certain strategy, report its results, change the strategy slightly (i.e changing the SL/TP or the times of the day trading happens) and repeating until they seem to find a proven strategy?
Do a lot of people do this? Or is this pointless because the market is too random? Or does noone do this to this extent because the processing power is too massive?
Of course I know that this idea isnt some sort of way to find a stategy that will guarantee profits forevermore… Just curious if anyone has any anecdotal evidence of this sort of approach
You’ll find this referred to as “back-testing”, and sometimes as “data mining” when someone wants to make a point about curve fitting. Either way, it’s an extremely common practice and their are whole businesses built around allowing traders to do exactly that. I’ll disagree with purplepatchforex on it being a waste of time. Researching different types of trading systems and methods can go a long way in helping one understand a lot of things, including what style of trading one is best suited to.
To each his own about whether or not back-testing is useful. Personally I am trying to force myself into more discretionary thinking and less mechanical thinking. Back-testing reinforces mechanical trading, and I would argue sometimes to a fault. The market itself is dynamic and always evolving. So to me, it makes sense that the trader also should strive to be dynamic and able to adapt to changing market conditions on-the-fly (discretion). One way I try to do this is to invent a baseline for the trading method. Find VERY basic things that work, things that I can trust. Then build on them. The things should not be rigid but instead fluid wherever possible. My goal is to master a dynamic discretionary method that I can someday take to any trading instrument or market, because it won’t matter. As such, back-testing or optimizing also would be irrelevant.
(Of course this is all theory and I could be totally wrong about it.)
I was aware of back-testing, particularly in testing theories that you already have, though I guess I was more curious to know if anyone had used software to find the patterns for them… It’s probably because I’m from a Computer Science background (rather than any kind of mathematical/statistical one) that I’d like to plug all the data into a program, set a couple parameters and let it tell me that EUR/GBP rises between 10:05am and 10:10am every wednesday more times than it falls, or something else as random.
Though I really can’t see it coming back with any valuable information I am still curious if anyone has attempted something like that
If one back-tests strictly with the plan to created an automated system then sure. If, however, one back-tests to see how different patterns, indicators, ect. work under different market conditions (the dynamic stuff you talk about) then it does anything but reinforce mechanical trading.
This is the data mining I mentioned, and it definitely gets done all the time. Neural Nets are sometimes used. I’ve personal done research on forex seasonal patterns (statistical, not NN).
