Hi folks, hope you are well. I'm quite new here. I am in the process of creating some signals of 30 FX pairs and am increasingly frustrated by the fact that the process does not seem to produce edge on all pairs in all directions when tested on new data. I start to wonder if the edge that seems apparent on some pairs may just be incidental to the dataset that I am using; is it real? But then it would make sense that some instruments are not sensitive to my filter for one reason or another. I suppose I need to answer the question; why does it work for some whilst it not for others to be fully satisfied with the results. What are your experiences in the process of generating edge, do you also have a mixed bag of results and how do you proceed when that happens? All comments welcome. Thanks.
Depends on how you are defining your edge. Mine works on all pairs and markets. Of course, No system or trading an "Edge" works 100% of the time. Expect loses on some of your trades and of course Trade Management is important to how profitable your trades will be.
Hi, thanks for the response. I define an edge as an increase in the break even win prob of 10%; which is my target. So when I test the filtered trades on new data, I expect to replicate that result. Works well on most but not some and makes me wonder.
Do you consider Duration that you will be in the trade to determine your edge. Someone who finds trades on say a daily chart or hour chart will need a different Target and Stop that someone who is employing say 5 minute charts for determining a setup and employing what they consider an "Edge"
So I trade exclusively on the H1 candle, for simplicity of executing multiple strategies in one chart and, H1 as the lowest timeframe (for more signals) with minimal signal noise. So I run 3 strategies at the moment and the stop loss and take profit are fixed dynamic, by that I mean they are a fixed percentage of the changing feature of standard deviation. Usually a stop loss will be 2 sigmas from the current price and take profit will be 1 sigma. The duration of the trades vary from 2hrs to 48 in general. So far I have found the trade durations to be uniformly distributed, but I will like to work on a better model for trade duration so that I can estimate trade times better and therefore, better quantify the correlation risk when trading multiple pairs simultaneously. I think the proper model for something like that might be some form of monte-carlo simulation on the correlation of the open trades. That is the correlation of the change in profit for 1 lot of each active pair. So obviously the exposure time; how long those pairs are all open simultaneously would be an important variable.
Suggested Read: "Trading In the Zone" by Mark Douglas. Should find it for Free on Google for both a PDF and an Audio File Version. That book points out the Key Points for Finding your Edge. If you wish you may contact me by PM on this and I will assist further in finding a Free copy of "Trading in the Zone" Also, Harmonic Numbers or some might call, Cycles or Market Maker Numbers may be of help. You may consider trading using Hedging instead of Stops.
The edge that you found is heavily curve fitted. Therefore you have to assume it is a much lower edge in reality. Also, nothing works all the time as movements of prices changes.