I have begun researching common technical analysis based strategies. The goal is to determine what works and what does not work by back-testing each of these strategies and publishing the results. I will be publishing both the results as well as the NinjaTrader source code used to perform the tests. A forum is not geared towards posting this type of information, so I will be hosting the files and results on my blog: http://thestrategictrader.com This is a free blog with no ads. The first indicator I am looking into is RSI. I chose RSI first because its a common indicator found in many strategies posted on these forums. My hope is that through this thread we can spur conversations around my results as well as suggestions for other things to try. A summary of what I have tested thus far: #1 Entering a trade when RSI above a threshold. (Fails misserably) #2 Entering a trade when RSI below a threshold. (Produces positive results beyond what is expected due to random chance) #3 Entering a trade only on the cross of a threshold. (Profit Factor and percent profitable remain fairly constant. Only the number of trades changes) #4 Entering a trade when RSI above/below a moving average. (Above the moving average is horrible. Below the moving average produces good results but not beyond random chance)
To discuss the RSI we need to talk about it's formula. The RSI is backward looking indicator, as all indicators are. How can you devise something that looks forward with regards to the RSI. In other words, how can you devise something that will tell when the RSI will turn before the RSI turns?
The RSI is an oscillator indicator, which means that it's a leading indicator. It leads price. This is the principle upon which divergence is based. Where in the post did the author says that he cares about finding when the RSI turns before the RSI turns. We care about when the RSI turns before the price turns. Are you trading the RSI or the price ? Evidently, you are trading the RSI. Good luck with that.
Oscillator indicators trail price kandlekid, the formula itself will tell you that. Price will move well before the oscillator, but dynamics WITHIN the oscillator will move before price, focus on that.
I agree the RSI is not a leading indicator. Usually people look at divergences of RSI, not at absolute values. A very interesting analysis in the price action blog yesterday showed that the Dow Jones RSI as of last Friday is at levels that have been reached only 3 times since January 2000. Only small corrections followed and the trends continued as before: http://www.priceactionlab.com/Blog/2011/02/is-the-dow-30-overbought/
You need to be very careful that your test data doesn't show any bias e.g. EURUSD has a distinct upwards trend over the last ten years and so your positive results may merely be a reflection of a slight buy bias that happens to suit the data set. You need to ensure that you test across a range of assets that include net risers aswell as net fallers. If it still looks strong in these cases then it warrants further investigation. I'd be surprised if you find anything useful based on a single indicator though. I've searched billions of combinations over several asset data sets and can tell you that the only combinations that I've found that show any reliability rely on a combination of 3 to 5 indicators. Good luck though.
I appreciate all the feedback thus far. One thing I would like to point out is that the initial focus is not to build a strong tradeable strategy. Instead its to look at the common ways of using these indicators and determine what really works and what is garbage propagated in forums. Initially we are using set exit points either 5 bars or 3 bars into the future. The goal is to find if an indicator can be used to provide an edge over the next few bars. One such use has already been discovered. When the RSI is below a certain threshold, there is a strong bias to the upside that results. Whereas a typical usage such as buying when above a threshold fails miserably over the next several bars. This is an example of what I am looking to find out. As far as how this will be useful is another matter. After all of the basic ideas have been explored to determine what provides an edge and what doesn't we will start to put the pieces together to build a tradeable strategy.
RSI is simply the first indicator I will be documenting. As I move on to others, I will show combinations as well as single use. Concerning a upwards trend. Refer to this post: http://thestrategictrader.com/?p=20 My goal before testing any indicators was to determine the bias of my test data. Only when the results meet the criteria outlined in that post do I consider it to be anything more than just random or following the natural bias of the testing set. I would love to discuss this aspect more. Does anyone think my qualifiers need to be more strict or look at other qualifiers?
RSI is an excellent oscillator. Your RSI implementation fails because you apply it on price series which are non-stationary and non-gaussian processes. Apply RSI to a stationary process and you'll get excellent results. Ninna
To the OP: While I agree that those who say the RSI is lagging (or all indicators are lagging) are simply dead wrong (you don't get meaningful divergences with lagging indicators), I find the RSI itself to be a klunky indicator that is a poor substitute for better, more direct measurements of momentum. But it is popular so examination of its performance is not out of line. To Ninna: I don't know what good applying the RSI to a stationary process will do, because neither price nor any of its derivatives are truly stationary, although log returns have a sort of quasi-stationarity about them. Is that what you meant?