Hello traders. I'm attempting to put together a statistical model for intraday trading. Basically, it's statistical sorting based on various parameters and metrics that I've compiled. So far, I got quite a bit of stuff to look at. In fact, I'm wondering if I perhaps have too much. Still tinkering around with it and trying to figure out what's relevant and not. I'm curious if there are any resources, either online or in books, where I could perhaps get some further inspiration or ideas on the topic? FWIW, my starting point was actually stuff I read about in Brett Steenbargers blog and in his latest book. I also seem to recall reading some good stuff on the Price Action Lab Blog, but I can't really say for sure. TIA.
I'd typed up a basic response talking about excel, but it seems you've already gone down that route in the other thread you start on stats. I take it you didn't get into MATLAB? If you want to go deeper than excel, then you'd be looking at either Python, Matlab or R for your analysis and models. I like Python personally (Pandas, SciKit), but R is meant to be great once you've got your feet wet. Just on excel: Adam Grimes has put together a fair amount of stuff on stat analysis using excel. He's also contributed a fair number of webinars at another website that's just changed its name (don't know if I can mention it or not). There is a brilliant thread on stat analysis in the Elite Section of the same website something like the homework thread from memory... http://adamhgrimes.com/blog/basic-backtesting-in-excel-graphs-and-first-calculations/
don't know of any online resources on the topic I could point out to you, but I did have a couple of questions. What instruments are you focusing on for your intraday testing? Stocks, futures? What software or language are you using to do this? I recently purchased intraday historical data for US stocks and it's on my todo list to do comprehensive backtesting on this dataset. My plan is to load all the intraday data into an Access database and then write a backtester in some language like VB.net or R which interfaces with that database. Curious to hear some specifics on what you are doing if you are willing to share.
Thanks a lot, man. So far, I've only compiled my metrics manually in Exce on an EOD basis. Very cumbersome and time consuming of course, but at this point, I'm just focusing on developing my model and figuring out stuff. I'm thinking perhaps I can hire someone to write out the code for me later on. I'll look into that site.
Currently, I've been looking at index futures and I haven't gone beyond Excel at this point, but I suppose I may need a database and possibly some other tool to automate things. I don't want to be an asshole or anything, but I've spent quite a bit of time on this, so I'm very careful about revealing stuff. I hope that's understandable. But I can say that I'm trying to look beyond merely OHLC data and code my dataset according to various parameters. For instance, I divide my data set into gap up and gap down days. Another parameter may be if the gap is closed that day or not. That further divides the dataset. Of course, I'm also looking at a lot more than just that, but that's where I'll keep my cards to my chest. Sorry. ; ) It's not mathematically challenging, certainly not at this point, but it's definitely a cognitive and logical exercise.
How nice would it be to use Matlab or similar matrix based manipulations to have price data represented in the frequency domain? You may be in the wrong place
I hear ya. No, I didn't expect you to show your hands. I'm also interested in including some type of stats into my current strategy but I question whether you can formulate a "pure" stat without any ties to price action. For example, "buy when price crossed above MA(20)" or "buy when close is above close of yesterday" are PA-based. But "buy when the mean or median of 20-period is 2% above close of x-days ago" might not be. I'm mostly interested in formulating volatility/momentum into the equation. Any suggestion for this noob?
There are numerous books on the topic. Just search for statistical learning or statistical pattern recognition or (nonlinear) time series and you will find multiple books (Amazon) and web pages. Unfortunately most of them are not in human readable form due to the fact that academia tends to make relatively simple things complicated for outsiders. I found that some of the books that are written for R or Python users tend to be more reader friendly. I guess machine learning also includes statistical learning since it is general term and you can include it in your searches as well. For basics in machine learning I would suggest Machine Learning in Action. There are relatively cheap ebooks on Amazon and elsewhere for few $ like Introductory R: A beginner's guide, 100 statistical tests in R and others; just to give you some ideas. There are also books addressing financial markets directly like Python for Finance (there are others like this for R, #F, Haskell, C++) and some of them include statistical concepts applied to markets. I am not recommending all of them since they may not suit your needs. Some are not that good except for few more or less useful chapters or they might be too difficult despite the fact that those were written for people without advanced degree in math or statistics. I think just searching and skimming through table of contents and few pages available for preview gives some idea if it is useful and as a bonus it can narrow your search criteria. Unfortunately I cannot recommend single best resource and I am not an expert in statistics nor I apply it directly to market data but rather use it as a set of utility tools.