This is another “trick” post to force me to step back, look at the several projects I want to work on and commit to them publicly (accountability again!).
For you readers, this post should give you a good idea of what’s coming in the blog… In no particular order:
This is my most “attention-grabbing” project at the moment. Coming to that realisation drove me to step back to reflect on why I write the Au.Tra.Sy blog and on other projects that matter. This will help me stay focus on the goals I aim to achieve with the blog (and avoid getting bogged down on less relevant mini-projects). For now, I will focus on updating the blog with progress and posts on developing automated trading systems – as well as developing the blog a bit further (current plans are to start a library page of useful book reviews and a “Trend Following Wizards” compilation of fund performances).
Trend Following System
I have wanted to build an automated trading system for a long time and have read and though quite a lot about different type of systems. I believe they can all be classified in a handful of groups, of which trend following is a very succesful one. A read of Trend Following by Michael Covel will provide you some evidence and potentially convince you of its effectiveness (traders track records, psychological/behavioural explanations as to why trend following works).
I have therefore decided to research an End-of-Day Long Term Trend Following system. This is my primary objective. The ideal steps to reach that objective will be:
- Research and develop satisfactory system
- Implement an automated interface to my broker trading the system
- Start trading the system with proprietary funds
- Build an audited track record
- Raise external funds to manage with the system as a CTA/CPO
DeMark indicators seem to be esoteric and have a certain aura about them. My initial thoughts were that one could use some of these signals to improve a trend following system. For example, TD Sequential is supposed to identify trend exhaustion. If used correctly it could be a way to improve exits for a trend following system by bailing out of the trade before the trend bends and eats away a big chunk of your open equity. I want to explore these indicators (as a trend following “add-on”) and possibly code them up for back-testing. So stay tuned for more posts on the intriguing DeMark and his indicators.
Systems Research Framework
I believe that Systems testing is too often reduced to measuring the “performance” of some entry/exit signal. An integrated approach to Systems research should include standard procedures for back-testing, variables such as portfolio allocation, position sizing and leverage, test results analysis.
This includes more “thinking and reading” rather than “doing”, which is always challenging. However I do feel it is an essential foundation for the overall process which will prove an invaluable investment. This encompasses the 3 following areas:
My first action point on that list is to finish reading Ralph Vince’s Handbook of Portfolio Mathematics which provides an extensive coverage of Money Management including optimal leverage and allocation using his Optimal f formula and Leverage Space Portfolio Model (I have updated the main page to reflect what I am currently reading. If the same book stays there too long, please nag me!).
I believe Money Management is an essential key to success as well as an integral part of an automated trading system. This has been acknowledged countless number of times by top traders and I want to beef up my knowledge of it.
Here, we need to make sure that we are testing in a robust way. A way which will give significant result analysis to decide whether the system is tradable or not. This might include procedures such as Walk-Forward analysis of other methods to avoid classic pitfalls such as over-optimisation, curve-fiting, historical biases, etc. In effect, for any new systems test, it should run through a well defined set of standard procedures which provide impartial results.
Back-testing trading systems generate a myriad of test results including large sets of numbers. Statistics are the best tool to make sense of these numbers. Interpreting back-testing results needs to be subject to robust statistics if we want historical results to have any predictive value. I will be re-educating myself (did lots of Maths when I was younger but I need a refresher) on standard statistics and probability. Areas I would like to explore further are non-parametric statistical inference, exploratory data analysis, John Tukey, etc.
TradersStudio back-testing software
I have purchased TradersStudio as it seems to be a very good alternative to TradeStation and other softwares available. This will be my main tool for back-testing trading systems. I need to learn how to use its programming language to code up custom systems and indicators, which I will make some available on the blog.
Real-time automated trading system (using Interactive Brokers API)
This is a secondary project but I would like to have robot trading in real-time (intra-day) by connecting to my broker’s API. The Trend Following system will require a platform connecting to Interactive Brokers so this should be an extension of it. This is more a Technology project at the moment but I would really like to trade some shorter timeframe systems on a virtual account. I have discussed with several traders how they use TD indicators and how their signals were effective for intra-day trading and this is something I would like to test.
This is the current state of the work stack. Some streams will have to run in parallel but the priority is to design a framework for Systems research (foundations first!). These projects will all be documented in the blog as they are being worked on.