Back to Community
What constitutes edge / why is my system working?

Some background: I am no quant PhD, and certainly no hedge fund guy. I am just an engineer with decent programming skill.

I have spent a small part of the last 2 years developing and building a pretty simple intra-day trend-following system (fully automated) focussed on currencies. No fancy indicators, just signals based of when price exceeds a simple moving average. After a bunch of backtesting and paper trading, I'm seeing that it actually makes good return and I just started live trading with it.

BT results: Sharpe > 1.7, with tight stop losses.

While I was happy, I was also quite suspicious to see this working as well. Can someone help me intuit how such systems work?

  • What are the market forces making this system work?
  • Is this effectively taking money from less sophisticated traders? Or, am I riding the wave with smart money who just don't mind taking smaller losses which are actually quite meaningful for retail traders?
  • Why haven't hedge funds/smarter people arbitraged or sucked the alpha out of my system?
  • I've read Ernest Chan's books and it seems that this might be related to 'capacity' but I am still not sure why it is actually working.

Thanks in advance.

4 responses

Intelligent question...Sridharan
..maybe we just see a small part of the full story and, this part is in gain now..

I'm very interested to know other users opinions about your question!

Hey there, do you mind sharing some more about the specifics? Here are some of my questions:

How are you backtesting your algorithms?

What the are assumptions and parameters of your backtesting system (trading cost, spillage, leverage constraints)?

What metrics did you use to evaluate performance?

Where are you live trading the algorithm?

I'm backtesting using a vectorized backtesting platform that I built. Sure, it isn't event driven but about a year of paper trading with out of sample data gives me some additional confidence in it. Optimization metrics used were Sharpe, Drawdown, Win Rates, Risk adjusted Return. Some slippage was modeled, as was spread for currencies. I didn't quite run Monte Carlo simulations but tested rigorously with out-of-sample data and it seems to be fairly profitable during non-trending timeframes. I'm live trading by hosting my code on a VPS that's colocated in my broker's region.

Cool, cool!

Over what time periods did you test in and how did it specifically (2010-2012, 2012-2014, etc..) do? What where the max draw downs, sharp, beta, etc... . Different types of markets exist over times and perhaps your algorithm in tuned to a particular one? I've had OK success with some very simple algorithms in backtesting on one set of years and then not so good performance over another set of years.