Build your own trading system? Hmm…
Not so long ago, designing a trading system was viewed as a highly technical problem that required talented developers, special background, and great skill. An individual investor would no sooner build a trading system than he would a refrigerator.
Advertisements for brokerage firms signal that this has all changed. Several of the leading discount brokers now bombard television watchers with the same message: You CAN do this at home. One firm explains how easy it is to develop and back test strategies with their online software. Another uses rotoscoped images to capture the indignation of some investors who are all smarter than their brokers. A third shows investors making smart moves while taking a minute away from running the restaurant or the construction site.
At "A Dash" we have tried to show the challenges in developing systems and interpreting data. Making powerful software simple to use and data more readily available just makes it easier for non-experts to lose a lot of money.
One graphic example is better than many admonitions from us. Check out the story of a rookie prop trader and his foreign exchange system, as reported by Tyro. Tyro’s friend was intelligent and methodical. He did a lot more work than most and took what he believed to be a cautious approach. After a month of awakening at 5 AM to test his system through paper trading, he was ready for the real show, going carefully with 2% positions. You can guess the result.
He did many things right, but it is so easy to go wrong. Go to Tyro’s site and read the entire well-told tale. Thanks also to Trader Mike for highlighting story.
Aspiring system traders should read (at least) Fooled by Randomness and the Portfolio Management Forumulas, the first book in the Ralph Vince "trilogy" (both featured on our recommended list at the right). These books will explain many of the traps as well as some good methods for testing a system.
If you do not find these books to be real page-turners, (we did), then system trading is probably not for you.
“going carefully” and “2% positions” don’t belong in the same sentence. It wouldn’t take a terribly long losing streak at 2% of equity per each loss to create an account-damaging and will-destroying drawdown.
Tested drawdowns, monte carlo simulation, and some odds being run on binomial distribution approximations for losing streaks of various magnitude are an essential part of system design.
Interesting story – while trading systems offer a consistent approach if nothing else, the task of creating one should not be approached lightly. One definitely has to be aware of assumptions made and performance monitoring required. While his story has the benefit of hindsight, many lessons can be learned.
With what seems like a martingale sizing mean-reversion strategy, to have 1:1 profit/stop exits one has to have at least 50% win rate to even break even… and if one is scaling in, this win rate has to increase even more to compensate. With many mean reversion strategies, if the trade diverges, one has to also look into fundamental shifts to see if an assumption is no longer true (RE: New Zealand announcement). Again, with assumptions, backtesting should be done in a very pessimistic way so as to be aggressive with slippage and spreads.
Finally, much of this possibly may have been avoided if he did a walk-forward test to see how the system performs on simulated money in realtime, in order to hammer out the kinks.
I would also add to the list, ‘Evidence-Based Technical Analysis’ by David Aronson and ‘The Encyclopedia of Trading Strategies’ by Katz & McCormick. I’ve found both to be very useful and relevant in the area of systems development.
Bill and Wilson —
Thanks for the comments. While I have not done much discussion of our own short-term trading methods, we obviously follow a careful backtesting approach and consider position size as a function of expected return and possible drawdown.
Wilson has outlined the method for doing this and provided some great additional references. I plan to highlight this issue in a future post.
Thanks again!
Jeff