Tuesday, May 26, 2009

Short-Term Trends In The Stock Market Are Not Your Friends

I thought I'd quickly follow up on the "Why Short-Term Traders Lose Money" post that I recently linked via Twitter.

Data come from the excellent Barchart site: We're looking at the S&P 500 Index (SPY) going back to late 2003. Our system buys SPY if it closes above the two standard deviation Bollinger Band surrounding the 20-day moving average. The system sells SPY if it closes below that band. We exit the trade once the average moves back within the envelope defined by the upper and lower bands.

The logic of the system is that we wait for a significant trend--one that is a two or greater standard deviation move away from an average price--and then jump on board.

The average holding time per trade is 3 days. The system gave us 105 trades since late 2003. Of these trades, 70 were long and 35 were short.

Of the 70 long trades, 25 were profitable and 45 were unprofitable. Of the 35 short trades, 14 were profitable, 20 were unprofitable, and 1 was scratched. In all, the system gave us 39 winners out of 105 trades--about a 40% winning percentage. The gross P/L for the system (not including commissions or slippage) was -39.70 points, or almost 400 ES points. This is because the average size of the losing trades was larger than the average size of the winners.

I in no way suggest that this is an indictment of Bollinger Bands. On the contrary, they appear to be a decent starting point for a winning system if one trades against the trend. Rather, the lesson is that once a trend becomes "significant", it is already long in the tooth. If we simply follow human nature and extrapolate the recent past into the present, we will be well on our way toward losing money consistently.
.

3 comments:

Daniel said...

This is a really great, illustrative study!

An interesting additional use of such studies is to keep an eye on them, by running a periodic equity curve on them.

(Much thanks by the way to Rob Hanna of Quantifiable Edges, whose methodology this is... the use of equity curves on "dumb strategies" as advance Warning indicators.)

As long as the dumb system shows the expected losing-curve slope it’s of marginal interest. However, if at some point the dumb strategy should start working (meaning the equity curve suddenly kicks up a significant amount) it can serve as a HEADS UP warning that some (temporary but lengthy) unusual period of time is ahead. Every dog will have it's day.

Dr. Brett’s study is of a system that can only make money in certain specific environments. Pop and snort environments, with sharp breakout moves by
many datapoints up against major resistance.

For example, like NOW. Many sectors and markets right at resistance. If all or some start popping thru, the result might be multiple and profitable BREAKOUT moves.

I'm not saying this will occur, but rather that this might be the meaning of an upside flip to the normal downward curve of the "idiotic strategy", if such a flip did occur. A litmus test saying, yes THIS IS EARLY BULL, and not a late BEAR headfake move. An environmental confirm that the next multiweek move is up rather than down.

Not all dumb strategies potentially work this way. Only ones where an infrequent extreme condition is exploited by the dumb system, in its few successful outcomes.

Daniel

Curtis said...

A key lesson here that is to be very wary of datamining a chart by eye.

On trends..

The novice trader does not know what a trend is, and he is better off for it.

The advanced novice learns to recognize trends.

The beginner learns to recognize multiple trends.

The advanced beginner learns to move before a trend.

The proficient learns to see infinite trends in all time frames each with an equal probability of developing.

The competent learns to identify what the market is doing in the present.

The guru he commands the market with his voice alone.

vmn said...

Stats are worth what they are worth. I'm short-term convinced and won't take your article for granted until i have access to full orders. There's definitely a problem somewhere. You said Bollinger? lol, this may explain that.

Trade & Twit
http://tradetwit.ning.com