Interesting Trading Strategy

All, I apologize for not posting over the past several days. I've been swamped at work and when at home have gotten absorbed in an idea I encountered on the traderfeed.com blog.

I'm not a trader - far from it - but this got me interested, especially with the ability to do automated programs. I didn't believe this example when I first read it because based on the "random walk" hypothesis something as simple as this shouldn't exist. Basically the idea is that if you buy the S&P on days when the S&P is down and sell 3 days later then those trades will double the performance vs. buying the S&P on up days and selling 3 days later. So I had to check the data and sure enough I got similar findings.

I pulled historical S&P data from Yahoo back through 1993 and ran the test. I found that 3 and 5 day trades buying the S&P on down days averaged gains of .21% and 0.34% respectively. Buying on up days returned 0.07% and 0.13% respectively. Again, very interesting. This shouldn't exist in efficient markets - but thinking about it I think what may be at work is simply playing off of hard-wired crowd behavior and the market reflects that interplay of exaggerated fear/greed/herd behavior. Buy when everybody else is selling. Don't buy when everybody else is buying.

Of course starting from this finding leads to a million different options, but I started looking at this from different angles, and returns of buying on down days can be enhanced if you restrict yourself further to only placing these trades in generally "down" markets. I'm defining "down" markets here as a market where the current market price is down over both 30 and 60 days (rough estimates- I'm not exact in constraining dates).

In these down markets buying on down days for the S&P and selling 3 days later generates 3 and 5 day returns of 0.55% and 0.76%. In contrast buying on up days in down markets generates -0.07% and 0.04% returns on the same 3/5 day trades. In effect - almost all the enhanced benefit/value of this strategy is captured in "down" markets - when fear is presumably sitting in with investors over extended periods.

All this is very interesting to me and I'm looking at feasibility/transaction costs/investment vehicles (SPY or futures)/generating leverage/size of positions/etc, but maybe trading in some type of automated format is workable considering how low transaction costs have become.

It looks like in an average year since 1994 this approach would generate around 30 trades per year. (with minimum of 0 trades in 1995 and 70 trades in 2002.) The strategy is generally historically profitable, although you would have lost money in 2001.

So far in 2007 the down market strategy described above would have generated 9 trades that averaged 3 and 5 day gains of 0.84% and 1.35%.

Here are the trades for 2007 to give an example.



Pretty cool stuff in my view. Right now I'm trying to understand how a practical and automated implementation of something like this could be done. The cool part is it gets you into the market at times when there's a good chance for strong gains. It reduces risk by getting you out into cash for most of the year. It could also be combined with some other strategies since it only occupies capital for a portion of the year.