More Out of Sample Results

Over the past couple of weeks Michael Stokes over at MarketSci has done an excellent series of posts on the state of short term mean reversion system. The series, and this one post in particular, showed clearly that mean reversion systems have been most healthy in the last decade. Prior to 2000 short term mean reversion systems would have provided a slow and steady decline in your equity.

After studying this post for some time I was interested in how the MultiStage Trading System would have performed during the years preceding 2000... during those years toughest on mean reversion systems. The MultiStage System is, at it's core a mean reversion system, but not purely. It is a hybrid and changeling system that also has trending characteristics and market adaptive characteristics.

This was especially interesting since prior I had only been testing my approach back to 1999, and while 1999 was an okay year, it wasn't the best and didn't really reach into the meat of the declining period. Since I have pretty good data back to 1990, I ran a backtest of the system for the period from 1/1/1990 to 12/31/1998. This was a nice out-of-sample test in a period I have not really looked at before.

Results

The results of the test in aggregate were very acceptable. I've clipped out the Amibroker report and pasted it here. You will see that the average return per trade of 3.58% was decent. The win ratio (not shown) was 82%, also very acceptable.

The most significant number on this chart, however, is the number of trades. At only 817 trades, the average per year was about a quarter of years in the following decade. The challenge with this time period was the lower volatility. Mean reversion systems thrive on volatility, and when it's missing they fall apart. In the case of the MultiStage System, it still works, but it's severely constrained.

Another interesting observation was that short trades didn't work. My later tests showed shorts to be equal to longs in return and win ratio. In this case the average short trade was a loser. In true MultiStage fashion, however, losers were dampened and there were only 34 trades making up the short side.

All in all, I was quite happy with the results of this test. Thanks to Mr. Stokes for pointing out the history. The MultiStage System is extremely robust and held up well under entirely different market conditions. If the market were to shift back to "trending friendly" I feel entirely comfortable continuing on with this approach.

Good Trading...