Tuesday, January 12, 2010

Futures Simple Moving Average Crossover

Last week I went over the S&P futures simple moving average crossover system. Today I'm going to briefly look at the performance of trading a futures basket using a simple moving average crossover system that is slightly modified to manage risk a little better.

The simple moving average cross has been used by traders for a long time now, and it appears that it's effectiveness has been waning over the last several years, or most likely, the parameters that I used to test this have become less effective to the benefit of other parameters in the current (somewhat volatile) market environment. This is the danger of curve fitting a trading system, although, if you have tested a system across a very long time frame, I would say to trade the curve fit system over any randomly selected variables. The reason is in the picture below.

If I were to trade this tested system, I would want to trade between 20-30 day breakouts because assuming that markets tend to be mean reverting, the optimal breakout days is likely to be in that frame in the future. Over shorter periods the 15 day or 35 day breakout may be ideal, but over time, the market has shown that the 20-30 day has been an optimal breakout. But I digress.

Futures Simple Moving Average Crossover

This system works just like the normal simple moving average crossover, but throws in a volatility based stop to limit the downside from what could be catastrophic lag in the moving averages crossing. The test was conducted across a basket of 30+ futures contracts with risk limited to .5% per trade.

CAGR: 41.7%
MAR: .76
Sharpe: .75
Max DD: 55.1%
Longest DD: 18.4 months
# Trades: 1387

Equity Curve

Monthly Returns

This system has performed well since 2000 and captured much of the gains from the inflationary uptrend in commodities through the middle of 2008. The systems reversed positions into 2009, but the lag caused the system to lose much of the gains from shorting the market at the end of 2008. This is about the time that institutional investors start to say trend following is dead and where astute trend followers start raising money. I would venture a guess that trend following is going to see a very good year following a year of flat returns in 2009. Trends will eventually reestablish themselves and force investors to recalculate risk premia quickly.

Overall, the simple moving average crossover on the futures side is still an effective system although it does appear that some of the edge may be deteriorating a bit. The problem that I see with the simple moving average crossover is how slow it can react to volatility. Much of the volatility in the system occurred in 2008-2009 when total market volatility spiked outside of what would be considered a "normal" market. Adding an "abnormal" market filter that reduced position sizes could improve the gut churning effects of this particular system. This is something I'm working on for stock and etf trend following. Hopefully I'll have some interesting stuff together in the coming weeks.

No comments: