Most of my trading models use Profund mutual funds as trading vehicles. These are broad market index or Dow Jones sector funds designed for short term traders. They are designed to move at 150% or 200% of the daily move of the sector or broad index they track. They may be traded once per day as late as 5 minutes before market close with no commissions or spread. The danger of using these funds is that if the model gets the direction right only half the time, and irrespective of the magnitude of the move, it will lose money. A day in which the fund is X% down will require more than X% up to recover fully. And this is even more true for the leveraged Profunds. It is a sobering fact of the math. Call it the "whittling" effect. Thus, a model with only 50/50 odds of being correct, trading in a trendless market, will steadily lose value. Rydex offers similar funds, but most sector funds are unleveraged.
In mid 2001, I developed a market timing model based on technical indicators and prices of the NASDAQ 100 index. I've made only one change to the model, a minor one, in early 2003. The model attempts to identify market trends and short term overbought and oversold conditions. It uses the Profunds UltraOTC, Profunds UltraShort OTC funds, and cash, as trading vehicles. The market may either be up or down, and the model may be either long, short, or in cash. Thus there are 6 possible outcomes. "Correct" corresponds to 3 of these - bullish on up days, bearish on down days, and in cash on down days. The model I settled on early in '03 I call Aggressive Hedge I. To show its performance in bull and bear markets, I back-tested it to the beginning of the bear market - January 14, 2000 - the high point of the S&P 500 index. Except for a 3 month period after the 2000 election when the NASDQ 100 index was moving chaotically, the model has performed well. While the model only predicts the next day's direction correctly 54% of the time overall, it tends to do better on the stronger moves. During the '00-'03 period, on winning days the model moved up on average ~3.80%, while on losing days it went down on average only ~3.05%. Accumulated over years this produced outstanding returns. Typically the model switches from long to short and cash several times per week, and one should not assume that the model will do better or worse in bull or bear periods. The model is purely objective. The charts below are updated by running a few computer programs on data I download. No tough, sweaty, subjective judgements are required.
Prior to Feb '03, it's fair to say performance of Aggressive Hedge I was destined to look good, subject only to the requirement that it be a single objective, quantitative-based mechanical strategy. In mid '04, the model's performance levelled off. From its design, there is more of a danger the model can get whipsawed. Aggressive Hedge II uses different rules which should preserve the good features and also limit losses due to whipsawing. There's another interpretation possible - that a close variant of this model has been discovered by the hedge fund community and now exploited so fully as to weaken the performance. Unlikely? Who knows, but a certain paranoia can be a healthy thing in this business. With time, the market mechanics have continued to migrate away from the sweet spot for this model. Lately, it's downright suicidal. The message seems clear - mechanical models, no matter how spectacular, should not be followed blindly. I suppose the implosion of the hedge fund Long Term Capital in 1998 has already delivered this message to the financial world.
The same strategy rules can be based on the S&P 500 index and the corresponding Profunds UltraBull and UltraBear funds in a similar way. While this model made strong profits during the bear market, it was not as striking for this lower volatility index as for the NASDQ 100. However, the strategy did very well, far outpacing the bull run of the SPX index through 2004, and like it's NDX-based sister, is now performing quite poorly.
The performance of Aggressive Hedge I using the NASDQ 100 is included on the Mechanical Models page.
2000: +27% (a great year turned mediocre during the NASDQ chaos of the final 2 months). NASDQ 100: -37%
2001:+165%. (NASDQ 100: -33%)
2002: +321%. (NASDQ 100: -38%)
2003: +237%. (NASDQ 100: +49%)
2004: +0%. (NASDQ 100: +11%)
2005: -3%. (NASDQ 100: +3%)
2006: -44%. (NASDQ 100: +7%)