In reference to this article titled Maybe the best market timing strategy ever I ran a series of backtest reports to verify the concept..
In a past article titled “Timing System Gone, Not Forgotten” from 2003, Mark Hulbert lists Norman Fosback’s past timing system rules as:
1. “To exploit positive seasonality around the turns of each month: Buy at the close of the third-to-last trading of each month, and sell at the close of the fifth trading session of the following month.
2. “To exploit positive pre-holiday seasonality: Buy at the close of the third-to-last trading day prior to exchange holidays, and sell at the close of the last trading day before a holiday.
3. “Exceptions: If the holiday falls on a Thursday, sell at Friday’s close rather than Wednesday’s. Also, if the last day before a holiday is the first trading day of the week, don’t sell until the day after the holiday. Finally, never sell on the first trading day after options expire; instead wait an extra day.”
I had a problem programming the 3rd rule above , the “Exceptions” . I tried a bunch of different approaches but couldn’t seem to get the “exceptions as a quantified rule” correct for the back test. The results that you find below are based on the first 2 rules listed above.
Is a 5.4% annual return really the “best market timing strategy ever”? I think not!
Using the following rules you can more than double the historical performance.
1) Buy at the close on the 5th day prior to the end of the month then sell at the close after 6 trading days since entry.
2) Buy at the close on the 5th day prior to a holiday then sell at the close after 1 trading day since entry.
Below is the performance report based on the 2 "improved" rules listed above, this report is based on taking a $10,000 position in the SPY for each signal since 1988. There have been 380 signals since 1988 and 236 have been winners for a 62% win rate. The average trade made .59% , a compounded return of 745% which is 9.51% per year.
6 comments:
Interesting but something seems off. If the July 2011 Hulbert article is still referencing the exact parameter testing as the 2003 article, Hulbert states an 11.4% annualized return in his recent article that you link to vs. the 5.43% you tested for the Fosback criteria. Some differences I noted are you using SPY vs. Hulbert using the Wilshire 5000, his test appears to go back to the "early 80's" vs. your 1988, and as you mentioned, you couldn't get the 3rd rule to code. Even with these differences, it seems like a big discrepancy in the annualized returns. Have you also included an interest rate proxy such as 90 day T-bills that Hulbert mentions when not in the market or are Hulbert's numbers way off?
Yes, those are good points. I used the SPY because I dont have the wilshire 5000 data and my Russell 2000 data only goes back to 1999. The point is timing to beat buy and hold, if it works for Wilshire it should work for S&P. Adding the performance for the 90 day T bills will certainly improve the results in the late 80s rates were around 8% and in the 90s 3-7%. I dont think the 3rd rule will boost returns much, but its possible. I will do a follow up when I get it correct.
Nice work.
I was wondering after reading both Hulbert's study and yours if placing protective stops after the long entries would have helped with the draw downs or would it whipsaw you out of profitable positions?
Loan wolf,
Using an 8.2% stop loss would have improved the results slightly by reducing the 10% loss from 8/25/98 - 9/3/98 down to an 8.2% loss and keeping you in all of the other trades.
You said your SPY data starts in 1988. Are you using SPX for 1988 - 1992, as I seem to remember that SPY first started trading in Jan. 1993?
Yes the SPX was used as a substitute prior to 1993.
Post a Comment