I have recently joined the group (this is my first post) and started studying TA a bit more seriously. I came across the J2L system a while back and did some work on it recently. I would like to share some thoughts with you and ask for your opinions.
Metastock plots TSF as the sum of: 1. the end-point of a linear regression equation and 2. the slope of the linear regression equation. Thus, TSF(opt1)-Linearreg(opt2) may be written as Linearreg(opt1)+Linregslope(opt1)-Linearreg(opt2). For further flexibility, one could use Linearreg(opt1)+Linregslope(opt1)-Linearreg(opt3). Using this idea, I have run an optimization routine on daily SPDR Trust Series 1, ticker SPY (system details attached). The in-sample optimization period is 21feb97 through 25jan06 and the optimized variables are opt1=15, opt2=16 and opt3=11. The out-of-sample backtesting period is 26jan06 through 22jan09.
I get nice results and a potentially interesting equity curve.
I was wondering about your opinion on this system and its results and if you would have any suggestions for improvement, perhaps aiming at improving the flattish equity growth periods (jul98-apr00 and sep04-feb07) and the excessive equity volatility starting in sep08. Some ideas for future study could include dynamic stops, dynamic position sizing, pivot/price pattern entry/exit, staggered buy/sell short position size, trend filter, etc. I have experimented with adding a long-term trend filter (closing price EMA) and trading J2L only with the trend. At first blush, results are as per expected: dramatically lower annualized performance but much lower equity volatility.
Many thanks.
Leonardo