Tclark and Capt_J,
The high returns can be partially accredited to treaking or curve-fitting and partially to the strategy. See the treaking numbers below (>2000000%). The curve-fitting part can not be expected to work for the future, but the strategy part should hold for the future.
The strategy is sound: ppo identifies trend, tend following via EMA crossing, extreme RSI anticipates trend change, and hard stop corrects bad luck.
Let's say 3/4 of the gain is due to teaking and 1/4 due to strategy. That gives 500000%, which is still excellent. But you would like to carry at least some of the curve-fitting gains forward. The question is how!
Here is a suggested plan on how:
1. test forward carrying capability: optimize the parameters based on N years (say 5) and apply the parameters for the following year, repeat this on a rolling base to get 20-N data points, analyze the forward carrying capability.
2. test if optimized parameter change can identify market condition change: optimize parameter for N years, optimize parameter for the following year, compare the two set, repeat this on a rolling base, analyze
3. test an adaptive version: optimise parameter over N years (a smaller N, say 1 or 2 or 0.5), optimize parameter from beginning to N years ago, use a weighted average of the two sets of parameter for next year, repeat this on a rolling base, analyze results
4. instead of carrying forward for one full year, try partial year.
Jim
Long Short Target
Stop: 15.33%, 1.85%, 18.20%
VTO bull: 25.75, 44; bear: 28, 37
PPO(3,10) <> 2.31
Short RSI(4) > 46.25
NDX EMA: 10
NASI EMA: 5, 24
1986: 25.90%
1987: 65.01%
1988: 29.43%
1989: 20.34%
1990: 34.54%
1991: 76.86%
1992: 103.73%
1993: 40.64%
1994: 37.08%
1995: 111.93%
1996: 85.38%
1997: 103.09%
1998: 70.59%
1999: 53.52%
2000: 200.30%
2001: 186.27%
2002: 75.81%
2003: 55.64%
2004: 16.10%
2005: 10.12%
86-95 Avg: 54.55%
96-05 Avg: 85.7%
20yrs Avg: 70.11%
compound: 63.66%
Accumulat: 2010845%
AI
