TraderMojo
16 posts
msg #98994
- Ignore TraderMojo |
2/9/2011 5:35:18 PM
What method do you use to backtest the RSI filter? Can the SF backtester do this (and if so, what settings)?
|
dune
4 posts
msg #98997
- Ignore dune
modified |
2/10/2011 3:58:07 AM
Kevin_in_GA
"Both approaches work, and have merit. The ETFReplay approach uses both performance and volatility rankings to make their selection, where the TSI or ROC approaches shown here use only performance.
This is why I have been requesting additional functions from SF such as a rank function or one that automatically calculates the Sharpe ratio for a user-specified time period. So far no progress on either front. "
*******************************************************************************************************************************
First Kevin, a Thank-You for all your contributions on this forum! Your efforts are truly appreciated.
When you mention the ETF Replay vs. your TSI/ROC weekly returns, what settings are you using on ETF Replay to give similar results/returns to your scans on Stock Fetcher?
1) Have you "curve fitted" the ETF Replay input data to match your results on SF, or
2) Are you using the default ETF Replay settings:
-Return A........3 month - 40%
-Return B........20 days - 30%
-Volatility.........20 days - 30%
3) Or are you using custom settings on the 3 input boxes on ETF replay?
Just curious as I have been playing a bit with the ETF Replay settings when trying to set up a weekly relative strength trading strategy (as you are currently using in your accounts as you've mentioned previous) and am interested in your current thoughts. Any insight you can provide on the ETF Replay setting you have found to be beneficial would be appreciated.
Also I agree 100% the best result would be for Stock Fetcher to write a syntax code for volatility adjustment that could be put into scans.
My personal preference would be a Sortino ratio over the Sharpe ratio given its focus on DOWNSIDE vs TOTAL volatility ..... there is no reason why they could not offer a number of different volatility syntax codes.
The benefit would be substantial and I hope others will press them to do so as you've done.
Cheers!
|
oldsmar52
104 posts
msg #99004
- Ignore oldsmar52 |
2/10/2011 10:21:22 AM
Could someone help, am confused:
1) The most recent filter being used is on Page 23, correct?
2) You get in the ETF based on what indicator?
Sorry, am dead after reading all 25 pages of thread & thanks for any help!!
|
oldsmar52
104 posts
msg #99005
- Ignore oldsmar52 |
2/10/2011 12:26:08 PM
After taking a break I think it is: run after close on Friday & buy (or keep) the highest tsi? Thanks
|
duke56468
683 posts
msg #99006
- Ignore duke56468
modified |
2/10/2011 3:08:46 PM
Kevin.....The outcome differs for the ROC(13) and the TSI(5,5,1) do you favor one over the other as a weekly trading indicator?
|
mystiq
650 posts
msg #99014
- Ignore mystiq
modified |
2/10/2011 10:55:27 PM
???
|
Kevin_in_GA
4,599 posts
msg #99021
- Ignore Kevin_in_GA |
2/11/2011 3:21:13 AM
Kevin.....The outcome differs for the ROC(13) and the TSI(5,5,1) do you favor one over the other as a weekly trading indicator?
+++++++++
I ran backtests on all possible combinations of the weekly TSI(x,y,1), letting x and y range from 1 to 10. Due to the symmetric nature of this indicator, TSI(4,7,1) is the same as TSI(7,4,1) so the total number of combinations was only 50. Backtests ran on IWM, SPY, EEM, and AGG from 12/29/2006 until 12/31/2010.
Most profitable combo was weekly TSI(3,7,1), which returned 30.5% compared to the ^SPX which returned -11.4%. For comparison, the weekly TSI(5,5,1) returned 20.1% over the same period. These numbers differ from my earlier hand calculations, but for comparative purposes I am willing to use them.
I also ran all combos of the weekly ROC(x,1) with x ranging from 1 to 30. The weekly ROC(13,1) returned 72.8% over the same time frame. Pretty clear difference in performance.
For now I am using the ROC(13,1). Since 1/2/2000 the 14 and 16 week ROCs were consistently the winner, but most the the gains came in the last four years.
|
duke56468
683 posts
msg #99035
- Ignore duke56468 |
2/11/2011 9:35:30 AM
Kevin.. thank you Sir, I really appreciate your contributions.
|
Kevin_in_GA
4,599 posts
msg #99041
- Ignore Kevin_in_GA |
2/11/2011 12:27:18 PM
No thanks needed. I am happy to share this type of information, and think that this particular approach makes investing (versus trading) a simple and profitable experience.
I also tested this on a series of Vanguard mutual funds for each asset class, just to be sure. These funds go back much longer than the ETFs, so it also allowed me to backtest from 1992 forward.
Since 1992, the data argues strongly for either a 14 week or 16 week ROC. Same for 2000 forward as well as 2005 forward. Once you look at 2007 forward, 13 weeks works about as well as any other, and the typical 3 month performance data for almost all ETFs and indices/mutual funds are readily available.
I am constantly impressed by both the simplicity and profitability of this approach. Trades happen about 3-4 times a year, which is awfully easy for most people to do, and fits nicely into most 401k portfolio strategies.
|
TraderMojo
16 posts
msg #99045
- Ignore TraderMojo
modified |
2/11/2011 5:07:33 PM
Kevin... Thanks for being so generous with your expertise on this methodology. You mentioned that in your 401K, trades happen about 3-4 times a year. Are you using the ROC(13,1) for this? And how often do you evaluate the need to re-balance (e.g. monthly? quarterly?).
|