duke56468
683 posts
msg #97494
- Ignore duke56468 |
11/14/2010 11:11:58 PM
EtfReplay suggests the settings at 40%-30%-30% for 3 month-20day-20day. Changing any of these settings can make significant differences in the outcomes of most portfolios. How do we decide on the optimal settings? It would seem that it may be different for different portfolios. The possible combinations are staggering.
|
Kevin_in_GA
4,599 posts
msg #97496
- Ignore Kevin_in_GA |
11/15/2010 6:39:57 AM
There are no optimal settings. The issue you are highlighting is really the opposite of what you are asking, that is trying to avoid curve-fitting a specific set of ETFs.
My suggestion (I have made this one already) is to look at a small set of uncorrelated ETFs and find settings that allow you to have beaten the S&P over the following time frames:
YTD
1 year
3 year
5 year
You can't go back further than 2003, and many ETFs were not even available until 2005. When you find a specific set of ETFs and one setting that consistently beats the overall market, stop. Focus on consistency and robustness, NOT optimal return!
What I don't like about using RS is that, by its very nature, it is RELATIVE to something else. If the market is in free fall and your top scoring ETF is falling as well but more slowly, RS makes it look like it is a good investment rather than a bad one. The basic principle of RS is sound, but in a correction it has this weakness.
|
wkloss
231 posts
msg #97499
- Ignore wkloss |
11/15/2010 2:33:29 PM
Kevin & duke,
Leslie Masonson wrote Buy Don't Hold (www.buydonthold.com/ ). He suggests using relative strength first followed by several technical indicators to determine if you should take the trade. One of the technical indicators/filters is whether the ETF is above its moving average (MA200, I think). If no, don't take the trade.I believe this attempts to address your concerns. I looked into using MA200 as a filter. I didn't do a formal backtest but my observation was that MA200 reduced the yield.
I liked the book and picked up a few good points. The downside is that Masonson's group of technical indicators, called the dashboard, produced some bad results. By comparison, ETFREPLAY using Kevin's ETF's and settings produced positive results during the same time periods.
The concept of using uncorrelated ETF's solves most of the problem. There is usually a bull market somewhere. It would be great if we could eliminate risk but we may have to settle for reducing risk.
Bill
|
TheRumpledOne
6,411 posts
msg #97500
- Ignore TheRumpledOne |
11/15/2010 8:00:29 PM
Risk can be eliminated at the expense of upside gains.
But making 2% or more per month, virtually risk free, is more than enough (for me).
|
duke56468
683 posts
msg #97501
- Ignore duke56468 |
11/15/2010 9:11:45 PM
Kevin_in_GA
msg #97496
What I don't like about using RS is that, by its very nature, it is RELATIVE to something else. If the market is in free fall and your top scoring ETF is falling as well but more slowly, RS makes it look like it is a good investment rather than a bad one. The basic principle of RS is sound, but in a correction it has this weakness.
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Does EtfReplay calculate the RS differently than this RSI definition?
"Developed by J.Welles Wilder in 1978, the RSI is an oscillator which is used to track the momentum of the stock price. Unlike other relative strength measures, the RSI compares the price of the stock relative to the actual stock, not against other stocks or indices. The basic premise of the RSI is that an advancing stock will trend close the high of the day, while a declining stock will trend toward the low."
|
Kevin_in_GA
4,599 posts
msg #97502
- Ignore Kevin_in_GA |
11/15/2010 9:56:41 PM
Yes, they use a combination of relative strength and rank. Not easily translated into SF code since we have no ranking function.
|
duke56468
683 posts
msg #97550
- Ignore duke56468 |
11/18/2010 6:55:33 PM
Looks like EtfReplay is going to start charging for the service. Hopefully Kevins TSI filter will do equally well or better for us.
|
Kevin_in_GA
4,599 posts
msg #97605
- Ignore Kevin_in_GA |
11/24/2010 6:26:27 PM
Please note that ETFReplay is using relative strength and NOT the RSI:
http://www.etfreplay.com/blog/post/2010/11/21/Relative-Strength-vs-RSI.aspx
This is basically just the price change in percent over the period selected - the ETF with the greatest price appreciation is ranked highest.
Extremely simple, extremely effective.
I still prefer using the weekly TSI, but their method works quite well. I have noticed that the weekly RSI(4) and the weekly RSI(13) usually rank order the same as the weekly TSI(5,5,1).
|
duke56468
683 posts
msg #97734
- Ignore duke56468 |
12/2/2010 7:00:01 PM
Free ride is over at ETFreplay. $30/month
|
campbellb75
101 posts
msg #97792
- Ignore campbellb75 |
12/3/2010 7:29:26 PM
Has anyone tried to build an identical filter in SF?
http://www.etfreplay.com/members/how_the_screener_works.aspx
|