StockFetcher Forums · Filter Exchange · PORTFOLIO SELECTION AND MANAGEMENT USING RISK/REWARD RATIOS<< 1 ... 8 9 10 11 12 ... 65 >>Post Follow-up
guymar
113 posts
msg #92452
Ignore guymar
5/11/2010 4:34:11 AM

It is perhaps a good thing to calculate the Sharpe ratio, maximum drawdown and maximum drawdown duration in of the resulting trades and not of the individual ETF's, they give a wrong idea of security....
Don't forget, you are evaluating a trading model and not the individual ETF's.

medowz
59 posts
msg #92453
Ignore medowz
5/11/2010 6:07:02 AM

Great thread everybody. Thanks!

Kevin_in_GA
4,536 posts
msg #92456
Ignore Kevin_in_GA
5/11/2010 6:53:03 AM

Here's something weird - if you use the ETFReplay analyzer with EEM, IWM, and SHY at 50% weight for 3 months on returnA and 50% weight for 20 days on returnB, 0% volatility you get 469.7% return since 2003.

Now reverse the settings - same weight and timeframes but make 3 months for returnB and 20 days for returnA. The return is now only 248.3%. Almost a 2-fold difference at the same weighting factors??? I'll send the guys a quick note to see what's up with this.

ciscoslaves
5 posts
msg #92471
Ignore ciscoslaves
5/11/2010 2:31:49 PM

Kevin,

I see what you mean. Let us know what you find out. Thanks.

Kevin_in_GA
4,536 posts
msg #92482
Ignore Kevin_in_GA
5/11/2010 6:23:24 PM

"good morning kevin,


the software needs a 'rule' for how to deal with ties. you have tapped into this --- if 2 ETF's have the same ranked score, the software will default to the ReturnA as the tiebreaker -- so the code actually reads differently if you are 50% 3-month 50% 20-day vs 50% 20-day 50% 3-month.


the issue with ties is a bit of a quirk when only using 3 ETF's like this -- especially when 1 is cash. there will be a lot of ties sometimes. our new portfolio app is really a better version of that because as you add more ETF's, you get less ties. this app will be out pretty soon and then you will get fuller appreciation of this issue you have tapped into.


if you would like to see the difference, type 51% into ReturnB rather than 50% and 49% for ReturnA. The first choice is to go with the larger % Return period -- but if they are even, as they are with 50/50 -- then it will always default to ReturnA. These are issues that have to be assigned rules -- and we think these rules are logical -- albeit may lead to some very marginal choices of one vs the other."


Good to know - clearly the default to the longer timeframe makes a big difference in total returns. Important to know as this concept is refined.

cwn6161
40 posts
msg #92490
Ignore cwn6161
5/11/2010 10:22:40 PM

how does this affect the filter or our picks, in terms of the code?

jeremytimko
4 posts
msg #92510
Ignore jeremytimko
5/12/2010 1:10:04 PM

I think the simple screen presented on page 1 is a fantastic filter.

I have been running tests with it since I discovered this thread Sunday night. The only issue I have is one of using past results to predict future behavior. Specifically, I'm referring to the usage of EEM in the screen. Yes, this has been a wonderfully performing ETF over (roughly) the past 5 years, but 5 years ago, would you have known to include that ETF in your screen? Basically what I'm saying is, how do we know what the 'current' EEM is? Emerging markets may underperform in the coming years...though now that I think about it, this filter will take that into account.

Anyway, I have been trying out the screen with a ton of different ETF combinations and finally came to the conclusion that the best way this would work (for me) is to limit the usage to US market ETF's. So, I'm running it using SPY, QQQQ, IWM, DIA, MDY and IWC. Using those along with IEI instead of SHY, a quarterly rebalanced backtest I just ran revealed a 100.03% return since the start of 2007compared to a loss of 17.49% for SPY.

Also interesting are the choices used for "cash" equivalent, as a lot of treasury ETF's have a low volatility and hold up quite well in adverse market conditions.

IMO, this is by far the best thread on SF. Thanks.

Kevin_in_GA
4,536 posts
msg #92517
Ignore Kevin_in_GA
5/12/2010 3:33:19 PM

I think the simple screen presented on page 1 is a fantastic filter.

I have been running tests with it since I discovered this thread Sunday night. The only issue I have is one of using past results to predict future behavior. Specifically, I'm referring to the usage of EEM in the screen. Yes, this has been a wonderfully performing ETF over (roughly) the past 5 years, but 5 years ago, would you have known to include that ETF in your screen? Basically what I'm saying is, how do we know what the 'current' EEM is? Emerging markets may underperform in the coming years...though now that I think about it, this filter will take that into account.

Anyway, I have been trying out the screen with a ton of different ETF combinations and finally came to the conclusion that the best way this would work (for me) is to limit the usage to US market ETF's. So, I'm running it using SPY, QQQQ, IWM, DIA, MDY and IWC. Using those along with IEI instead of SHY, a quarterly rebalanced backtest I just ran revealed a 100.03% return since the start of 2007compared to a loss of 17.49% for SPY.

Also interesting are the choices used for "cash" equivalent, as a lot of treasury ETF's have a low volatility and hold up quite well in adverse market conditions.

IMO, this is by far the best thread on SF. Thanks.
+++++++++++++++++++++++

1. I agree that the simple filter works better than the more complex one I wrote in the first post, but it also does not give you any sense of the volatility or risk measure of the ETFs under consideration. I posted a more comprehensive version which also showed you the 21 and 63 day Sharpe ratios, volatility, and alpha for each time frame. I currently use that one since it gives me the same selections but lots of added information on risk and return.

2. As for the use of EEM, remember that in this instance it is really meant to be a proxy for emerging markets, IWM is the proxy for small caps, and SHY is the proxy for cash. I think that in this way the general trend of the market becomes very clear - if a specific sector or market is gaining strength, it is easily seen. EEM, VWO, EFA or any of the other ETFs that represent international/emerging would be fine (your yield may not be as high but the overall ranking should be the same).

3. I just finished backtesting this idea on IWM, EEM, and SHY back through mid-2003. Quarterly rebalancing provides a 314% return versus monthly rebalancing at 245% (this is through 3/31/2010, the last full quarter). Since then, both systems have had you in IWM, so the returns listed here are proportionally the same today.

4. If Emerging Markets underperform over the next few years, this filter will not have you invested in them. Same for if the market takes off - you will be in either EEM or IWM, not SHY.

Kevin

davesaint86
263 posts
msg #92533
Ignore davesaint86
5/12/2010 8:49:18 PM

Kevin - I agree with Jeremy. Great filter. I'm really pumped about it. I selected EEM, IWM, GLD, SHY as my picks. GLD was at the top this. I want to add a bond or two that historically gains more than cash. AGG has a pretty good long-term record of returns - 5.8%.

I've decided I'm going to use this strategy for my IRA, my 401k and my wife's 401k and use nothing else. I've deleted my other SF filters. You stated that you test back to the middle of 2003. Weren't you numbers better before? Did you change the weights?

Dave

Kevin_in_GA
4,536 posts
msg #92535
Ignore Kevin_in_GA
5/12/2010 9:40:20 PM

On page 2 of this thread I posted my manual backtesting of the filter since Jan 2007 (return of 113% with monthly rebalancing).

I downloaded the historical data for IWM, EEM, SPY, and SHY (EEM data only goes back to mid 2003, so that's when I started). Since July 2003 the returns were as I posted earlier. The ETFReplay results are done a little differently, as they are based on RANK rather than absolute score. Also, I am more comfortable with my manual backtesting since I know that I did it myself, and I stuck to the simple rule of switching to the top rated ETF even if the 1st and 2nd place scores were very close. I'll look at rank-based backtesting as well (if it works better, then why not use it?).

If you use the expanded filter I wrote (on page 5 of this thread), you can very easily see how each ETF has done over the two timeframes, including Sharpe ratios and excess return versus the SPY.

ETFReplay is about to unveil its expanded version of this backtester, which will also be linked to personal portfolios you can set up on their site. Should be VERY cool.

StockFetcher Forums · Filter Exchange · PORTFOLIO SELECTION AND MANAGEMENT USING RISK/REWARD RATIOS<< 1 ... 8 9 10 11 12 ... 65 >>Post Follow-up

*** Disclaimer *** StockFetcher.com does not endorse or suggest any of the securities which are returned in any of the searches or filters. They are provided purely for informational and research purposes. StockFetcher.com does not recommend particular securities. StockFetcher.com, Vestyl Software, L.L.C. and involved content providers shall not be liable for any errors or delays in the content, or for any actions taken based on the content.


Copyright 2016 - Vestyl Software L.L.C.Terms of Service | License | Questions or comments? Contact Us
EOD Data sources: DDFPlus & CSI Data Quotes delayed during active market hours. Delay times are at least 15 mins for NASDAQ, 20 mins for NYSE and Amex. Delayed intraday data provided by DDFPlus