StockFetcher Forums · Filter Exchange · Kevin_in_GA -- would you repost a script?<< 1 2 3 4 >>Post Follow-up
glgene
534 posts
msg #100547
Ignore glgene
5/3/2011 11:55:10 PM

Kevin,

Would you be so kind to show us (again) your script for ranking 5-10 non-correlated ETFs on a weekly basis? And your reasons for the specific ETF selections, and methodology for buying on a weekly basis? I'm not sure if you have updated it since your last posting. Either way, a reposting of it would really be appreciated by me, and I'm sure others who pay attention to your valuable posts.

Thanks!

Gene in Ohio

Kevin_in_GA
4,538 posts
msg #100589
Ignore Kevin_in_GA
5/6/2011 9:40:10 AM

Sure. It's not really all that complex.

In my previous posts, I demonstrated how simply investing in the most profitable asset classes over the past 3, 6, 9, or 12 months almost always beat the market as a whole, often by more than double the average CAGR. This was based on the work of Mebane Faber, who extended this basic idea to a tactical asset allocation strategy (TAA) that rotated between asset classes based on the same relative strength concept. Going back as far as 1920, he was able to show that this approach was remarkably consistent in its outperformance of the market as a whole.

Diversification was discussed, but not just holding a bunch of different stocks or ETFs, because if they are highly correlated they will not effectively smooth out any volatility in the markets. What is best is a set of uncorrelated assets so that as one goes down another is likely to go up or at least remain stable. Standard diversification still outperforms the overall market because it limits downside risk, but it generally fails when compared to a tactical asset rotation strategy using the same assets, just invested sequentially rather than in parallel.

If I were to sum it all up it would look like this:

1. Start with a manageable set of investments that are not highly correlated to each other. I use just three – International/Emerging markets, US Small Cap Stocks, and Bonds. These three alone provide pretty good diversity. For safety I would also have one money market fund in the mix, so that you can move entirely into cash in case of another meltdown like in late 2008 (in that case, all asset classes dropped at the same time, so there was nowhere to put your money that was safe other than in cash).

2. Every Friday, determine the return for each over the last three months. I use weekly close data, and a 13-week look-back period. If you read the Faber paper, you know that you can use 2, 6, 9, or even 12 month periods for this calculation and still handily beat the market.

3. Invest in the highest performer, and only re-allocate when another investment shows better three month performance. Most people are not all that comfortable with the “all your eggs in one basket” approach I take, so it is also OK to use a larger set of investments and put funds into the top 2-3. Just re-allocate when any of them are replaced by a better performer.

I look at the 3 month performance of the three asset classes each Friday, and if there is a change required I make the re-allocation on line, and I usually know that it is due several days in advance.

Before I go any further, a little comparison of how this strategy has fared over the last 4 years:

Here I am using only three assets – equities (using VB as a proxy for the US small caps), emerging markets (using VWO), and bonds (using AGG). Performance is calculated as I described, and reviewed either each week or on the last Friday of each month. For comparison, I have also done the static diversified portfolio (equal amounts of each) as well as the overall market (S&P 500).

REVIEW WEEKLY GAIN 1/3/2007 - 12/6/2010: 121.03%
REVIEW MONTHLY GAIN 1/3/2007 - 12/6/2010: 109.51%
DIVERSIFIED GAIN 1/3/2007 - 12/6/2010: 24.17%
OVERALL MARKET GAIN 1/3/2007 - 12/6/2010: -12.87%

In the case of the weekly review, you only made 25 re-allocations over 48 months. For the monthly review, you only made 13 re-allocations over the same time period. But look at the difference that small amount of time each Friday made in your account performance. These are numbers most professional fund managers would kill for.

Here is a simple filter that let's you do this (you can sub in your own ETFs if you want):

Fetcher[

symlist(VB,VWO,AGG)
add column weekly roc(13,1) {13 week performance}
sort on column 5 descending

]





davesaint86
264 posts
msg #100638
Ignore davesaint86
modified
5/7/2011 9:32:38 AM

I agree with what Kevin's states about correlated assets. I sold at a 3% loss on Wednesday for three ETFs on bought on Monday. I have a list of 18 ETFs in my list. Backtest results showed an overall gain of 180% since the beginning of 2008. The three ETFs that showed up at the top were SLV, GLD, and DBC. All of these are correlated and all overbought. I'll keep my same list of ETFs, but next time, I'll groups the 3 correlated ETFs purchased as 1/3. So in essence I would have bought 5 ETFs, (DBC, SLV, GLD) 33.33, VB-33.33% and VNQ-33.33%. Since SLV was so parabolic I probabaly in the future either would not buy it or lower the allocation to 5% of the 33.33%. I would still be sitting at a loss if I bought the 5 ETFs but not as big of a loss.

Dave

Kevin_in_GA
4,538 posts
msg #100639
Ignore Kevin_in_GA
5/7/2011 10:31:37 AM

Dave:

If you look at SLV as a fourth option to the filter I posted above, you would have rotated into SLV on 2/11/2011 at 29.21, and would still be holding it as of this Friday at at gain of just over 18% (even after this week's bloodbath).



glgene
534 posts
msg #100640
Ignore glgene
5/7/2011 12:22:23 PM

Thanks so much, Kevin, for your reposting. I (and others) really appreciate it.

Curious...would you ever include any contra ETFs in the set?

Also, have you discovered that weeky vs. monthly rebalancing gets better results? Does weekly get you too many whipsaws?

Gene

davesaint86
264 posts
msg #100642
Ignore davesaint86
5/7/2011 2:10:53 PM

Kevin,

Say you didn't get in SLV and just starting using the ETFS in the new list on Monday would you have bailed on SLV? if not you would be sitting a large loss. I purchased a subcription to Smartstops.net and glad I bailed. I can always get back in at a lower price. Also the daily buy filter you developed a long time ago gets you out pretty early with most of your gains intact. I've been using the filter for my 401k it's been doing good. Waiting to see if it is going to switch to AGG by the end of the month. Keep up the good work and thinks for doing what you do with the filters.

Dave

Kevin_in_GA
4,538 posts
msg #100647
Ignore Kevin_in_GA
5/7/2011 9:09:24 PM

Simple solution - wait for the leadership to change before entering the trade. This would mean that you had to wait for SLV (and would still be waiting) while missing a great run up.

Othrwise you could use a stop at close - 2 x ATR(20) or something like that whch incorpoates recent volatility.

I would probably use DBC as the preferred asset class, simply because the precious metals have too much volatility. The components for DBC are heavily weighted toward oil/energy, as you can see below:

DBC Top Ten Holdings
HEATING OIL FUTR JUN11: 15.32%
BRENT CRUDE FUTR MAR12: 15.19%
GASOLINE RBOB FUT Dec11: 15.12%
WTI CRUDE FUTURE JUL11: 13.91%
GOLD 100 OZ FUTR Aug11: 8.08%
SUGAR #11(WORLD) JUL11: 5.90%
NATURAL GAS FUTR Oct11: 5.35%
Lme Copper Future Mar12: 4.87%
Lme Pri Alum Futr Sep11: 4.49%
LME ZINC FUTURE MAY11: 4.23%

The correlation betwen DBC and the other asset classes is as follows:

DBC - VB 6 month correlation: 0.38
DBC - VWO 6 month correlation: 0.47
DBC - AGG 6 month correlation: -0.13

Definitely fits the "weakly or negatively correlated" category. Adding this to the simple filter from the post above, you would have been in on DBC on 2/18/2011 and out (back into small caps) on this past Friday, losing a small -0.31% on the trade.



pcarmasino
2 posts
msg #100672
Ignore pcarmasino
5/9/2011 10:07:20 PM

kevin - what would be the best way to backtest this. I tried with some different settings and did not get the same results you did.

thanks for sharing your insights.

Kevin_in_GA
4,538 posts
msg #100675
Ignore Kevin_in_GA
5/10/2011 7:36:38 AM

kevin - what would be the best way to backtest this. I tried with some different settings and did not get the same results you did.

++++++++

Different settings is likely to end up with different results.

I back test this manually. Look at the top scoring ETF each Friday. Stay in it until it is no longer the top. Reallocate to the new leader when that happens. Rinse. Repeat.

I have not figured out how to backtest this automatically, except in other software (e.g., Stratasearch). The reason is that SF will not implement a "rank" function.

pcarmasino
2 posts
msg #100677
Ignore pcarmasino
5/10/2011 12:13:06 PM

i'll have to look at that platform... will that give me a spreadsheet output? do you post on there as well, because i'll use stratasearch to run this backtest, but with my "not so good" knowledge of writing formula i could use some help. can you post the formula for stratasearch here?

thanks again...i was up for hours reading your old posts...

pc

StockFetcher Forums · Filter Exchange · Kevin_in_GA -- would you repost a script?<< 1 2 3 4 >>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