May Results for Main Model

Enclosed find the Annual results from inception through May for the main model.

ACTIVE ETF OIL MODEL OUTPERFORMS LONG ONLY OIH

The Active ETF OIL model has it data start in July 2002. $1 invested in the active model would now be worth $5.4 vs $2.6 for a long only position representing a significant improvement as a result from an active approach. It out performs the long only approach in six out of ten years often significantly so. Also in years when there were enormous drawdowns in a long only position for example 2008 the model was down 18% vs a down 60.15% for the underlyer. Similarly in 2011 the underlyer is down over 18% and the model is down 11%. The model volatility is half that of the underlying ETF. The OIL Model is currently long and has been so since March 27.

EEM MODEL SIGNIFICANTLY OUTPERFORMS UNDERLYER

The large scale EEM model is currently neutral. The previous long signal ended on 4/17 close of business. Net net the model has underperformed the long only EEM strategy for 2012. However this model has significantly outperformed the underlier since inception. Since the inception of the model the value of $1 invested in it would be $7.28 on april 20 2012. The value of $1 invested in a long position in the underlyer would be $2.47. As you can see this is significant out performance. There are times of course when the model lags the underlyer. However while it is tempting to assume that because of demand from china that EEM soared throughout this period there were some huge bumps in the road. For example in 2008 when the underlyer was down a whopping 49% due to fall off of global demand for commodities and materials the model soared by following the trend when long and exiting prior to the most major drops.

Results Of the EWJ (JAPAN MODEL)

Japan has been in the doldrums for two decades. If you invested $1MM in the EWJ ETF in July 2002 it would be up to $1.34MM today. On the other hand our timeing model would be worth $3.598MM. The worst year for the model occurred in 2006 down 7.81% when the underlying ETF was up 5%. However in 2008 when EWJ was down 26.95% the model was up 35.17%. The model has 71% winning trades and is considerably less volatile. Overall the Japan Model outperforms the underlier overwhelmingly in the last ten year period. Enclosed find the results. For 2011 the underlier was down 16.5% but the model was up 4.09%. This year the underlier is out paceing the model. This model uses almost ten years of end of day data and so appears fairly robust.

Results of EEM (Emerging Markets Model) from inception to Feb 2012

Note the goal here is to have a timing model that is not necessarily correlated with the underlying ETF. EEM is the second largest ETF block by size with net assets at just under 41 Billion dollars. As can be seen 2011 was a tough year for the underlyer while our model broke even but in relative terms outperformed. This year has seen some role reversal while we are up 4% the underlyer is up 15.76%. However overall since inception our strategy has significantly outperformed the underlyer and avoided the massive drop that occurred in 2008. Notably in 2008 our model was up almost 30% while the undelyer itself was down 48%. Overall the model returns are much less volatile than the underlying maket and the winning trade ratio is 71%. This is a highly scalable model which is important when allocating capital. Tommorrow we will look at Brazil

Annual Results of ETF Model From 2006-2011 Summary

As can be seen by the table the dynamic ETF model put in a respectable performance on a risk adjusted basis. While it did not perform as well as in 2010 it did well with respect to any reasonable bench marks in a year when most hedge funds lost money or under-performed. Overall the sharpe ratio for the strategy which uses end of day prices as input still has a good annual rolling sharpe ratio of 2.74. For 2011 the sharpe dropped to 1.69 however the Sharpes rolling can move around and we have seen this in other rocky years. Note that one thing that affects the rolling sharpes is whethet there are is more down months in a given year. In 2011 there were three down months and quite a bit of variation in the monthly returns which led to an increase in volatility. However several factors have came into play the main one of course will Europe be as bad as the Lehman crisis will a meltdown in greece lead to the same in the rest of the PIGs and what in turn will be the impact on the global financial system and what exposure to Europe do our own banks have. Will there be another Lehman style collapse in Europe? There are so many uncertainties that one might consider taking up farming. While the US seems to be bouncing back and certainly that is evident in the declines in inventory is distressed real estate markets such as Miami which has benefited from an influx of rich latins. Note that the Real has soared vs the dollar in recent years so that for a Brazil based investor that million dollar condo in South Beach is much more alluring. Overall the strategy has performed well over the last six years and lets see what 2012 brings.

Summary Of Results for Intermediate Model Oct 2011

The above presents the summary results for the Intermediate trading model through Oct 2011. The second table shows monthly returns since inception of strategy and the graph compares the monthly returns of the strategy vs the monthly return of the SP500 for comparison. Note that the strategy is much less volatile than SP500 and less subject to tail risk and black swan events. The strategy is highly scalable and can take over 1billion in capital so is ideally suited for a money management firm or hedge fund that has the capacity to scale.

September Results 2011

I have been working on some adjustments of the model where we improve returns considerably with a bit more volatility. The adjustments are due to changing the weighting schemes applied to each ETF strategy to be less dynamic ie the weights are not adjusted daily the idea here is to avoid sharp migrations in the weights on a daily basis. This well also reduce transaction costs. Enclosed find the summary results note that the annual Sharpe Ratio on average is 2.87 which is favorable for a strategy using daily prices.

Summary Of Results for Intermediate Model Aug 2011

Enclosed find the summary performance through the end of Aug 2011. Overall the strategy has performed well since inception in 2005 with a max drawdown of approx 1.7% from high to low. The reason for this is because the strategy carefully weights each particular ETF (proprietary weighting scheme) and may be half in cash at any given time. The average rolling 12 mth sharpe ratio exceeds 3.5. I can provide mthly results but in this table I am producing annual to date.

New Long Positions Established

On August 18th my models went long at the close on the following assets
ETFs EWZ, QQQ, OIH, SPY, EWJ, XLF. As of Aug 22 the system is down just over 6% which compares favorable with both the SP500 and the QQQs which are down 12.84% and 13.4% respectively. The massive EEM ETF faired even worse down over 15.8%. The last time this particular model saw such a down-turn was in 2008 where the downturn was even more pronounced. Remember that the idea was to create a model that has robust returns but avoids some of the major downturns. Now the volatility can be controlled by adjusting the amount of money an investor allocates in the model and the amount allocated to cash. Now that everything is more automated I will be updateing my positions more frequently.

Legal Disclaimer: Note that I am not a registered investment advisor so anyone using my trading signals does so at their own risk. This is not an offer to buy or sell securities

Summary Of Results through Aug 12 for intermediate model

The average annual return since inception in July 2005 is 18% with an average sharpe ratio of 1.87; The average winning month is 3% and the average losing month is 1.5%; Winning months outperform losers 2:1. For 2010 we were up through the end of July but are now flat due to an almost 5% drop in August. This also reduced our rolling sharpe ratio due to the sharpe increase in volatility in early august. However over all this model has been fairly robust since inception and held up well in 2008 relative to market indexes. It has performed well in August on relative terms given the extremely high level of volatility and the problems spreading through the euro-zone as well as the political stale-mate in washington. The table below indicates the summary results

.

Current Trading Signals

Below is a list of my current signals for each ETF that I currently follow and the date on which they were received.  Note that I get my signals on the close of a particular day and I follow highly liquid ETFs this makes for a scalable strategy.   The next phase of development is to come up with a tactical asset allocation model based on the ETFs I am tracking, Im not there yet but im giving a lot of thought as to the best way to implement it.  

XLF   Long on 06/01/2011   @ 15.31
EWJ   Long on 06/15/2011   @ 10
EWZ  Long on 06/06/2011   @ 73.01
FXI    Long on 06/01/2011   @ 44.48
SLV   Long on 06/01/2011   @ 35.75
SPY   Long on  06/15/2011  @ 126.39

QQQ, OIH, EEM, GLD are giving no signal at the moment.   I realize that there are a wide range of ETFs that are also fairly liquid but I have focussed my attention on some of the largest ones.   Also I cover a fairly broad spectrum of the markets for example Brazil, the emerging markets block, China, SP500, Nasdaq, Silver, Gold and Japan.   I hope to add a few bond and currency ETFs just to see how my models performs there.    Note these are signals determined based on my Intermediate Mean Reversion Methods. 

Note I should qualify that I am not a registered investment advisor and I am not making any recommendations to buy or sell securities.

Updated Intermediate Model Results Table

I added two new ETFs to my Intermediate Mean Reversion Model.  If anyone wants any more details on results for any of the individual ETFs feel free to ask me.

Results For Intermediate Models

Above find the results for my "Intermediate Mean Reversion Models".  This is a summary using data back to Feb 2 2002.  The reason I chose that date was for consistency.  All of the ETFs considered had data going back to at minimum this date and they were among the most liquid actively traded ETFs.  The reason I of course select the most liquid ETFs is for ease of scalability.  In each instance the Strategy Volatility for the Model is considerably less than for the underlying ETF.   I also include the returns of a long only position in the SPY ETF for comparisons sake.  I have more data available for each individual model upon request.

My New Model

I realize it has been a while since I have posted but I have been focussing my attention on developing a short/medium model.  I started with the oil market and I am using an extreme mean reversion model with stops.   Overall in a highly trending market the longer dated models work very well.  If the model is too short dated it takes the trader out of trends are they are being established.  However on the way up it is never a smooth path so I sought to develop a model with about half the volatility of the ETF but that trades more frequently than my long dated model.   This model has a 70% winning ratio (ie the number of winning trades exceeds the number of losers).  The overall volatility of this model is half that of the underlying ETF.  The model traded three times in the month of may exiting on 05/20/2011.    

Results For Shorter Dated Oil Model Through March 2011

Below find the results for my shorter dated model applied to the OIH ETF.  The goal in this endeavor was to find a much shorter dated model that would trade more frequently than the long dated models that I use.   The other goal was to develop a model that is much less volatile than the underlying ETF.  In this case the volatility of the model is 12% vs over 50% for the underlying ETF.   Additionally the number of uptrades to down trades is 12:2 which is extremely high and the dollar value of an up trade is over two times the dollar value of a down trade.  This indicates that the model is fairly robust.  It should be noted that the model has not made as much money as the underlying ETF but has eliminated some of the huge draw-downs associated with the ETF.   In 2008 for example the model was up over 9% while the underlying OIH ETF was down 60%.
 

	Trading Statistics
	Model
UpTrade	12
AvgUpTradepnl	$100,472.21
DnTrade	2
AvgDnTradepnl	-$38,366.28
AvgTotTradepnl	$80,638.14
Total Trades	14
WinRatio	%85.71
StratVol	%12.24
MktVol	%50.94

An Article on FundFlows From IndexUniverse

http://www.indexuniverse.com/sections/features/8626-2010-etf-fund-flows-vanguard-leads-all-firms.html

An Interesting Article on Factor Models

Below find an interesting article on the use of Factor Models applied to ETF's from seeking alpha
http://seekingalpha.com/article/261570-a-simple-macro-factor-relative-strength-model?source=email_etf_daily.

An interesting article on the global Economy

I am posting the following link to seeking alpha
http://seekingalpha.com/article/261001-2-resources-related-etfs-signal-caution-for-the-global-economy

A New Study on Hedge Fund Performance

Sometimes in this blog I will provide links to articles from other blogs that I thought looked fairly compelling.  This piece was written by Gary Kaminsky who is a frequent guest on CNBC and was previewed on the oxtones blog.  http://oxstones.com/kaminskys-call-hedge-funds-do-worse-than-market/