Signal Type	Trade Date	Return since issued
		World	U.S.
			Nasdaq 100 (QQQQ) Russell 2000 (IWM) S&P 500 (SPY)

Market Update

Stocks experienced a sharp retreat over the five-day span. This week's sell-off on heavy trade within the context of the long uptrend that started last March caused our Model to issue a Cash signal after the close today.
The trouble started Monday as the major averages gave up solid initial gains to finish in the red after the dollar rebounded, resulting in a 1.2% loss for the S&P 500. News that consumer confidence fell in October caused stocks to lose more ground Tuesday. The selling intensified during the next session following disappointing housing data: an unexpected 3.6% drop in September new-home sales triggered a sharp decline on heavy volume, causing the Nasdaq Composite to shed 2.7%. Stocks were able to rebound smartly Thursday after GDP growth for the third quarter came in at 3.5% but the gains did not last, as another sell-off on strong volume occurred Friday after a drop in consumer spending renewed concerns over the strength of the economic recovery. The S&P 500 lost 2.81% on the day to cap a week of heavy losses for the market.

The Russell 2000 (IWM), Nasdaq 100 (QQQQ) and S&P 500 (SPY) respectively lost 6.21%, 5.03% and 4.18% over the five-day span. All three ETFs have now crossed back below their 50-day exponential moving average (EMA) but remain located above their 200-day EMA.

For its part, our World portfolio underperformed its U.S. counterparts this week with a loss of 8.32%. The portfolio consists of the 5 top-ranked world ETFs as of October 9, which marked the beginning of the current 4-week holding period. Please note that since we now have an active Cash signal, the World approach calls for selling your holdings if you follow the "Long Only" or "Long and Short" strategy. Only if you follow the "Buy and Rebalance" strategy should you remain invested in the top 5 ETFs, as the strategy calls for staying invested at all times. Please go to the "Our Service" page for all the details.

We now have a Cash signal in effect.

Trend Timing School

Standard deviation as a measure of risk

We have frequently discussed risk in these pages as well as methods to manage the risks we take, such as diversification. Unlike gains and losses which are real, risk is in general just a vague and abstract notion, on a scale ranging from low to high. We instinctively know that in order to achieve superior returns we have to take higher risks. For example, a money market fund or government treasury fund would have a very low risk but also lower returns than say a portfolio of blue-chip large companies which, for comparatively more risk, will also deliver higher returns. At the high-end of the risk/reward curve you can find small cap indices or funds with the highest return potential but at the cost of high volatility. Today we will attempt to get at the metrics which characterize risk and how it can be measured.

Risk can be viewed in many different ways: as volatility of price, as risk of losing the investment, or simply the possibility that something different than expected will happen. To demonstrate the shifty nature of what is meant by risk, we take the example of keeping a large part of our assets in cash. You could say that the risk is non-existent because you are guaranteed to have the same amount of cash in the future (except for fires and thieves, of course), or you could say the risk is infinite because you have the certainty of steadily losing money or purchasing power to inflation by keeping your assets in cash. In this case our opportunity risk is what counts, because cash holdings are not actively growing our wealth as they should be.

Most of the time risk is linked or even equated to volatility. When an investment rises and falls drastically over short period of times it is also considered high risk because its performance could change quickly in either direction at any time. If two investments provide the same return, we clearly prefer the one that does so with the least amount of volatility. By far the most common and widespread volatility measurement is what statisticians call standard deviation. It is the measure of how widely dispersed from the mean a series of values are. The values can be anything such as price or return of an investment or index. All it takes is a number of data samples at fixed intervals over a period of time, and a good dose of elbow grease.

The best way to explain the standard deviation calculation is to use an example, and in our case it is to look at the 10-year TimingCube returns of the QQQQ with a Long and Short strategy (see Table below). The steps are:

• Calculate the average (mean) annual return over the 10 years (sum divided by 10)
• Find deviation for each period (return minus 10-year mean return)
• Calculate the square of each period's deviation (deviation times deviation)
• Sum all the squared deviations
• Divide by 10
• Take the square root of that number to get the standard deviation

Standard Deviation (volatility) of TimingCube Returns for the QQQQ (Nasdaq-100 ETF),
'Long and Short' Strategy

	Return	Deviation	Deviation squared
2000	90.36%	0.50	0.25
2001	113.26%	0.73	0.53
2002	57.75%	0.17	0.03
2003	50.27%	0.10	0.01
2004	23.54%	0.17	0.03
2005	22.28%	0.18	0.03
2006	24.84%	0.15	0.02
2007	-3.94%	0.44	0.20
2008	-7.14%	0.47	0.23
2009 (partial as of 10/29/2009)	31.88%	0.08	0.01
10-year mean return	40.31%
Sum of squared deviations			1.34
Sum of squared deviations/10			0.13
Standard deviation			0.37

In the example, the 10-year standard deviation is 0.37 which does not mean anything by itself.
Since you could vary the intervals and the time period, say from 10 years to 36 months, to obtain a different standard deviation number, what really counts is comparisons with other references. In general a higher number means a more volatile and risky item. Indeed, the same calculations for the QQQQ 10 years, Long Only strategy yields a much less volatile 0.16, but this time for an annualized return of +14.6%, much less than the +33.9% obtained with the Long and Short strategy.

Another comparison sheds some light on a major flaw of standard deviation, which is that it is directionless. Looking at an investment gaining exactly 10% every year, its standard deviation is zero because none of the years deviate from the 10-year average. Ironically, an investment losing 10% every year has the same perfect standard deviation of zero!

Standard deviation really measures volatility by how much the individual samples deviate from the average of all samples, with no regard to gains or losses. This is best demonstrated by looking at the same 10-year period for the QQQQ but this time with a Buy and Hold strategy (which lost 54% as compared with the Long and Short gain of 1775%). We get a standard deviation of only 0.33, marginaly better than the 0.37 obtained when following the market trend. While the two strategies are very close in terms of volatility, the Long and Short approach is able to leverage that volatility on the positive side (making money in both up and down markets), showing a much better return at the end for the same level of risk.

This just goes to prove that risk or volatility by themselves have really little meaning and that what we should focus on instead is risk adjusted performance, which we will address in a future weekly article.

FAQ of the Week

Question: What is the volatility of the World Long and Short strategy?

When computing the standard deviation of the World Long and Short strategy, we get a reading of 0.31, which is a little less than the 0.37 obtained for the QQQQ (Nasdaq 100 ETF) as mentioned above.

This tends to prove that geographical diversification is good not just in terms of performance optimization but also in terms of return stability (for reference, when computed since January 2001 the annualized return of the World Long and Short strategy is 39% compared to 31% for the QQQQ Long and Short strategy).

Warm wishes and until next week.

The TimingCube Staff