Beruflich Dokumente
Kultur Dokumente
Article 6: “Percent b” 24
(Active Trader, April 2004).
.30
The longer the
.20
lookback period,
.10
28-8:30 12:30 29-8:30 12:30 2-8:30 12:30 3-8:30 5-8:30
the fewer signals
July
Source: CQGNet will be generated.
FIGURE 4 THE HEAD FAKE Short lookback
A Squeeze is sometimes followed by a false breakout. In this case, the BWI
narrowed to below .10, and the market broke through the resistance level periods will identify
and penetrated the upper Bollinger Band. However, the stock then reversed
and trended back down to make a new low in the downtrend. low-volatility
IBM (IBM), daily 130.00
situations more
125.00
120.00
117.72
often.
115.00
Source: CQGNet
30.00
T
27.50
F
he technique for trading the B E
Resistance
W-bottom pattern, as de-
25.00
scribed by John Bollinger in
his book Bollinger on
Bollinger Bands, blends classic technical A C 22.50
analysis and modern indicator-based D
analysis. Volume Volume spike
It’s classic because the W-bottom con-
2M
cept consists of a double bottom sup-
ported by traditional volume character-
istics: heavy volume on the first low, a 1M
price recovery and a retest of the low on
lighter volume.
The high volume on the first low indi- 0
cates a climactic selling spree: Sellers are 16 23 30 7 14 21
October
jumping ship, most likely because of a
news event. However, this low marks a Source: eSignal
bearish sentiment peak. The retest —
when the market fails to make new lows The approach also takes advantage of Bollinger Bands can help find stocks that
— occurs on low volume, which sug- the more modern statistical price meas- are setting up W-bottoms.
gests there are no more shares to sell at urement of Bollinger Bands. Augmenting The following trading approach com-
this price level. As a result, the market is W-bottoms with Bollinger Bands enables bines the implications of how price
set up to rally if there is any good news you to determine whether price action is behaves relative to 20-day Bollinger Bands
on the horizon. within a typical range. In addition, with classic price-volume chart analysis.
Regardless of the time frame, W-bottoms usually exhibit the same approach is to wait
price-volume relationships. Long trades are signaled when price pushes above
the resistance level of the W-bottom, which subsequently acts as support. until the market
Microsoft (MSFT), weekly
65.00
breaks above
60.00
the resistance level
Resistance
B
55.00
established during
D
50.00 the rebound
C Support
E
45.00 following the first
A
B 47.3185
Bollinger Bands:
Resistance Trading tool and scanner
Bollinger Bands are a statistically
45.0000
derived measurement of extreme price
A C action relative to average price action. A
42.5000
close outside the Bollinger Bands repre-
Volume Volume spike Second highest sents excessive price movement, but it
volume does not automatically mean the market
will retrace back to the moving average
50M (or the opposite band). However, a valu-
able analytical template is created by
adding the price and volume character-
istics of a classic double bottom.
0 Bollinger Bands enable you to screen
19 26 3 9 16 23 30 7 14 21 28 4
September October November for setups consisting of the combination
of a close below the lower Bollinger
Source: eSignal
Band occurring on volume greater than
the 30-day average volume. This will
ing (on declining volume) just above the weekly reversal high at bar D; the resist- produce a list of stocks with the poten-
moving average of the Bollinger Bands at ance level subsequently functioned as a tial to form W-bottoms. The support and
point B. Next, the stock turned back again, support level. resistance levels defined by the W-bot-
and at bar C it closed at the upper end of A closer look at this chart indicates Bar tom pattern allow you to construct a set
that week’s range. At this point, a success- E and the preceding bar may have of procedures for managing trades once
ful low-volume test appeared to be com- formed a double bottom of their own. the pattern is complete.
plete because the price held above the Let’s look at those two weeks using daily
lower Bollinger Band. The market should bars (see Figure 3). At bar A, the stock
be set to rally. closed below the lower Bollinger Band on
BY THOM HARTLE
BY THOM HARTLE
109
Uptrends and downtrends
In technical analysis, as in most other areas of
30-day
simple
life, you always give up something to gain
moving something else; for every advantage an indi-
average cator or approach possesses, it also has a
RSI A B C D E F G H I J drawback.
75 For example, the longer a moving aver-
age’s lookback period (the number of price
50
bars used to calculate the average), the less
February 25 likely it will result in false signals (when price
27 3 10 18 24 repeatedly moves above and below the aver-
Source: CQGNet
age), but the more it will lag changes in mar-
ket direction. Although there are several types
The RSI peaks during this period were all below 75, while downward
price thrusts pushed the indicator below 25.
The relative strength index (RSI) is an oscillator that The simple moving average (SMA) gives every price point in
ranges from 0 to 100. The formula is: the average equal emphasis, or weight. Weighted moving
averages emphasize more recent price action. The exponen-
RSI = 100 - (100/[1+RS]) tial moving average (EMA) weights prices using the following
formula:
where
EMA = SC * Price + (1 - SC) * EMA(yesterday)
RS = relative strength = the average of the up closes where
over the calculation period (e.g., 10 bars, 14 bars) SC = a “smoothing constant” between 0 and 1, and
divided by the average of the down closes over the cal- EMA(yesterday) = the previous day’s EMA value.
culation period.
You can approximate a particular SMA length for an EMA
When calculating a 10-day RSI, if six days closed higher by using the following formula to calculate the equivalent
than the previous closes, subtract the previous close from smoothing constant:
the current close for these days, sum the differences,
and divide the result by 10 to get the up-close average. SC = 2/(n + 1)
(The sum is divided by the total number of days in the where
lookback period and not the number of up-closing days.) n = the number of days in a SMA of approximately
For the four days that closed lower than the previous equivalent length.
day’s close, subtract the current close from the previous
close, sum these differences, and divide by 10 to get the For example, a smoothing constant of .095 creates an
down-close average. If the up-close average was .8 and exponential moving average equivalent to a 20-day SMA
the down-close average was .4, the relative strength over (2/(20 + 1) = .095). The larger n is, the smaller the constant,
this period would be 2. The resulting RSI would be 100 - and the smaller the constant, the less impact the most
(100/[1+2]) = 100 - 33.3 = 66.67. recent price action will have on the EMA.
Further experimentation
The RSI was used here to find trade opportunities in the Trading approach highlights
direction of the trend, as well as to indicate the trend was
shifting direction. Using the RSI to catch swing points
This is not the only tool that could function in this role 1. Define the trend: If the n-day (default: 30-day) simple
— other oscillators could be used as well. The key is to moving average (SMA) is rising, the trend is up; if the SMA is
review market action and determine practical criteria falling, the trend is down.
based on the typical behavior of the oscillator within the
context of a trend. Once that behavior is quantified, you 2. Use a short-term RSI to find swing entry points: If the 30-
can develop entry strategies and risk management points day moving average is rising and the three-day RSI drops below
for trades. 50, go long on the next up close. Reverse for short trades.
RETRACEMENT TENDENCIES
Traders often calculate likely retracement levels, but
BY THOM HARTLE is there any validity to such techniques? Objectively
defining trends and retracements makes it possible to
T
analyze the price swings in a market and determine
he market never moves in a if retracements follow any noticeable patterns.
straight line — it repeatedly
swings up and down on a
shorter-term basis as it moves
in the direction of the longer-term trend.
During a bull trend, the market FIGURE 1 DEFINING PRICE MOVES
advances and then retraces part of each
previous rise, repeating the process until Solid green lines indicate uptrending conditions and dashed red lines repre -
the long-term trend turns down. Similarly, sent downtrending conditions.
during a bear trend, the market zigzags
A Nasdaq 100 index-tracking stock (QQQ), daily
downward, with upswings retracing the
down legs of the prevailing trend. 42.00
As such, retracements represent oppor-
tunities to get on board a trend. Traders
often project retracement, or pullback lev- 40.00
C
els to use as entry points. K
It’s easy to see retracements on a chart E
in hindsight. If a market advances or
38.00
declines overall for six months, it’s obvi- G
ous how each countertrend swing repre- B
sented a retracement and potential entry D
36.00
point. I
To take advantage of retracement
opportunities as they develop, it’s neces- F
sary to first have objective definitions of 34.00
both trends and retracements. Then, you J
can determine if retracements display any H
noticeable patterns that could be used in
the future. For example, if you could 1/2/2002 2/1/2002 3/1/2002
determine with a measure of certainty the Source: Fibonacci Trader
next retracement in an uptrend would be
50 percent of the previous upswing, it would greatly simplify Before analyzing percentage retracements, we must first
entering the market on the long side. establish a definition for a trend.
Some traders believe the market frequently retraces in per-
centages derived from the numbers in the Fibonacci series (see Defining trends
“The Fibonacci series”). The most common Fibonacci There are a number of ways to define trends, such as moving
retracement targets corrections of 38.2 percent, 50 percent and averages and regression analysis. All trend-defining tech-
61.8 percent of the previous price move. niques have limitations, the primary one being lag. No matter
To determine if there are such identifiable retracement per- what a trend indicator says, you can never be sure a trend has
centages, we analyzed two years of daily Nasdaq 100 index- changed direction until after (sometimes well after) the fact.
tracking stock (QQQ) prices from January 2002 to October We’ll use a price-based technique here. Let’s start by defin-
2003. The January 2002 opening price was $42.11. The market ing an uptrend as a pattern of consistently higher highs and
fell to a low of $19.76 in October 2002 before climbing back to higher lows and a downtrend as a pattern of consistently lower
$36 in October 2003. Because, the QQQ experienced both highs and lower lows. This means support levels will progres-
falling and rising trends, we can be confident there is no sively climb (higher lows) during uptrends while resistance
unusual bias to the data in that regard. levels will fall (lower highs) during downtrends.
FIGURE 2 RETRACEMENT SIZES Although two of the follow-through moves after retrace -
ment were noticeably large, there was no pattern to
There were 28 retracements, but there was no particular
these price swings.
percentage size that appeared unusually often.
100% 300%
90% 250%
80%
70% 200%
60% 150%
50%
40% 100%
30% 50%
20%
10% 0%
1 2 3 4 5 6 7 8 9 10 1 11 21 31 41 51 6 1 71 81 9 2 02 12 2
0%
1 3 5 7 9 11 13 15 17 19 21 23 25 27
BY THOM HARTLE
T he eternal quest for would-be short-term traders sents the oversold threshold and the moving average repre-
is to find a way to identify swing points consis- sents the median line.
tently and mechanically. It cannot be done, but The strategy discussed here uses the percent b indicator both
traders never tire of experimenting with this or to define the prevailing market bias and, combined with a
that indicator in the hope it will lead to a universal rule for three-bar price pattern, to signal trade entries in the E-Mini
identifying short-term exhaustion points, or overbought and S&P 500 futures (ES) on an intraday basis.
oversold levels.
Oscillators are the indicators typically
enlisted for this role, and although there
are literally dozens of tools that fall FIGURE 1 PERCENT B
under this heading, virtually all of them
rely on similar momentum calculations The percent b indicator shows where the closing price is relative to the
that measure how much price has moved upper and lower Bollinger Bands (indicator levels 100 and zero, respectively).
up or down over a recent time period.
Typically, a high indicator threshold Continuous S&P 500 E-Mini futures (ES), five-minute
defines prices that are theoretically over- 1,074
bought and a low indicator threshold Upper Bollinger Band
identifies prices that are theoretically
oversold. A middle equilibrium or medi- 1,072
an line separates rising and falling
momentum conditions.
Bollinger Bands are a similar tool that 1,070
use a statistical calculation to measure
the degree to which price has moved
above or below its average. The indicator 1,068
consists of lines placed (by default) two 20-bar simple
standard deviations above and below a moving average
moving average of closing prices (see 1,066
Lower Bollinger Band
Figure 1, right). For a more detailed dis-
cussion of Bollinger Bands, see Percent b 150
“Bollinger Bands, variance, and standard
deviation,” and the references listed at 100
the end of the article.
The percent b indicator, which was 50
developed by John Bollinger, measures
the closing price relative to the upper 0
and lower Bollinger Bands. In a way, the 15,000
Volume 10,000
indicator is a more typical oscillator-type 5,000
depiction of Bollinger Bands for which 0
ˇ 16-8:30 9:30 10:30 11:30 12:30 13:30 16-14:30
the upper band is equivalent to the over-
Source: CQGNet
bought threshold, the lower band repre-
Bollinger Band, variance ket has become cheap relative to the market bias, or trend.
The percent b readings in Figure 2 indicated the bias or trend
and standard deviation was down. In this case, prices above the 20-bar simple moving
average meant the market had become expensive relative to
Bollinger Bands consists of three lines, as shown in the bias.
Figure 1 (p. 66): Shrewd traders seek to buy when price is cheap in an
uptrending market and sell short in a downtrending market
Upper band = 20-period simple moving average + 2 when price is relatively expensive. Next, we’ll illustrate a pat-
standard deviations tern that can be used to place trades when these conditions are
in effect.
Middle line = 20-period simple moving average of
closing prices Trade entry
An extreme percent b reading dictates the market bias or trend.
Lower band = 20-period simple moving average - 2 Although this may run counter to traditional interpretations of
standard deviations oscillator-based approaches (in that extreme readings are usu-
ally considered signs of an overbought or oversold market,
The upper and lower bands represent levels at which and an indication of an impending reversal), careful scrutiny
price is trading higher or lower, respectively, than it reveals that during trends, oscillator readings shift. That is,
has approximately 89 percent of the time over the they have persistently higher high and higher low readings in
past 20 days. uptrends, and lower low and lower high readings in down-
trends.
Standard deviation measures the degree to which With this in mind, we’ll trade based on the following con-
values in a group (say, closing prices from day to day) cept: If the trend is up, we’ll look for a buy (price) setup once
vary from the mean value. Standard deviation is price trades back below the 20-bar simple moving average —
derived from the variance of a set of numbers which, in other words, when percent b moves below 50. In this case,
mathematically, is the average squared “deviation” we’ll use the simple “pivot low” pattern shown in Figure 3 (p.
(or difference) of each number in the group from the 70): a three-bar pattern in which bar 1 is followed by a bar with
group’s mean value, divided by the number of ele- a lower high and lower low (bar 2), followed by a bar with a
ments in the group. For example, for the numbers 8, 9 higher high and higher low (bar 3).
and 10, the mean value is 9 and the variance is: Figure 3 depicts an ideal setup. The percent b indicator trad-
ed above 100, signaling an uptrend. Next, percent b dropped
{(8-9)2 + (9-9)2 + (10-9)2}/3 = (1 + 0 + 1)/3 = 0.67 below 50, indicating price was cheap on a relative basis. At this
point, you would watch for a pivot low and confirm percent b
Now look at the variance of a more widely distributed has not dropped below zero. Go long when price moves above
set of numbers — 2, 9, 16: bar 2’s high. If the market is strong, this pivot low (the low of
bar 2) should act as a support level and should not be taken
{(2-9)2 + (9-9)2 + (16-9)2}/3 = (49 + 0 + 49)/3 = 32.67 out. Therefore, an appropriate risk point is just below the low
of bar 2.
The second group of numbers are more widely distrib- Profits can be taken based on an evaluation of recent volatil-
uted and thus has a higher variance. Standard devia- ity (i.e., how much the market has tended to move lately), a
tion is simply the square root of the variance. So, the
standard deviation of 8, 9 and 10 is:
BY THOM HARTLE
Although the beginning of the first quarter of 2001 was very bullish, all but one of the ETFs — the iShares MSCT-Taiwan Index
Fund (EWT) — were in the red by the end of the quarter.
The fourth quarter was the only one in 2002 that had more than 50-percent bullishness among all ETFs. Three of the top-five
ETFs at the beginning of the quarter were still near the top at the end of the quarter.
Nearly 82 percent of the ETFs had positive returns in the first two weeks of the first quarter of 2003. This kind of broad-based
strength was more likely to result in the early-quarter leaders maintaining their dominance through the rest of the quarter.
30.00%
Jan. 12, 2001, return
20.00%
of the previous quarter. For instance,
performance for the first quarter was 10.00%
calculated by comparing the March 30, 0.00%
2001, closing prices to the Dec. 29, 2000, -10.00%
closing prices. Finally, the percentage -20.00%
returns from the top-five performers -30.00%
were compared to those of the QQQ and -40.00%
March 30, 2001, return
the SPY. -50.00%
Tables 1 through 3 show the quarter- -60.00%
ly returns for 2001, 2002 and 2003, -70.00%
respectively. The five ETFs that per-
-80.00%
formed best in the first two weeks of
each quarter are shown, along with the Source: data - eSignal; chart - Excel
final quarterly returns and where each
ETF ranked at the end of the month. The FIGURE 2 BEGINNING VS. END OF FIRST QUARTER, 2002
results for SPY and QQQ are included
for each period as well. As was the case in 2001, only a handful of ETFs were positive in the first two
In Table 1, although the first several weeks of the first quarter of 2002, while the rest were flat to down; just 30
trading days of 2001 were spectacular percent had positive returns.
(the top two ETFs were up nearly 20
40.00%
percent), the market quickly turned Jan. 11, 2002, return
down and all but one of the ETFs — the 30.00%
iShares MSCT-Taiwan Index Fund
(EWT) — posted losses for the quarter. 20.00%
Figure 1 provides some explanation
10.00%
for this. Ranking all 40 ETFs by their
performance in the first two weeks of 0.00%
2001’s first quarter, along with their
respective end-of-quarter 2001 returns, -10.00%
reveals only a relative handful of the March 28, 2002, return
funds had strong returns in the first two -20.00%
weeks of the quarter; the remaining -30.00%
funds posted near breakeven to negative
returns. This narrowness in the leader- Source: data - eSignal; chart - Excel
ship during the first two weeks of
January indicated there was a great deal FIGURE 3 BEGINNING VS. END OF FIRST QUARTER, 2003
of weakness behind the scenes in the Early in the first quarter of 2003, buying was strong across the board. The
market and, as a result, when the lead- final first-quarter 2003 returns were mixed, but the action that occurred in
ers finally gave up, the entire market the first two weeks of the period set the tone for the rest of the year.
tumbled.
To reflect the extent of the leadership
25.00%
in a given quarter, Tables 1 through 3 Jan. 10, 2003, returns
also include (in the last row, labeled “% 20.00%
15.00% March 31, 2003, returns
Bullish”) the percentage of ETFs that
had positive returns in the first two 10.00%
weeks of the quarter. 5.00%
The tables also show the QQQ and
0.00%
the SPY were below the top five the
-5.00%
majority of the time — not surprising,
since QQQ tracks 100 stocks and SPY -10.00%
tracks 500. Although this diversification -15.00%
spreads risk, it also lowers return. -20.00%
Also, what the QQQ and the SPY did -25.00%
in the first two weeks during the second,
Source: data - eSignal; chart - Excel
45.00%
third and fourth quarters of each year
40.00% indicated the outcome for the top-five
June 30, 2003, returns
35.00% ETFs: Positive QQQ and SPY returns
30.00% correlated to positive ETF returns and
25.00%
negative returns also were hand-in-
hand.
20.00% Figure 2 is similar to Figure 1 in that
15.00% it shows only a handful of ETFs were
10.00% positive in the first two weeks of the
5.00%
first quarter of 2002, while the rest were
flat to down; just 31 percent had returns
0.00%
April 11, 2003, returns greater than zero.
-5.00% Figure 3, on the other hand,
Source: data - eSignal; chart - Excel shows how broad based the buying was
in the first two weeks of January.
FIGURE 5 4TH QUARTER 2001 Although the final first quarter 2003
returns were mixed, most of the action
The Merrill Lynch Internet HOLDR (BHH) was a steady climber in the fourth occurred in the first two weeks of the
quarter of 2001. The remaining ETFs followed suit, but at a slower rate of period, and this strength set the stage
ascent. for the rest of the year. Table 3 shows
nearly 82 percent of the group had pos-
120.00% itive returns in the first two weeks.
Figure 4 shows how demand was
100.00%
BHH high in the first two weeks of 2003’s sec-
80.00% HHH ond quarter, as there was just a handful
60.00% IIH of ETFs with negative returns during
SMH this period, and even those ended up
40.00% SWH positive by the end of the quarter.
SPY An important aspect of the fourth
20.00%
QQQ
quarter for each year is, despite 2001
0.00%
and 2002 being bad years, there was
-20.00% substantial recovery. In 2003, the fourth
quarter was a continuation of the
strength of the previous three quarters
(refer to Tables 1-3).
Source: data - eSignal; chart - Excel However, the tables don’t show the
day-to-day price action and the risk you
FIGURE 6 4TH QUARTER 2002 may face by simply owning the top-five
ETFs from the first two weeks. Figures
In the fourth quarter of 2002 the ETFs peaked in early December and then 5, 6 and 7 show the day-to-day relative
declined. BHH again led the ETF pack. performance of the top-five ETFs and
QQQ and SPY. The day-to-day percent-
0.6 age changes are tracked and adjusted to
0.5 a start price of zero on the last day of
the previous quarter.
0.4 BHH
HHH Figure 5 shows BHH was a steady
0.3 climber in the fourth quarter of 2001.
SMH
0.2 SPY The remaining ETFs followed suit, but
0.1 SWH at a slower rate of ascent. In the fourth
QQQ quarter of 2002 (see Figure 6), the ETFs
0.0 XLK peaked in early December and then
-0.1 declined. Figure 7, fourth quarter for
-0.2 2003, was more volatile, with a sharp
descent in the middle of November,
and then a recovery.
This study indicates if broad-based
Source: data - eSignal; chart - Excel buying or selling in the top-volume
BY THOM HARTLE
The majority of high MFI readings during the analysis period occurred during
the 11:30 a.m. and 12:15 p.m. bars.
The market tended to trade sideways Time 8:30 9:15 10 10:45 11:30 12:15 1 1:45
for the first half of the day and as the vol- (Central) a.m. a.m. a.m. a.m. a.m. p.m. p.m. p.m.
ume decreased, the MFI rose. Bar A, the MFI > 87 0 0 6 27 47 54 20 4
first bar of the day, gapped up from the
previous day’s close. The volume (his-
togram bar a) was nearly 100,000 con-
tracts, producing an MFI of 32.70. The next bar (B), which TABLE 2 FAVORABLE AND UNFAVORABLE MOVES
started at 9:15 a.m., was an inside bar (lower high and higher AFTER HIGH-MFI BARS
low than the preceding bar) and the volume dropped off to The median MFEs and MAEs following high-MFI bars
just under 55,000 contracts (b). Consequently, the MFI rose to
50.16. Bar C’s range was less than bar B’s and the volume Occurrences Time MFE MAE
trailed off to less than 38,000 contracts (c), and the MFI edged 6 10 a.m. 3.125 -0.875
up to 52.86. At bar D, the range expanded but the volume
27 10:45 a.m. 1.500 -0.500
declined to 24,000 (d) and the MFI climbed to 106.82.
Bar E, which began at 11:30 a.m., was another inside bar and 47 11:30 a.m. 0.750 -2.000
the volume dropped to slightly below 16,000 (e), pushing the 54 12:15 p.m. 1.750 -1.125
MFI to 109.59. Bar F had a larger range but the volume only 20 1 p.m. 1.250 -1.500
reached 25,000 (f), so the MFI increased to 117.32. Note these
4 1:45 p.m. 1.375 -0.750
two bars (E and F) cover the time period most traders in
Chicago and New York would be at
lunch.
Bar G was a wide-range down bar FIGURE 2 FAVORABLE MOVES: COMPARING THE 11:30 AND 12:15 BARS
and the volume hit its peak for the day The maximum favorable excursions (MFEs) for the 11:30 a.m. bars (green)
at nearly 145,000 contracts (g). The MFI are less than those for the 12:15 p.m. bars (blue) except for three occur-
dropped to 84.21. Bar H and bar I had rences.
narrower ranges, but the volume
remained high (near 80,000) and the MFI 8.00
Maximum favorable excursion
Reflecting the tendency shown in Figure 2, the maximum adverse excursions Let’s take a closer look at the difference
(MAEs) of the 11:30 bars are consistently worse than those of the 12:15 bars. between the 11:30 a.m. bar and the 12:15
The 12:15 bars had larger favorable moves and smaller unfavorable moves p.m. bar. Figure 2 shows the MFEs for the
than the 11:30 bars. two periods sorted from smallest to high-
est. The MFE was zero for both bars four
0 times. Notice how the MFEs for the 11:30
bars (green) trail below the MFEs for the
Maximum adverse excursion
-2.00
12:15 bars (blue) except for three occur-
-4.00 rences.
-6.00 Figure 3 shows the MAEs for the two
11:30 bars time bars sorted from smallest to largest
-8.00
negative move. The MAEs for the 11:30
-10.00 a.m. bars are consistently worse than those
12:15 bars for the 12:15 p.m. bars. (By contrast, the
-12.00
12:15 p.m. bars with MFI readings below 87
-14.00 had a median MFE of 1.50 and a median
MAE of -2. The 64 times the MFI was below
-16.00
87 during the 11:30 a.m. bars, the median
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
Number of occurrences MFE was 1.50 and the median MAE was -
Source: Excel 1.25.)
Also, recall the price action in Figure 1
FIGURE 4 8:30 A.M. BARS showed the 8:30 a.m. bar had no follow
through that day after it closed and the
This chart shows the MFEs and MAEs for the one hundred and eleven 8:30 MFI was below 40. How common is this?
a.m. bars with MFI readings below 87, sorted by MFE. Figure 4 shows the MFEs and MAEs for the
one hundred and eleven 8:30 bars with
10.00 MFI readings below 87, sorted by MFE.
Not until bar 66, which spans approxi-
8.00
mately 60 percent of the observations, is
8:30 a.m. bar MFE and MAE
A squat bar (highlighted red) has both higher volume and a lower MFI reading
than the preceding bar.
33.00
This chart supports a general contention that many trends end (within one to three bars from a top or bottom) with
squat bars, but many squat bars are part of continuation moves. Squat bars with trading ranges smaller than the
previous day’s range are marked with dots below them. However, these bars failed to signal trading situations.
39.00
Nasdaq 100 index-tracking stock (QQQ), daily
38.00
37.00
36.00
35.00
34.00
33.00
ume must be greater than the previous bar’s. Figure 1 shows a ranges smaller than their preceding bars. The theory is if the
daily bar chart of the QQQ and highlights several squat bars squat bar’s range was narrower than the prior day’s range, that
(in red). day was a tight battle between the bulls and bears. However,
Because the MFI measures range relative to volume, the lower the chart doesn’t indicate these bars signal particularly favor-
the MFI, the more volume per unit of range. If volume increases able trade opportunities.
from one bar to the next, we can conclude there was more activ- If the squat bar does represent a battle between buyers and
ity on this bar than the previous bar. This is a squat bar. sellers, a short-term trading opportunity might exist because the
Williams used the term “squat” to convey the idea the mar- day after a squat bar, one side of the market will be right and the
ket was squatting and preparing to make a big jump because other will be wrong. Those who are wrong will be forced to react
the increased level of activity on such bars reflected a battle and exit their positions, which could create a trend move.
between the bulls and the bears; the battle is intense because For example, if you sold short one day and the next day the
the bar has high volume and relatively little price movement. market rallied, you’d probably take a loss at a certain point.
Williams stated almost every trend ends with a squat bar as Similarly, if you went long one day and the market headed
one of the three top or bottom bars, but he also said not every lower the next, you might be forced to sell. This idea is the
squat bar signals the end of a trend. opposite of the concept mentioned earlier of low volume indi-
Figure 2 is a daily chart of QQQ from January to September cating market participants are on the sidelines awaiting some
2004 with squat bars highlighted in red. The figure indicates event to trigger a price move.
Williams’ statement is accurate: Squat bars do often occur
within one to three bars of the end of trends, but many others The study
are followed by continuation moves. To test this idea, we analyzed daily QQQ data from Jan. 2, 2003,
The squat bars marked with dots are those with trading
to Aug. 31, 2004. Of the 418 days tracked, 85 days (20 percent) Using the squat bar’s probabilities
were squat bars, which means they appeared roughly once a The bottom line of this study is the market remains within the
week. squat bar’s range only 11.8 percent of the time. If the market
Squat bars were followed by inside days (bars with lower opens within the squat bar’s range, which happens 68 percent
highs and higher lows than their preceding bars) only 10 times of the time, there is a good chance the prior high or low will be
(11.8 percent of the time); eight squat bars were followed by broken. Therefore, you can set trade targets outside the squat
bar’s range. If you get the direction right for your original posi-
tion, the odds are good you’ll be able to capture the price move.
Additional Active Trader reading The squat bar offers a unique way to look at price and vol-
“The Market Facilitation Index,” by Thom Hartle, ume action. Many trading setups are based on momentum and
October 2004, p. 70. volume reaching low points. However, the squat bar repre-
sents the opposite approach, because its setup is based on vol-
You can purchase past articles at ume hitting a high level. Ý
www.activetradermag.com/purchase_articles.htm
and download them to your computer.
BY THOM HARTLE Company Depositary ReceiptS. HOL- DRS with their respective 2004 percent-
DRS offer the opportunity to trade a age returns. For comparison, three of the
group of stocks with a single instrument, most popular index-based ETFs — the
C
and thereby profit if a particular indus- Nasdaq 100 tracking stock (QQQQ), the
urrently, there are more try, group, or sector does well. The S&P 500 Depositary Receipts (SPY), and
than 160 exchange-traded advantage of the group ownership rep- the Dow Industrial Diamonds (DIA) —
funds (ETFs) traders and resented by a HOLDRS is you don’t also are included.
investors can analyze or have to pick the best stock. Few invest- The Market 2000+ HOLDRS (MKH)
trade. These instruments, which trade ment decisions are as disappointing as represents 50 companies trading on the
like common stock, represent a wide owning an individual company that is New York Stock Exchange, American
range of market indices, sectors, and having problems while its industry is Stock Exchange, or the Nasdaq National
groups. Here, we will look at one group doing well. (Of course, there’s always Market System that were considered to be
of ETFs — Merrill Lynch HOLDRS — the risk that a single company’s woes among the largest in terms of worldwide
and review their salient features. will negatively impact an entire group.) market capitalization on July 7, 2000. The
HOLDRS stands for HOLding Merrill Lynch has created 17 HOL- Europe 2001 HOLDRS (EKH) represents
DRS. Overall, there is a 50 of the largest European companies
TABLE 1 MERRILL LYNCH HOLDRS slight emphasis on the (worldwide market capitalization on
Internet, but groups such as Nov. 14, 2000) whose equity securities
The 2004 percentage return of the HOLDRS banking and utilities are were listed for trading on a U.S. stock
ranged from 42.08 percent (Internet) to –19.57 also represented. Table 1 market.
percent (Semiconductor). Seven HOLDRS out per- lists the seventeen HOL- Figure 1 shows the HOLDRS’ 2004
formed the broader market indexes, such as the
Nasdaq 100, the S&P 500, and the Dow Jones. FIGURE 1 2004 HOLDRS RETURNS
Symbol Name 2004 Return In this histogram bar format returns are sorted from lowest to
HHH Internet HOLDRS 42.08% highest. Seven of the HOLDRS outperformed the Nasdaq 100 track-
OIH Oil Service HOLDRS 37.21% ing ETF (QQQQ). Utilities HOLDRS benefited from a low interest
UTH Utilities HOLDRS 24.59% rate environment and Oil Services did well due to higher oil prices.
BDH Broadband HOLDRS 24.16% 2004 percentage returns
WMH Wireless HOLDRS 22.54% 50.0
IIH Internet Infrastructure 21.69%
40.0
BBH Biotech HOLDRS 13.01%
QQQQ Nasdaq 100 9.46% 30.0
2004 percentage returns
BBH
IIH
WMH
BDH
UTH
OIH
HHH
OIH was also relatively strong in 2004 — ison, other ETFs such as the QQQQ or the performers in 2004, just as the
buoyed by record oil prices — and SPY can be bought and sold in any Pharmaceutical HOLDRS (PPH) was
unlike the HHH, its strength carried amount (e.g., 10 shares of the QQQQ, 35 one of the worst performers because the
over into early 2005. shares of the SPY). problems of Merck & Co. and Pfizer, Inc.
These types of charts are good tools Exchange traded funds are excellent For detailed information on HOLDRS
for understanding the behavior of a par- tools for putting capital to work in spe- performance, see the ETF Snapshot on p.
ticular HOLDRS vs. the broader market. cific equity groups, sectors, industries, 49. To see which HOLDRS are perform-
as well as countries and fixed-income ing the best on weekly basis, refer to the
Finding a theme products. Weekly HOLDRS review on the Active
Unlike other ETFs, HOLDRS can only be Tracking the performance of the HOL- Trader Web site
bought and sold in round lots of 100 DRS can provide insight into what ( w w w. a c t i v e t r a d e r m a g . c o m ) .
shares. As a result, they are not always an investment “themes” are working or not Additional information on HOLDRS can
affordable option. For example, as of Feb. working, or which areas of the market be found at www.holdrs.com.
4, the Biotech HOLDRS was trading at are in or out of favor. Considering the
145. 45, meaning a minimum investment rise in oil prices, it is not surprising the
in the BBH would be $14,545. By compar- Oil Service HOLDRS was one of the top
The 10-year T-note futures contract is the highest-volume U.S. treasury futures contract. Over the past year, the
contract has been in uptrending, downtrending, and trading-range environments.
111-00
110-00
109-00
108-00
107-00
106-00
105-00
104-00
1,500,000
Total
1,000,000
500,000
March 2004 April May June July Aug. Sept. Oct. Nov. Dec. 2005 Feb.
Source: CQGNet, Inc.
rather than casual observation. The following analysis takes the pulse
60
50
Price has been converted to decimal format.
T reasury bonds and notes are
debt securities issued by the
U.S. Treasury. Bonds and
notes are considered debt because
by purchasing these instruments you
are, in effect, loaning money to the
Frequency
1.75
2.00
2.25
2.50
2.75
3.00
0.00
0.25
0.375
0.50
0.625
0.75
0.875
1.00
1.125
1.25
25
large amount of unredeemed bonds
20 traded in the open market.)
15 T-bonds are priced as a percent-
age of the notational value with the
10 combination of the handle, 100 for
5 example, and 32nds of 100. For
example, 98-14 is a price that trans-
0
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
90.00%
100.00%
Of the 124 times the market closed up for the day, it also closed below the
open only eight times, and only twice at the opening price.
10
5
64 days (the peak reading), or 25 percent
of the time. The daily range was less 0
0.000
0.125
0.250
0.375
0.500
0.625
0.750
0.875
1.000
1.125
1.250
1.375
1.500
0.125
0.250
0.375
0.500
0.625
0.750
0.875
1.000
1.125
1.250
1.375
1.500
Yield
A strategy that uses a multiple time frame
approach: Two indicators applied to daily
bars work together to determine the
trend; two others, Bollinger Bands and the 5
moving average convergence-divergence
(MACD) indicator, identify entry and exit 3
1965 1970 1975 1980 1985 1990 1995 2000
signals on an intraday basis.
“Relative volume analysis,” Source: Excel, Data: CQGNet, Inc.
Active Trader, July 2003.
Traders commonly use total volume to take
the pulse of price moves, but when it FIGURE 11 EMPLOYMENT FRIDAY: ONE-MINUTE CHART
comes to determining whether the bulls or
With expectations for a rise of 200,000 non-farm payroll jobs, the number
bears are in charge of the market, up and
came in at 146,000. The T-note contract started falling before the number’s
down volume speaks volumes.
release at 7:30 (CT), but rallied when the news actually came out.
“Familiarity breeds profitability,”
Active Trader, September 2002. 10-year T-note (ZN), One-minute
The author analyzes price patterns to
determine the odds that different kinds of 110-00
price moves will occur.
“Up-down volume and next-day 109-28
follow-through,”
Active Trader, December 2004.
109-24
What kind of trading is likely to occur
today? Yesterday’s balance of up volume
and down volume — and whether the mar- 109-20
ket establishes the high or low of the day
first — provides guidance in the Nasdaq
100. 109-16
Similar to the overall T-note data in Figure 3, the size of the Employment
Friday daily ranges has been declining.
Employment for a rise of 200,000 non-farm jobs, but
3.0 the number came in at 146,000.
Price has been converted to decimal format.
Price actually started falling before
2.5
the number came out at 7:30 (CT), and
2.0 then rallied when the news was
Points
4/1/04
5/1/04
6/1/04
7/1/04
8/1/04
9/1/04
10/1/04
11/1/04
12/1/04
1/1/05
2/1/05
12 employment Fridays in the analysis
period (Figure 11 represents the last
Friday in Figure 12). Similar to the over-
Source: Excel, Data: CQGNet, Inc.
all data shown in Figure 3, the size of the
employment Friday ranges has been
FIGURE 13 T-NOTE RANGES WITHOUT EMPLOYMENT FRIDAYS declining. If we remove the employment
Removing the Employment Fridays from the data decreases the typical daily Fridays from the data, the typical daily
ranges. The linear regression line is shifted slightly lower, but the slope is range drops. Figure 13 is the same as
nearly the same. Figure 3 except that employment Fridays
are removed. The linear regression line is
shifted slightly lower, but the slope is
2.0
nearly the same (-.0009 versus -.001).
Price has been converted to decimal format. Figure 14 is the frequency distribu-
1.5
tion of daily ranges without the
Points
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
2.50
2.75
3.00
The QQQQ/SPY
spread
BY THOM HARTLE
When the economic outlook sours, investors will move money out of growth
stocks and into so-called “defensive stocks” — those companies that offer products
and services consumers tend to purchase despite the weak economy, and which often
Comparing growth and defensive stocks is one way to characterize the broader
market environment and trend: When growth stocks are leading defensive stocks,
the outlook for the broad market is good. If defensive stocks are outperforming, the
outlook is negative.
A relative strength calculation of two of the most popular exchange traded funds
(ETFs) — the Nasdaq 100 index tracking stock (QQQQ) and the S&P 500 Depositary
Receipts (SPY) — allows you to gauge the stock market’s overall trend and set up
swing trades that are in sync with that trend. The setup outlined here uses an entry
concept based on the work of Richard Wyckoff, a stock market analyst and educator
From late January through early March, SPY led an uptrend against the direc- close (with the risk point above the entry
tion of the QQQQ/SPY ratio. Don’t short SPY simply because it is climbing and bar). Four days later, SPY made a new
the ratio is falling. low in the downtrend.
Springs
Nasdaq 100 index tracking stock/S&P 500 Depositary Receipts Figure 7 is an example of the QQQQ
(QQQQ/SPY), daily leading the way and trading SPY from
0.32
the long side. The RSI of the QQQQ/SPY
ratio began generating bullish readings
0.31 with a close at 61 on Sept. 20. Later in the
month, the RSI dropped below 50 for
just one day and climbed back above
RSI that level, providing further bullish evi-
60
dence.
40 On Oct. 20 SPY dropped through the
September support level and closed back
February March 20 above it, completing a spring setup and
18 24 1 7 14 22 1 7 14 signaling a long trade on the close. Three
Source: CQGNet days later, the market traded below the
entry bar’s low, so the trade was stopped
news and key market indices are making market took out the support level out. But the market reversed again, clos-
new highs. However, if the market was defined by the late-February low, as sell- ing near the high of the day and back
actually strong, QQQQ should have ing would likely accelerate, creating above the entry bar’s low. The situation
been leading the pace, not lagging so sig- good liquidity for covering short posi- qualifies as a new spring because the
nificantly — capital should have been tions. close (A) is above a support level, even
moving into technology, not out of it. A second, shorter-term trade example though the support level is three days
These conditions set up a short sale in occurred in April 2005. The RSI of the away. However, you could enter the
SPY or the S&P 500 futures. QQQQ/SPY ratio was still indicating a market after it trades above the bar’s
When the market closed back below (weak) downtrend was in place by high (B) or wait for a bar that closes back
the December high, the false breakout climbing above 50 only twice and revers- above the September support level (C).
was confirmed and a short sale was sig- ing immediately each time. A move above the early October high
naled on the close. The stop point would On April 7, SPY closed above the pre- is a logical profit-taking level, as you
be above the high over the entry bar. A vious week’s high, then closed back would be selling into a very optimistic
logical place to take profits was after the below it, signaling a short sale on the
116.00
market.
Related reading
“Trading the Wyckoff way: Buying springs ing people out. Active Trader Staff (Active Trader, June
and selling upthrusts,” by Henry O. (Hank) 2002). Relative strength can be measured
“RS System No. 1,” by Thomas Stridsman
Pruden (Active Trader, August 2000). and interpreted in several different ways.
(Active Trader, October 2000). This is a
Understanding Wyckoff’s springs and
short-term relative-strength (RS) system “Indicator Insight: Relative Strength
upthrusts can help you identify low-risk,
that compares the stock price with its Index,” by Active Trader Staff (Active
high-reward trade setups based on false
underlying market index, but also looks Trader, August 2001). The relative
breakouts.
for confirmation from volume. strength index (RSI) is a momentum oscil-
“Wyckoff axioms jumps and backups,” by lator used to identify short-term momen-
“Relative strength bands,” by Thomas
Hank Pruden (Active Trader, January- tum extremes (overbought and oversold
Stridsman (Active Trader, March 2001).
February 2001). Not every breakout will points). J. Welles Wilder, developer of the
This system uses a mix of relative
turn into a good trade, but using RSI, provides step-by-step instructions for
strength (not RSI) analysis and Bollinger
Wyckoff’s concepts of “jumps” and calculating and interpreting the indicator.
Bands to identify markets that are about
“backups” can help you sort the good
to break out of congestion areas. Special notice: From Aug. 1 to 31, these
trades from the bad.
articles will be available for 30 percent
“Finding strength with the ratio spread,” by
“The telltale spread,” by Thom Hartle off through the Active Trader online store
Thomas Stridsman (Active Trader, August
(Active Trader, November 2003). (www.activetradermag.com/purchase_art
2000). The ratio spread provides a better
Analyzing the relationship between the E- icles.htm). You can download them
alternative to moving averages for deter-
mini Nasdaq 100 and the E-mini S&P 500 directly to your computer as PDF files for
mining trend and support and resistance.
can indicate when the broad market is easy viewing and printing.
making a genuine move or when it’s fak- “Indicator Insight: Relative Strength,” by
BY THOM HARTLE
At point A, the Euro/U.S. dollar peaks, and its 10-day SMA turns down three
days later. The retracement (point B) doesn’t affect the SMA much, and price
Simple moving average rallies above its moving average well before the SMA turns up (point C).
Our first example uses a 10-bar SMA,
but technicians commonly use moving
averages as long as 200 days (approxi-
mately one year in trading days) to
define the long-term trend, as well as a
combination of moving averages such
as four-, nine-, and 21-day periods to
simultaneously track three trends.
To calculate a 10-bar SMA, add the
last 10 bars’ closes and divide by 10.
When the next bar closes, the most dis-
tant closing price in the look-back peri-
od is dropped, the new bar’s closing
value is added, and the new sum is
divided by 10.
Figure 1 shows simulated prices with
a 10-bar SMA. (Notice it takes 10 closes
to plot the first SMA value.) At point A,
the simple moving average is near the
Source: eSignal
middle of the price range. As price
begins to rise, the simple moving aver-
FIGURE 3 — 10-DAY WMA VS. SMA
age turns up a few closes later, illustrat-
ing the moving average’s inherent lag. The weighted moving average tracks price action closer than the simple moving
The average continues to rise despite a average. The WMA’s turning points (points A and C) occur before the SMA’s.
short-term retracement (point B), and Point B’s price rise did not materially affect the WMA.
both values climb after that.
Next, price forms a double top and,
as it declines, the simple moving aver-
age crests and turns down (point C).
During the drop, the SMA trended
down despite a final countertrend rise.
Figure 1 shows the benefits and
drawbacks of simple moving averages.
Despite short-term, counter-trend
movements, the SMA filters out this
noise and continues in the direction of
the trend. But the SMA’s lag is notice-
able when price changes direction.
Figure 2’s daily chart of the Euro/
U.S. dollar currency pair shows the
same characteristics as in Figure 1.
Here, the simple moving average fol-
lows the market as it peaks and then
turns down (point A). The market
Source: eSignal
FIGURE 4 — 10-DAY EMA VS. SMA point B’s countertrend move did not
derail the descent of the WMA.
The EMA turns each time price closes above or below it (points A and B), and
the EMA turns up slightly at point C.
Exponential moving averages
The exponential moving average
(EMA) uses a special algorithm (the
“smoothing constant”) that weights
price by a percentage factor. There are
two versions of the EMA calculation.
Both versions use a smoothing con-
stant to weight the closing price and
adjust the previous day’s EMA value.
The constant uses a formula to approx-
imate the value of a SMA:
SC = 2/(n + 1)
where
n = the look-back period for a simple
moving average
SC = a smoothing constant between 1
and zero
Source: eSignal
Therefore, if n = 10:
retraces (point B), but the moving average continues to SC = 2/(10+1) = 2/11 = 0.1818
trend lower. At point C, price rallies above its moving aver-
age and the MA turns up. We see the same lag at points A The first version of the EMA calculation multiplies
and C when the market reverses the trend. today’s close by the smoothing constant and yesterday’s
EMA by 1 minus the smoothing constant:
Weighted moving averages
One reason the SMA lags price is because all prices are (today’s close * 0.1818) + (yesterday’s EMA * 0.8182)
equally weighted — the current close and the first close
have the same impact on the average. A weighted moving The second approach uses the following formula:
average (WMA) uses specific multipliers to weight each
price, giving the most weight to the most recent price and (today’s close - yesterday’s EMA)* 0.1818) + yesterday’s
reducing this emphasis the further back in time you go. EMA
Thus, the most current price impacts the WMA more than
the last price in the look-back window. The formulas show how important today’s close is rela-
For example, a 10-bar WMA multiplies the current price tive to the previous day’s EMA. If today’s close is above
by 10, the next most recent price by 9, the next by 8, and so yesterday’s EMA for the first time, the EMA will immedi-
on. The sum of these weighted prices is divided the sum of ately turn up. If today’s close is less than yesterday’s EMA,
the weights, which is 55 (10 + 9 + 8 + 7… 1) in this example. the average will immediately turn down.
Figure 3 compares a 10-bar WMA (blue) to the 10-bar SMA There is no lag between today’s close and the direction of
(red). Notice the WMA crests and troughs (points A and C) the EMA. During trading ranges, however, the EMA will
ahead of the SMA. The WMA responds much quicker to the flip back and forth.
reversal patterns at the top and bottom, which took approxi- Figure 4 compares a 10-day EMA to the 10-day SMA. At
mately five days. In contrast, the SMA required either more points A and D the EMA reverses direction once the Euro
time or a more dramatic price move to reverse its trend. But crosses above or below it. It is a little less noticeable, but at
If the 20-day EMA is rising, the trend is up and price moves below the three-
day WMA represent buy opportunities (green area). If the EMA is falling, the
area between the EMA and WMA can be considered a sell zone (red).
opportunities in certain
sectors.
Name Symbol
Merrill Lynch Biotech HOLDRS BBH
DIAMONDS Trust Series I DIA
iShares Dow Jones Select Dividend Index Fund DVY
iShares MSCI Emerging Index Fund EEM
iShares MSCI EAFE Index Fund EFA
iShares MSCI Germany Index Fund EWG
iShares MSCI Hong Kong Index Fund EWH
iShares MSCI Japan Index Fund EWJ
iShares MSCI Malaysia Index Fund EWM
BY THOM HARTLE iShares MSCI Taiwan Index Fund EWT
iShares MSCI Mexico Index Fund EWW
iShares MSCI South Korea Index Fund EWY
iShares MSCI Brazil Index Fund EWZ
Merrill Lynch Internet HOLDRS HHH
F
iShares NASDAQ Biotechnology Index Fund IBB
iShares Lehman 7-10 Year Treasury Bond Fund IEF
irst launched in the early 90s, iShares Goldman Sachs Software Index Fund IGV
Exchange Traded Funds iShares S&P MidCap 400 Index Fund IJH
(ETFs) have continued to iShares S&P MidCap 400/Barra Value Index Fund IJJ
proliferate on Wall Street.
iShares S&P SmallCap 600 Index Fund IJR
Today, there are more than 170 ETFs list-
ed for trading — an increase of more iShares S&P 500/BARRA Value Index Fund IVE
than 50 in less than two years. These iShares S&P 500 Index Fund IVV
instruments, which trade like regular iShares Russell 1000 Value Index Fund IWD
stocks, represent everything from major iShares Russell 1000 Growth Index Fund IWF
stock indices to specific market sectors iShares Russell 2000 Index Fund IWM
and instrument groups (bonds, etc.). iShares Russell 2000 Value Index Fund IWN
The appeal to investors and traders is iShares Russell 2000 Growth Index Fund IWO
twofold. First, people can buy assets that
iShares Dow Jones U.S. Real Estate Index Fund IYR
closely replicate the returns of popular
MidCap SPDR Trust Series I MDY
market indices fund managers use as
benchmarks. Second, traders can use iShares S&P 100 Index Fund OEF
sector ETFs to capitalize on specific mar- Merrill Lynch Market Oil Service HOLDRS OIH
ket areas that are outperforming the Merrill Lynch Pharmaceutical HOLDRS PPH
broad market. NASDAQ-100 Index Tracking Stock QQQQ
“Following the money: Quarterly Merrill Lynch Regional Bank HOLDRS RKH
momentum and ETFs” (Active Trader, Merrill Lynch Retail HOLDRS RTH
June 2002) detailed a strategy for identi-
iShares Lehman 1-3 Year Treasury Bond Fund SHY
fying ETF leaders early in the quarter.
Merrill Lynch Semiconductor HOLDRS SMH
The premise was that money managers
will commit capital to stocks offering the SPDR Trust Series I SPY
best opportunity for profit for the quar- iShares Lehman 20 Year Treasury Bond Fund TLT
ter. Thus, the ETFs leading the market at Merrill Lynch Telecom HOLDRS TTH
the beginning of a quarter had the poten- Merrill Lynch Utilities HOLDRS UTH
tial to maintain their dominance for the Select Sector SPDR Fund - Basic Industries XLB
entire quarter. Select Sector SPDR Fund - Energy Select Sector XLE
The following analysis reviews the
Select Sector SPDR Fund - Financial XLF
findings of that article and, using the
Select Sector SPDR Fund - Industrial XLI
same procedures, explores how its con-
cepts have held up in the eighteen Select Sector SPDR Fund - Technology XLK
months since publication. Select Sector SPDR Fund - Consumer Staples XLP
Select Sector SPDR Fund - Utilities XLU
Narrowing the ETF field Select Sector SPDR Fund - Health Care XLV
Because there are so many ETFs to con- Select Sector SPDR Fund - Consumer Discretionary XLY
sider, the first step is to narrow the field
Source: eSignal
to a reasonable number by focusing on
The first column lists the ETFs from the “Following the money” article that did not make the volume cut for this
analysis. The second column shows the ETFs that replaced the deleted ETFs.
Source: eSignal
2005 performance
The first quarter of 2005 got off to a
rocky start, with just 14 percent of the 50
top-volume ETFs showing positive
returns (-2.83 percent median return),
and QQQQ down 3.72 percent and SPY
down 2.18 percent. Only 18 percent of
the top-50 ETFs were positive at the end
of the quarter (-2.46 percent median
return).
However, the initial top-five ETFs
had an average return of 2.16 percent in
the first two weeks and four of these
Source: eSignal
ETFs enjoyed a dramatic rise for the quarter, as just two failed to show
positive returns at the end of the quarter. were still in the top 10 at the end of the
quarter. Figure 5 shows the two sets of
returns sorted by initial review period.
At the start of the second quarter of
2005, only 44 percent of the top-50 ETFs
were up, but by the end of the quarter
78 percent were positive. The first two
weeks had a median return of -0.13 per-
cent. The median return for the quarter
for all 50 ETFs was 2.69 percent.
In the first two weeks of the second
quarter, the average return for the top-
five ETFs was 2.79 percent; it had
climbed to 7.22 percent by the end of the
quarter. QQQQ gained 0.57 percent for
the quarter vs. being up 0.19 percent at
the beginning of the quarter. SPYadded
an additional 1 percent from the second
Friday to the end of the quarter. Figure
6 compares the initial and end-of-month
returns for this quarter.
Source: eSignal
Although Q4 2004 reinforced the ten-
dency highlighted in “Following the
money” for fourth quarters to have the
FIGURE 5 BEGINNING VS. END OF FIRST QUARTER 2005 best returns, it is worth noting that the
period between the second Friday and
Despite the overall weak reading of positive ETFs at the start of the quarter the end of the quarter can be a volatile
(just 14-percent with positive returns and median return of -2.83), three of period. Figure 7 shows the relative price
the initial top five were positive at the end of the quarter. action based on percentage change for
the top five ETFs on Oct. 8, 2004 from
the start of the quarter to the end. Once
October had passed the ETFs started
climbing — except energy (XLE), which
was flat.
Only 44 percent of ETFs showed positive returns by the second Friday, but
the early leaders performed well, as did ETFs overall.
quarterly performance
half.
BY THOM HARTLE
The S&P 500 was only marginally positive at the end of the second quarter, and
only three of the top-10 ETFs outpeformed the index.
The average performance of the top-10 ETFs was far superior to the S&P 500’s
performance during this quarter.
After a bullish start, both the top-10 ETFs and the S&P 500 posted strong returns
this quarter, with the ETFs closing the quarter slightly stronger.
2005 performance
Table 5 shows the first quarter for 2005.
At the start, only 14 percent of the 50
most-active ETFs were positive on the
second Friday of January 2005. Those
seven are in the top-10 list, and two of
them are fixed-income instruments. The
group’s returns were wide-ranging over
Source: data — eSignal; chart — Excel
the quarter. The top performer, Select
Reviewing the best-performing ETFs from Jan. 14, 2005 shows that despite the mar-
ket’s weak start, the average end-of-quarter return for the top-10 ETFs was nearly
2 percent better than the S&P 500’s return.
The S&P 500 gained less than 1 percent from April 8 to June 30, while the average
gain for the top-10 ETFs came in at 4.13 percent.
by a mere 0.01 percent in Q1 2004,
and in Q2 2004 the ETFs posted an
average loss of 3.50 percent while Name Symbol 4/8/05 6/30/05 Difference*
the S&P was up 0.13 percent.) iShares MSCI South Korea Index Fund EWY 3.74% 0.60% -3.14%
For the entire six-quarter period, iShares NASDAQ Biotechnology Index Fund IBB 2.86% 6.85% 3.98%
the sum of the averages for the top- iShares MSCI Emerging Index Fund EEM 2.71% 5.92% 3.21%
10 ETFs was 19.67 percent, while the Merrill Lynch Biotech HOLDRS BBH 2.60% 18.83% 16.23%
sum for the S&P 500 was 9.15 per-
Select Sector SPDR Fund — Health Care XLV 2.01% 3.92% 1.91%
cent. Execution costs, slippage, and
dividends were not considered in Merrill Lynch Pharmaceutical HOLDRS PPH 1.75% 1.92% 0.17%
this review. iShares MSCI Hong Kong Index Fund EWH 1.30% 7.91% 6.60%
Tracking the top 10 performers in Select Sector SPDR Fund — Utilities XLU 1.03% 8.27% 7.24%
the first two weeks of a quarter pro- iShares MSCI EAFE Index Fund EFA 1.03% -1.07% -2.10%
vides a useful watch list for swing Merrill Lynch Utilities HOLDRS UTH 0.89% 8.09% 7.20%
trading. For example, if an ETF’s
S&P 500 SPX 0.05% 0.91% 0.86%
performance starts to lag the S&P
500, you can assume this ETF has Average change for the top 10 4.13%
lost its luster. Such ETFs are likely Percentage of 50 bullish 44.00%
weaker candidates for long-trade *Results in the Difference column may vary due to rounding.
setups, but may offer potential on
the short side (especially if the over- Source: data — eSignal
all market is weak).
Also, for the past six quarters, either
energy or biotech (sometimes both at the
FIGURE 6 Q2 2005 RELATIVE PERFORMANCE
same time) have been in the group of
top-10 ETFs. This simple screening The S&P ultimately finished the quarter with a minor gain. However, eight of
process would have alerted you to their the top-10 ETFs came in with stronger performance, led by the Merrill Lynch
strength early on. Biotech HOLDRS (BBH).
Related reading
Other ETF articles
by Thom Hartle:
BY THOM HARTLE
The day before the employment report (March 9, 2006), the MFI readings were unusually high.
Time (CT) 8:30 9:15 10:00 10:45 11:30 12:15 13:00 13:45 14:30
MFI 83.55 155.52 115.44 236.79 242.84 230.65 382.70 211.80 163.37
On March 9, 2006, the market trended lower but volume was very low. High
minute bar MFI readings was used to
MFI readings were the result.
determine what constitutes a “high”
MFI reading. In this case, any MFI read-
ing greater than the mean MFI plus one
standard deviation was considered high.
However, we need to look a little more
closely at the data before finalizing this
threshold.
The standard deviation for the current
study was 24.91, compared to 23.98 for
the previous study. The fact that the
standard deviation increased slightly
while the mean MFI reading dropped 21
percent implies some outliers (extreme,
anomalous values) are skewing the MFI
readings higher.
Sorting the MFI readings from high to
low revealed that March 9, 2006 — the
day before the monthly employment
report release — produced some very
high MFI readings relative to the rest of
the analysis period. Table 1 lists the MFI
readings and Figure 3 is a chart of that
day. Given that days immediately pre-
ceding important economic reports are
often exceptionally quiet, the high MFI
readings are not surprising.
Source: CQGNet, Inc. Removing the March 9 data lowers
the mean from 49.75 to 48.51, but the
from Jan. 2, 2004 to June 10, 2004 was respective three-month simple moving standard deviation drops from 24.91 to
63.16. For the new period, the mean MFI averages). It shows the average monthly 19.48. Adding the new standard devia-
dropped to 49.75 — a 21-percent decline. volume is higher and the average tion (19.48) to the new mean (48.51)
Figure 2 offers an explanation monthly range is lower in the last six results in a high MFI threshold of 67.99.
as to why this change occurred. It’s a months compared to the first six months. Table 2 shows the number of times the
monthly chart of the E-Mini S&P 500 Smaller ranges and higher volume trans- MFI exceeded 67.99 during the new
from January 2004 to March 2006 with late into lower MFI readings. analysis period. Table 3 shows compara-
volume and monthly range (and their The standard deviation of the 45- ble statistics from the original study,
Time (CT) 8:30 9:15 10:00 10:45 11:30 12:15 13:00 13:45 14:30
MFI> 87 0 0 6 27 47 54 20 4 0
when 87 was the MFI threshold. The original study required the MFI to exceed 87, while the new study
One thing the two studies have in required the indicator to surpass 67.99. The 13:00 and 13:45 bars had higher
common is the MFI readings for the first MFI readings during the second study period.
and last 45 minutes of the trading day
never exceeded the MFI high threshold.
This makes sense, as the beginning and
end of the trading day have the largest
volume — participants react to
overnight news in the morning and
adjust to the day’s trading going into the
close.
One difference between the studies is
the greatest number of high MFI read-
ings occurred at the 11:30 bar in the new
study and the 12:15 bar in the original
study. Also, the new study suggests
there are more high MFI readings dur-
ing the afternoon (specifically, the 13:00
and 13:45 bars) compared to 2004.
Figure 4 compares the two studies
side by side. The chart implies ranges
are not expanding and volume is not
climbing in the latter part of the day. In FIGURE 5 MFE-MAE STUDY COMPARISON
other words, the opportunities implied
The 12:15 bar has a higher average MFE for the 2005-2006 study (purple) and
by a high MFI reading at one bar are not
a more negative MAE (burgundy) than the same bar in the 2004 study (gold
developing after a series of consecutive
for the MFE and light blue for the MAE).
high MFI readings.
Time (CT) 8:30 9:15 10:00 10:45 11:30 12:15 13:00 13:45 14:30
MFI > 38.44 1 4 7 18 39 39 22 13 1
TABLE 6 E-MINI NASDAQ 100 MFE direction after this bar. The more follow-through price movement is
AND MAE READINGS current study produced an occurring in the direction of the bar’s
The average net points for the MFE of the MFE of 1.42 and an MAE of close (above the midpoint indicating an
e-mini Nasdaq 100 futures contract is higher -1.34, so this tendency up move and below the midpoint indi-
than the E-Mini S&P 500 contract. seems to have dissipated. cating a down move) after these bars.
In the current study the
Occurrences Time MFE MAE MFE is largest after the E-Mini Nasdaq 100 and Russell
1 8:30 3.000 -1.000 12:15 bar, coming in at 2000 results
4 9:15 4.000 -2.125 2.2434 — but at -2.0461, the The same analysis was applied to E-Mini
MAE after this bar also is Nasdaq 100 (NQ) and E-Mini Russell
7 10:00 3.143 -1.357
the largest. This indicates an 2000 (ER2) futures. The E-Mini Nasdaq
18 10:45 3.083 -1.917
increase in volatility but not 100 futures contract had a median MFI
39 11:30 3.090 -2.846 necessarily directional reading of 27.37 and a standard devia-
39 12:15 3.038 -3.859 movement. (Nonetheless, tion of 11.07, so an MFI reading greater
22 13:00 3.500 -2.500 the edge is towards the than 38.44 was considered high.
13 13:45 2.231 -3.346 direction of the close rela- Table 5 details the results for
1 14:30 0 0 tive to the bar.) The final October 2005 through March 2006
two 45-minute bars also (again, with March 9 removed). The MFI
had higher MFEs in the new did exceed 38.44 once each in the first
months in the second study and zero study, but the MAE for the and last 45 minutes of the day. The great-
times in the first study, this bar can be final bar in the new study was est number of high MFI readings
ignored. Although high MFI readings for -1.8125 points compared to just -0.75 occurred in the 11:30 and 12:15 bars
the 10:00 bar also were rare (only five points in the original study. More impor- (both had 39).
and six occurrences in the new and old tantly, the MFE for the final bar was Table 6 details the MFE and MAE
studies, respectively), the second study 1.5156 for the new study. In other words, numbers. The peak MFE (4.00 points)
showed a drop in both the MFE and the setup saw a negative return (+1.5156 occurred after the the 9:15 bar, but there
MAE for this bar. The 10:45 bar has near- vs. -1.8125). were only four occurrences. After the
ly the same MFE readings in both stud- The fact that over the 124 days of the 12:15 bar the MAE was larger (-3.859
ies, but the MAE reading nearly tripled current study (March 9, 2006 was points) than the MFE (3.038 points).
in size in the new study, implying non- removed) there were 133 high MFI read- Even more challenging for directional
directional price movement following ings from the 10:45 through 13:00 bars — traders is the price behavior after the
the high MFI reading. all of which have higher MFEs than 13:45 bar — an MAE of -3.346 vs. an MFE
The 11:30 bar had an MFE of 0.75 and MAEs — indicates the high MFI reading of 2.231, indicating a reversal tendency
MAE of -2.00 in the first study, which could be the basis of a trading strategy. during the afternoon trading.
indicated the market tended to reverse In other words, the higher MFEs suggest
The 11:30 bar in the latest study has a larger MFE and MAE than the original
study. The 12:15 bar has a larger MFE and a smaller MAE. The 13:00 bar has a
larger MFE and MAE.
In the E-Mini Russell 2000 futures, the MFI exceeded the high reading in the first 45 minutes of trading four times.
Time (CT) 8:30 9:15 10:00 10:45 11:30 12:15 13:00 13:45 14:30
MFI > 35.02 4 1 8 25 43 40 14 11 8
From the close of the 10:00 bar to the The analysis was built upon a very
close of the 12:15 bar, the MFEs edge out simple signal. Additional rules could be Related reading
the MAEs, with the 10:45 bar having the included to fine-tune the identification
widest margin. The afternoon trading in of a reliable trend following a high MFI “The squat bar”
the E-Mini Russell indicates a tendency reading. An additional filter could be by Thom Hartle
towards reversals, as the MAEs are larg- applied that might shift the high MAE (Active Trader, January 2005).
er than the two MFEs. readings over to the MFE column. Traders often anticipate that trend
moves will follow low-volume, low-
The lessons of the data volatility periods. The squat-bar
The analysis provides some pattern shows high-volume bars
useful insights into the can also lead to tradable price
TABLE 8 E-MINI RUSSELL 2000 MFE moves. The MFI is used as one of
ways these markets typical-
AND MAE READINGS two parameters to classify price
ly move intraday. Not sur-
prisingly, all three tend to bars into four categories.
During the afternoon, the MAE tended to be
slow down during worse than the MFE for the E-Mini Russell 2000
lunchtime in Chicago. “The Market Facilitation Index”
futures.
However, the S&P 500 By Thom Hartle
tended to trend more dur- Occurrences Time MFE MAE (Active Trader, October 2004).
ing this period in the new Traders are always looking for ways
4 8:30 1.425 -1.250
study compared to the to determine when the market is
1 9:15 0 -1.700
original 2004 study. poised to make a move. The
8 10:00 1.238 -0.825 Market Facilitation Index is a tool
The E-Mini Nasdaq 100 25 10:45 1.344 -0.952
futures had a higher MAE you can use to identify market lulls
43 11:30 1.535 -1.349 that can precede price trends.
than MFE following the
40 12:15 1.638 -1.600
close of the 12:15 bar. All
three markets had larger 14 13:00 0.893 -1.757 You can purchase and download
MAEs than MFEs follow- 11 13:45 1.236 -1.355 past articles at
ing the close of the 13:45 8 14:30 0 0 www.activetradermag.com/pur-
bar. chase_articles.htm.