Hull Moving
Average
by Alan Hull
www.alanhull.com
The Hull Moving Average solves the age old dilemma of
making a moving average more responsive to current price
activity whilst maintaining curve smoothness. In fact the
HMA almost eliminates lag altogether and manages to
improve smoothing at the same time. To understand how it
achieves both of these opposing outcomes simultaneously we
need to start with an easily understood frame of reference.
The following chart contains a 16 week simple moving average
which constantly lags the price activity and has poor
smoothness.
Firstly, solving the problem of curve
smoothing can be done by taking an average of the average,
i.e. 16 period SMA(16 period SMA(Price)). The bad news is
that it causes a huge increase in lag as seen below.
Solving the problem of lag is a bit more involved and
requires an explanation with numbers rather than charts.
Consider a series of 10 numbers from '0' to '9' inclusive
and imagine that they are successive price points on a chart
with 9 being the most recent price point at the right hand
leading edge. If we take the 10 period simple average of
these numbers then, not surprisingly, we will determine the
midpoint of 4.5 which significantly lags behind the most
recent price point of 9. Here's the clever bit…first let's
halve the period of the average to 5 and apply it to the
most recent numbers of 5,6,7,8, and 9, the result being the
midpoint of 7.
Finally, to remove the lag we take the midpoint of 7 and
add the difference between the two averages which equals 2.5
(7 - 4.5). This gives a final answer of 9.5 (7 + 2.5) which
is a slight overcompensation. But this overcompensation is
very handy because it offsets the lagging effect of the
nested averaging. Hence the result of combining these 2
techniques is a near perfect balance between lag reduction
and curve smoothing.
The HMA manages to keep up with rapid changes in price
activity whilst having superior smoothing over an SMA of the
same period. The HMA employs weighted moving averages and
dampens the smoothing effect (and resulting lag) by using
the square root of the period instead of the actual period
itself…as seen below.
WMA (2 x WMA(Price,Integer(Period/2)) -
WMA(Price,Period),Integer(SquareRoot(Period)))
The following formulas for the Hull Moving Average are
for Ensign Windows
but can be easily adapted for use with other charting
programs that are capable of custom indicator
construction. A template named HullAverage can be
downloaded from the Ensign web site using the Internet
Services form in Ensign Windows.
The HullAverage template has taken the
visual of the Hull Average line one step further by plotting
the study line in rising and falling colors of Green and Red
using the dual color line marker on Line J. (This template
requires Ensign Windows with a version date of 12-29-2005 or
later.)
Trading Tip:
Bradley Model 2005 Update
by Howard Arrington
An article about the Bradley Stock Market Model was
published in the
November 2002 issue of Trading Tips newsletter and
updated in the
May 2004 issue. Now that 18 months have gone by, it is
time for a follow up article to document the correlation of
the Bradley model for 2005 with the stock market.
The Bradley model is a forecast of the market based on
astrological relationships. Because astrological
relationships can be defined with mathematics, the Bradley
forecasts can be made decades in advance. The following
chart shows how the Bradley model correlated with the stock
market in 2005.
The end of 2004 and the first two months of
2005 had excellent correlation with the timing and direction
of the turns in the market. However, in my opinion, it
would have been difficult to trade the stock market using
the Bradley model for the balance of 2005. For a portion of
the summer and fall, there is better correlation with the
market if the Bradley model is plotted inverted.
There are three characteristics of the
Bradley forecast: time, direction,
and price. Each of these will be discussed.
The primary characteristic to be extracted
from the Bradley model is time. The model is
basically a clock based on the motions of objects in our
solar system. We readily acknowledge the influence in our
lives of the daily rotation of the earth, and the monthly
orbit of the moon, and the cycle of seasons from the annual
orbit of the earth around the sun. But beyond those three
accepted astronomical clocks, skepticism increases that life
on earth is also influenced by other astronomical bodies.
It is hard to accept that other astronomical bodies have any
influence on the business cycles of the economy of the
United States or the world. But that is what the Bradley
model is attempting to show. When the timing of market
turns is in synch with the Bradley forecast as was the case
for the turns in December 2004 (top), January 2005 (bottom)
and February 2005 (top), the forecast can be very
impressive.
As is shown in the 2005 chart, the market
and the Bradley forecast can be out of sync. For example,
the two turns in August 2005 were well aligned for time, but
out of sync for direction. The market put in a top
in mid August when the forecast was for a bottom turn. And
the market made a bottom swing at the end of August when the
forecast was for a swing top. When this happens, the
forecast is described as being 'inverted'. The timing of
the turns is still well aligned or correlated. It is the
direction into these turns that is inverted. Why does
inversion happen? Answer: I do not know and I have not
found a good answer from those who regularly work with the
Bradley model. Is there a way to know in advance when
inversion will happen? Answer: No. You can suspect the
model is inverted when it is happening, and still take
advantage of the time forecast.
The third characteristic is price,
and this should have a low priority in comparison to time
and direction. The Bradley data values are in the
range of -200 to 200 and thus to plot the Bradley forecast,
the data set has been resized and repositioned so that it
shows on the INDU chart as an overlay. The Bradley curve
has been stretched and shifted vertically until the curve
fit nicely on the INDU chart. How the curve is shown on
the chart is completely arbitrary for the convenience of
seeing the Bradley curve near the market data. At times
there is good correlation in comparing the size of one
Bradley swing with another and seeing a similar ratio in
swing amplitudes in the market. At other times, there is no
price correlation. The 2005 Bradley model suggested the
annual top of the market would be in July 2005. There was
a swing high in the summer, but it definitely was not the
top of the market for 2005. The stock market's 2005 top
occurred in March with a retest of that top in November
2005.
Be sure and read the other two articles about the Bradley
model to see other examples where there was better
correlation between the forecast and the market than was
experienced in 2005. Neither article shows the full
correlation for 2004 so the 2004 chart is shown here.
This update on the Bradley model concludes by showing the
forecast for 2006. Only time will tell whether the market
and the forecast will be well correlated, or inverted, or
down right useless as a tool for trading the stock market.
The forecast does not show a price scale on purpose. Use
the curve primarily for timing the turns, and then secondly
for the direction into those turn dates. In general the
model shows an up trend for the first half of 2006 and then
a down trend into Thanksgiving. However, the forecast into
the end of 2005 appears to be inverted, so the 2006 forecast
may also be inverted in part or in full.
|