introduction Bradley siderograph http://www.amanita.at/e/faq/e-bradley.htm
A half year ago the Bradley was discussed the last
time, so in this article I'd like to analyze what has happened in the
meantime, present the model for 2007 and also the results of a little
study.
The standard model had 5 turning dates since April which produced 3
intermediate-term and 1 short-term reversal, only 9/15/06 didn't mark
anything of significance:
- 4/11/06: 6 days afterwards 4/17 medium-term low resp.
double-bottom April 11 and 17, in this case it nailed the turn to the
day
- 5/20/06: 3 days afterwards 5/23/06 short-term low
- 6/20/06: 7 days before 6/13/06 medium-term low
- 7/23/06: 6 days before 7/17/06 medium-term low
- 9/15/06: - (nothing of significance)
The 2 remaining dates 2006 (window:
usually +/- 4 calendar days, sometimes up to +/- 1 week) are:
- 10/11/06
- 11/28/06
This is the standard model for 2007:
Next year there are only 8 potential turning points, with the last 4
being more significant (bold letters) than the first 4 (window:
usually +/- 4 calendar days, sometimes up to +/- 1 week) :
- 3/10/07
- 3/20/07
- 4/20/07
- 5/4/07
- 6/14/07
- 8/26/07
- 10/17/07 (most important date)
- 12/22/07
Some Bradley analysts would also include the "micro-spikes" of the 1st quarter but I strongly
advise against that because that's not significant enough and should be
interpreted as "white noise". For unknown reasons there are even minor
differences between the different software programs that all use the
original formula of Donald Bradley. While major turns in the Bradley chart
are more or less identical you would get entirely different results if you
zoom in too much.
Often the Bradley is misunderstood and interpreted as if it were able
to predict the market ups and downs as displayed in the chart but that's
definitely not the case, the following little study refutes this claim. I
have already been thinking about dropping the chart altogether and only
publish the dates to avoid misunderstandings, however this has the major
advantage that the magnitude of a turning point gets entirely lost (among
other drawbacks), that's why I continue to release the chart.
The study examined whether the polarity of Bradley reversal dates
correlates with the polarity of S&P 500 index turns, i.e. if a Bradley
high is also a market high (same for lows). I analyzed the standard model
since 2002 (N=34), here is the list including whether the polarity was
predicted successfully or not (date format: D.M.YY; H = high, L =
low):
(1) 27.1.02 H: incorrect
(2) 11.4.04 L: incorrect
(3) 27.7.02 L:
correct
(4) 22.8.02 L: incorrect
(5) 10.10.02 L: correct
(6)
24.11.02 H: correct
(7) 10.1.03 H: correct
(8) 13.3.03 L:
correct
(9) 26.2.03 H: correct
(10) 2.7.03 H: incorrect
(11)
14.9.03 H: correct
(12) 21.10.03 H incorrect
(13) 22.11.03 L:
correct
(14) 26.1.04 L: incorrect
(15) 6.3.04 L: incorrect
(16)
26.4.04 H: correct
(17) 17.5.04 L: correct
(18) 24.6.04 H:
correct
(19) 14.8.04 H: incorrect
(20) 28.9.04 H: incorrect
(21)
25.10.04 L: correct
(22) 27.12.04 H: correct
(23) 25./26.1.05 L:
correct
(24) 4.3.05 L: incorrect
(25) 10.6.05 L: incorrect
(26)
28.7.05 L: incorrect
(27) 30.8.05 H: incorrect
(28) 16.12.05 L:
incorrect
(29) 15.1.06 H: correct
(30) 3.4.06 H: correct
(31)
11.4.06 L: correct
(32) 20.5.06 H: incorrect
(33) 20.6.06 L:
correct
(34) 23.7.06 H: incorrect
conclusion: the polarity was called
18 times (P=53%) and missed 16 times (P=47%). That's almost like throwing
the dice, while the results do point in the right direction the deviation
from the 50% to be expected by chance is by no means statistically
significant, so one can only conclude that the Bradley siderograph does
not reliably predict whether a turning
point will be a high or a low - only the date matters.
