Aug

21

1st column: Date of first Hindenburg
Omen Signal

2cnd: # of Signals
In Cluster

3rd: DJIA
Subsequent
% Decline   

4th: Time Until
Decline
Bottomed

4/13/2004 (1)   5       5.4%    30 days
6/20/2002       5       15.8%   30 days
                              23.9%   112 days
6/20/2001       2       25.5%   93 days
3/12/2001       4       11.4%   11 days
9/15/2000       9       12.4%   33 days
7/26/2000       3       9.0%    83 days
1/24/2000       6       34.2%   44 days
6/15/1999       2       6.7%    122 days
12/22/1998 (2)  2       0.2%    1 day
7/21/1998 (3)   1       19.7%   41 days
12/11/1997      11      5.8%    32 days
6/12/1996       3       8.8%    34 days
10/09/1995      6       1.7%    1 day
9/19/1994       7       8.2%    65 days
1/25/1994       14      9.6%    69 days
11/03/1993      3       2.1%    2 days
12/02/1991      9       3.5%    7 days
6/27/1990       17      16.3%   91 days
11/01/1989      36      5.0%    91 days
10/11/1989      2       10.0%   5 days
9/14/1987       5       38.2%   36 days
7/14/1986       9       3.6%    21 days

Looking back at historical data, the probability of a move greater than 5% to the downside after a confirmed Hindenburg Omen was 77%, and usually takes place within the next forty-days.

The probability of a panic sellout was 41% and the probability of a major stock market crash was 24%.

Phil McDonnell comments: 

The HO signal is negated when the McClellan advance decline oscillator turns up. It turned up briefly, hence no signal. This is the case even though the MCO has now turned negative again.

The probability of making money when you sell at the high of a given move and buy at the low is 100%. Now would someone kindly tell me when those are going to occur?

Jonathan Bower writes:

I'd like to know when they would occur too!

Phil's assumption is based on the notion that 100% of the position is in place at the high. If one were to follow a strategy (not suggesting that one should for any number of reasons) that added to the position as the market fell (perhaps one was not 100% sure it was the high?) and cover some of the position if a big enough pull back occurred (oops, maybe I'm wrong about this move), then one could get chopped up with sufficient vigor such that covering the entire position on the low would not generate sufficient P/L to cover the losses that occurred from over trading. At least that's one way you could sell the high and buy the low of a move and not win….

Phil McDonnell responds: 

Perhaps I was a bit too indirect and subtle. The people who believe in the HO cite 'back testing' that is seriously flawed. Their back testing methodology requires perfect knowledge of the future. In order to duplicate the claimed results one would have to sell at the high and buy back at the low. And yes they are selling 100% at the high and 100% at the low but that is irrelevant to the broader issue of flawed analysis. This is very flawed because you have to know exactly when the high and low will occur (knowledge of the future).

Looking at results this way you need to ask yourself how often large drops occur at random. It turns out that they are a regular feature of markets. Then we ask the question whether these particular results were unlikely to be due to chance. Remember the distribution of highs in a random walk is controlled by the Arc Sine Distribution not the Gaussian. Same for the lows. The Arc Sine is a U shaped distribution which is ALL tails and not much in the middle. For a discussion of Arc Sine see Feller's An Intro to Probability Theory & Applications. Vol 1 deals with binomial random walks and Arc Sines and Vol 2 moves on to Normal random walks.
 


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search