Hand Study of Reverals of Length 1, from Victor Niederhoffer
I have received many ironic comments about my speculations that a quantification of the point and figure approach might be particularly applicable to the Dow Jones. My speculation was prompted by the inordinate tendency of the Dow to bounce around the round numbers during the day. For example, during the last two days the Dow has broken above and below 12,300 by at last five points, 12 times on a intra day basis. I felt that there must be many who look inordinately for such reversals, and that they wouldn't come when the masses expected them. Indeed, it seemed guaranteed that there was a higher form of deception here, akin to what birds, fish, and mammals do all the time.
To get a feel for it, I decided to do a hand study of the Dow from the beginning of the year when the Dow conveniently started on 12 28 06 at 12,501. I decided to count only moves that went through a new round number of 50 or 100 using closing prices.
To quantify it, I decided to go back to my roots. One day, some 45-years ago in Lamont Library I was looking at some old journals, which in those days were bound conveniently for perusal on the second floor. I came across the Monthly Meteorological Review and saw an intriguing article by Jones on calculation of runs. It turned out he worked for Alfred Cowles at the Cowles Foundation. And Cowles divided the price moves into sequences and reversals.
A sequence is a plus followed by a plus or a minus followed by a minus. A reversal is a plus followed by a minus or a minus followed by a plus. I must remark here that Cowles insights into stock price movements circa the 1910s and 1920s seem much superior to any I have come across in the many years since, among academics, except for those of my esteemed friend MFM Osborne.
I decided not only to take into account the sequences and reversals but to quantify them by length. In particularl I have always felt that a series of prices ++++- was quit different in its predictivity than +-+-+-.
Here are some closing prices of the Dow from 1228
Date Closing Pos Pos Neg Neg
price seq rev seq rev
12 28 12501
12 29 12463
01 03 12474
01 04 12480
01 05 12398 1
01 08 12423
01 09 12416
01 11 12514 1
01 12 12582 2
01 22 12477 1
01 24 12621 1
01 25 12502 1
01 26 12487 1
01 30 12621 1
02 01 12673 2
02 09 12580 1
02 13 12654 1
02 14 12741 1
02 15 12765 2
2 22 12686 1
2 23 12647 1
2 27 12216 10
3 02 12114 12
3 05 12050 13
3 6 12207 1
3 8 12260 2
3 12 12318 3
3 13 12075 3
3 15 12159 1
3 19 12226 1
3 20 12288 2
3 21 12447 4
3 22 12461 5
03 27 12397 1
03 28 12300 1
What does one conclude from a hand study like this? There were many moves in Jan, a stasis in Feb leading up to a negative sequence of length 13, which wouldn't have been able to get into until 10 of them had elapsed and would have led to a slight loss. Since then there have been more frequent moves without much ground being covered. The negative reversals of small length as I had hypothesized were not good to reverse. Doubtless there are a million other ways of analyzing or tabulating this data. Indeed, I wrote a program to do this with the Wiz about 10 years ago, but because there was so much qualitative involved, one has not exhumed it. However, I think it's a good suggesting things for meals for a lifetime.