Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

A High Probability Buy Signal Just Occurred

|Includes: SPDR S&P 500 Trust ETF (SPY)

In all of the testing that I have done of objective market indicators, which is extensive, those based on NYSE breadth have worked the best. And the best of the best do not signal very often. (By "objective" I mean those that require no interpretation and can be computed algorithmically to create precise buy and sell signals -- in contrast to subjective things such as chart patterns, Elliott Waves, etc.)

About 20 years ago I learned of a then new book entitled Hidden Patterns: Power Money in the Stock Market, by Robert Kinsman, and that it was based primarily on NYSE breadth. I purchased a copy and began to study it and test its strategies in real time. As happens with nearly all such books that I have read, most of the time its methods did not work. Moreover, it was obvious that most of the "patterns" were of such complexity that they were over-optimized and thus most likely were doomed from the start.

However, unlike nearly all such books that I have read, it does contain a few discoveries of value. One of the best patterns, called the "Stopper", occurs during a strong rally. Kinsman states that it predicts a pause, after which the rally may resume, or it may fail. The way he says to use this pattern works often, perhaps even most of the time, but in 2003 I discovered an alternate way to use it with more predictable results in a strong bull market. I later refined this into a very precise algorithm. If breadth during and after the stopper's development is sufficiently strong, there is a very high probability that the rally will continue. Perhaps more importantly, there is very little chance that the market is on the precipice of a huge decline.

In developing the algorithm I used only NYSE breadth data beginning in 1965, as that was all I had. (Kinsman's data began in 1972 and ended in 1994 when his book was published.) There have been four signals since I completed the algorithm in 2012, and all of those have worked. Then in 2015 I obtained NYSE breadth data from 1928 to 1965, and was quite surprised to discover that there had been 12 instances of this pattern in that data, and 11 of them had worked. Thus out-of-sample success = 15/16. There have now been 38 instances of this pattern since 1928, and 33 of them have been profitable. Results using S&P 500 daily closes are as follows:

Begin date | End date || Result ||| 6 mo. result

======= ======= ====== ========

1932-08-13 1932-08-15 +7.20 pct. +0.58 pct.

1935-08-19 1935-10-15 +4.73 pct. +21.28 pct.

1938-10-28 1938-11-09 +4.95 pct. -15.83 pct.

1943-03-10 1943-03-29 +5.41 pct. +8.26 pct.

1947-07-22 1947-08-11 -3.40 pct. -4.03 pct.

1949-08-23 1949-10-06 +4.42 pct. +14.37 pct.

1954-08-06 1954-09-21 +4.64 pct. +20.97 pct.

1958-10-14 1958-12-16 +4.51 pct. +12.62 pct.

1961-02-13 1961-03-02 +4.43 pct. +11.69 pct.

1962-12-10 1963-01-14 +4.71 pct. +12.82 pct.

1963-05-06 1963-09-03 +4.50 pct. +5.08 pct.

1964-07-24 1965-01-27 +4.52 pct. +4.82 pct.

1967-02-17 1967-04-18 +4.52 pct. +6.51 pct.

1971-02-18 1971-04-07 +4.53 pct. +1.73 pct.

1972-01-21 1972-03-06 +4.94 pct. +3.74 pct.

1972-12-12 1973-02-02 -3.63 pct. -11.98 pct.

1975-02-24 1975-03-17 +5.61 pct. +3.67 pct.

1976-02-26 1976-07-09 +4.86 pct. +1.96 pct.

1976-12-20 1977-02-11 -3.31 pct. -2.92 pct.

1980-06-19 1980-07-14 +4.67 pct. +18.00 pct.

1982-09-20 1982-10-07 +5.13 pct. +24.73 pct.

1985-02-21 1985-05-20 +5.29 pct. +3.95 pct.

1985-12-23 1986-02-14 +5.37 pct. +19.67 pct.

1986-03-25 1986-05-28 +5.07 pct. -1.06 pct.

1989-02-10 1989-04-18 +4.79 pct. +18.37 pct.

1991-02-26 1991-04-02 +4.60 pct. +9.28 pct.

1997-06-23 1997-07-08 +4.57 pct. +6.16 pct.

2001-02-16 2001-02-21 -3.55 pct. -10.47 pct.

2003-05-19 2003-05-30 +4.65 pct. +13.21 pct.

2003-06-19 2003-09-18 +4.51 pct. +9.88 pct.

2004-01-28 2004-03-24 -3.29 pct. -2.55 pct.

2009-05-13 2009-06-01 +6.67 pct. +23.71 pct.

2009-08-11 2009-09-10 +5.01 pct. +10.11 pct.

2010-03-25 2010-04-23 +4.42 pct. -1.55 pct.

2012-02-10 2012-03-15 +4.47 pct. +4.68 pct.

2013-02-04 2013-03-14 +4.51 pct. +13.49 pct.

2014-03-11 2014-06-09 +4.48 pct. +6.31 pct.

2016-03-15 2016-06-02 +4.43 pct. +6.65 pct.

==============================

6 mo. result = gain/loss at 130 trading days from begin date.

The algorithm just signaled again on August 2. Its history concludes, with 0.86 probability, that the S&P 500 (daily close) will reach 2252 or higher before it reaches 2086 or lower. If the profit point is reached, in most cases the rally has further to go: 24 of the 33 initially successful were profitable after 6 months. Similarly, if a loss occurs, the most likely outcome is further decline. However, not shown are the facts that the greatest decline that ever has occurred during the 130 days following a signal is -16.4% (after the 1938 occurrence), and only 3 other occasions (1972, 2001, and 2010) resulted in a decline of more than -8%. This strongly suggests that those hoping for an imminent, huge bear market will be disappointed.

Disclosure: I am/we are long SPY, IWM.