The Meaning of Minor McClellan Oscillator Changes

To the best of our knowledge, it was the late Kennedy Gammage who came up with this hypothesis and was its chief proponent.  Gene Morgan also used to make mention of it on his Charting The Market television show in Los Angeles during the 1970s through the 1990s.  Others have also picked up on the belief in it, and so we took a look at it a few years ago from a statistical point of view, just to see if there was anything to it. 

One of the problems we face in conducting such a study is defining the terms.  How "minor" is a minor change?  What's a big move?  And how soon is the big move supposed to come?  Ken Gammage used to contend that it was most likely on the next trading day after the minor Oscillator change, but it could take up to 4 days to see it.  So we ran through several permutations of thresholds, as you can see.
 
The two tables below list the results of our own analysis.  We used data from 1990 to 2003 as our sample population of data.  In the first table, we ask what is the probability of a big move tomorrow, if today is a minor change day.  For measuring the magnitude of the price move, we used the percentage change in the NYSE Composite Index.  We used the absolute value of that move, making no distinction between up moves or down moves.  We are just looking for magnitude.

Amount    >.5%       >1.0%   >1.5%    >2.0%    >2.5%    >3.0%
of Chg                       
<1 pt.     35.71%    14.29%    8.57%    4.29%    1.43%    0.00%
2            38.82%    15.79%    8.55%    3.29%    1.32%    0.66%
3            39.66%    15.95%    6.47%    2.59%    1.29%    0.43%
4            40.13%    15.61%    6.69%    2.87%    1.27%    0.32%
5            41.25%    17.75%    8.36%    2.61%    1.31%    0.26%
6            42.38%    18.10%    8.17%    2.65%    1.32%    0.22%
7            43.67%    18.15%    8.32%    2.84%    1.32%    0.19%
                      
Random: 46.28%    21.23%  10.08%    4.17%    2.03%    1.18%

So looking in the first row, for example, if the Oscillator were to change by less than one point, then the statistics show that there was a 14.29% occurrence of a price change of more than 1% up or down the next day.  But the curious thing is that for that study period, 21.23% of the trading days saw a move of that size or greater.  So for that particular set of criteria, our study found that the market was LESS likely to see a big move on the day after an unchanged reading in the McClellan Oscillator.

Since the effect of a minor Oscillator change might be delayed, we continued the study to see what the probability of occurrence of a big move was over the next four trading days.  Here are those results:

Amount    >.5%       >1.0%   >1.5%    >2.0%    >2.5%    >3.0%
of Chg                       
<1 pt.     81.03%    32.76%   10.34%    5.17%    1.72%    0.00%
2            80.18%    29.73%    9.91%    4.50%    0.90%    0.00%
3            79.11%    30.38%    8.23%    4.43%    1.90%    0.00%
4            78.67%    28.44%    9.00%    5.21%    1.90%    0.00%
5            77.52%    29.46%    9.30%    4.65%    1.94%    0.00%
6            77.74%    30.65%    9.35%    4.19%    1.94%    0.00%
7            77.46%    29.58%    9.58%    4.23%    1.97%    0.00%

                      
Random: 88.03%    53.28%  29.19%   13.01%   6.27%    3.78%

As you can see in the table, if we define a big move (in the NYSE Comp) as anything greater than 1/2 a percent, then there is a random probability of 88% to see such a move within any 4-day period (over the years we studied).  That sort of result makes any sort of predictor of such an event self-fulfilling. 
 
We really wish that the minor change "rule" did work, because it would be a nice clue to have, especially for a options spread trader.
 
One improvement we might have made to the study would be to examine the magnitude of the moves in terms of Average True Range instead of absolute percentages.  In other words, if you have a really quiet market, like 1993 for example, then a 0.4% move might be a "big" day compared to what has been going on recently.  We could pretty easily do that study in a few hours' time, but given these result presented here it does not appear that there is enough to this supposed phenomenon to be worth further study.