Point/ Counterpoint By Clifford S. Sherry, Ph.D.
I believe there are some major flaws in the statistical reasoning in Curtis McKallip's article
"Investigating Chart Patterns Using Markov Analysis" (STOCKS & COMMODITIES December
1986). Using the raw numbers from his transition matrix, I believe that the observed probabilities for the
pairs of states should be as follows: p1,1 = 58/309 = 0.1877; p1,2 = 15/309 = 0.0485.
Thus the probability of a 1, 2 or 3 is 78/309 = 0.2524; 139/309 = 0.4498 and 92/309 = 0.2977,
respectively. If we assume independence, the probability of a 1,1 is equal to p1 times p1 or 0.2524 times
0.2524 = 0.0637.