A logarithmic point-and-figure construction with weekly breadth scores

Why the first point-and-figure chart was set aside

Early chart study began with formations on line charts in 1930. Point-and-figure was tried and then set aside, because the line-chart carried more information and could be placed on a logarithmic-scale judged better for price changes.

Point-and-figure construction was also judged inferior because it does not express volume and because an arithmetic scale distorts trend comparison. A logarithmic-scale was preferred for trend geometry.

The later hybrid for individual names

A later hybrid kept the point-and-figure idea of filtering minor moves. It applied a percentage-filter on logarithmic charts rather than posting unfiltered arithmetic boxes. For individual names, the working construction was a logarithmic point-and-figure price chart.

How the line-chart was read

A line-chart was treated as a glance-readable map of support, resistance, and trend. Computer output was required to be seen in chart form before any decision, not accepted as an automatic buy or sell mandate.

What sat beside the price chart

Market-wide work was not left to that geometry alone. It was checked against breadth and short-interest series that had been scored by batting-average.

The advance-decline-divergence-oscillator was constructed by regressing the Dow against cumulative advance-decline totals and measuring how far the average sat ahead of or behind that line. One contemporaneous panel was labeled neutral.

Odd-lot-shorts were built as the five-day average of odd-lot shorts divided by the five-day average of odd-lot purchases plus sales. Later testing described bearish calls as poor and bullish calls as useful at 13, 26, and 52 weeks.

The price-dividends-ratio of the Dow Jones Industrial Average was used as an inflation-canceling expensiveness scale. Readings above 28 were rated expensive. A long-record map marked above 28.30 as expensive and below 18.10 as cheap. A reading as high as 35 was treated as a dangerous extreme seen only once in the prior 60 years, in 1987.

A weekly batting-average and two test errors

Indicator opinions were scored every week over a 10-year sample against subsequent market direction at one, four, 13, 26, and 52 weeks, so each series could receive a batting-average by horizon.

Two statistical construction errors were named. A type-one-error is a short test that fails to show significance that a longer test later finds. A type-two-error is a short test that reports significance that a longer test removes.