Two-gate breadth divergence and a trend filter for rally tops

Two gates, not a tape read

Editorial reading: a mid-rally advance can look like a durable regime change or like another bounce that is already losing sponsorship. The archive does not settle that identity from price alone. It waits for two separate conditions before a topping hypothesis is treated as timed.

The first gate is a price-indicator divergence. The index keeps advancing while a smoothed market-breadth oscillator rolls over, implying thinner internal sponsorship behind the rise. The second gate is a moving-average trend filter, used only after that split is already visible, to test whether the intermediate direction has flipped.

Two readings from the same four inputs

The market-breadth oscillator multiplies advancing issues by advancing volume and declining issues by declining volume, then divides the difference of those products by their sum. The reading is confined between -1 and +1, and the positive side is read as constructive.

An older two-ratio breadth index uses the same four inputs. It is described as neutral at 1, constructive from 1 toward 0, and weakening above 1. The constructive side is bounded, the weakening side is unbounded, and the polarity is opposite most oscillators.

Smoothing for the intermediate horizon

For inspecting intermediate swings on daily bars, a 22-session exponential average of the market-breadth oscillator is applied. Intermediate horizon here means a swing lasting several weeks to a few quarters. A long horizon is treated as longer than one year.

A rising composite with weaker sponsorship

On the Nasdaq Composite from August 2001 through July 2002, price continued higher after late November 2001 while the 22-session average of the market-breadth oscillator turned lower. The archive frames that split as deteriorating internal sponsorship behind a rising tape.

Confirmation after the split

After such a split has been visible, confirmation is a change from an uptrend to a downtrend. For that composite, the archive uses a 35-session exponential average of an ease-of-movement series crossing below zero. A 50-session exponential average of price is offered as a slower alternative.

Illustrated confirmations include a zero-cross two sessions after that composite's all-time peak, a reading at the end of May 2001 after an April 2001 advance, and a 6 December 2002 flip after a November split on a rally that began in October 2002.

The same sequence on a large-cap composite

For a 500-name large-cap composite whose own issue-level breadth is not readily available, the write-up substitutes the sum of national-market and exchange-wide breadth, citing that composite as more than 80 percent of domestic capitalization. The moving-average trend filter on that series is a 25-session exponential average of ease of movement.

The same sequence, a market-breadth oscillator divergence with price and then the trend average turning lower, is shown on that 500-name composite in December 2001 and again after a February 2002 bounce. The stretch includes a brief early-February 2002 instance that the first composite did not show.

What the sequence is for

Editorial note: the archive is not asking the first gate to name a top. It is asking the second gate to say whether the intermediate direction has flipped after sponsorship has already thinned. That is a timing drill, not a claim about present-day markets.