1996issue C011-5
Constructing a smoothed advance-decline trend filter
This article teaches market breadth as a three-layer construction: turn the daily advance-decline count into an exponentially smoothed oscillator, mark a rolling six-month normal range, and withhold a medium-term trend-change hypothesis until a second same-direction outside spike appears after a gap of at least a month.
- Build market breadth from the composite tape as each session's advancing issues minus declining issues, and keep an advance-decline line as the running sum.
- Turn the daily net into a seven-day exponential oscillator by combining 25 percent of the current net reading with 75 percent of the prior average, and allow about 40 sessions for a start-date move to fade.
- Read that oscillator against a six-month norm of ordinary highs and lows. Treat a print at least 40 percent beyond the band as an outside spike, and count spikes inside two to three weeks as one event.
- Withhold a medium-term trend-change hypothesis until an outside spike, a return to the pre-spike normal range, and a second same-direction spike at least a month later.
A three-layer construction
Editorial framing: TradersWeek reads the archive workflow as a construction sequence with three layers. The first layer records market breadth. The second layer turns that daily count into a smoothed oscillator and a six-month norm. The third layer is a trend filter that withholds a medium-term trend-change hypothesis until a second same-direction outside spike appears after a gap of at least a month.
From the daily count to a smoothed oscillator
The advance-decline line is a running sum of the number of stocks that close higher minus the number that close lower in each session. Market breadth is that daily participation, used as a market-wide signal rather than as a single-index price path.
Net advance-decline figures are taken from the composite tape of listed issues rather than from an exchange's first internal advance-decline print.
A breadth oscillator is formed by applying a 25 percent, seven-day exponential average to daily net advances minus declines. Exponential smoothing here means combining 25 percent of the current net reading with 75 percent of the prior average.
The exponential series may be started on any session. About 40 trading sessions is described as enough for a large move near the start date to stop distorting the reading.
A six-month norm and outside spikes
Six-month clusters of prominent highs and lows on the smoothed series define ordinary reversal bands. Prints well outside those bands are treated as candidate medium-term trend alerts. That typical high and low cluster over a recent half-year that contains no large spikes is the six-month norm.
For an April-through-October 1993 window described as spike-free, the average of ten prominent highs was +229 with tops near +300, and the average of prominent lows was -222 with troughs near -300.
Readings at least 40 percent beyond the prevailing band are treated as more informative than 20 percent or 30 percent overshoots. Spikes clustered inside two to three weeks are counted as one event. An outside spike is a smoothed reading at least 40 percent beyond the current high or low norm.
The confirmation sequence
A medium-term trend-change sequence is specified as an outside spike, a later return to the pre-spike normal range, and a second spike in the same direction separated by at least one month. The trend filter upgrades an oscillator extreme to a medium-term trend-change hypothesis only after that full sequence.
Major trend changes from 1973 through the mid-1990s are described as each showing an outside spike within six months of the matching S&P 500 top or bottom, and most within two months.
When the ordinary range shifts
During a later correction the lower six-month band widened from about -220 to -300 toward about -200 to -500, while the upper extreme stayed near +300 and typical rally tops moved closer to +100.
All readings on this track · 71 readings
- 1987How a failed rebound, weak breadth, and cycle dates broke the 1987 bull case
- 1988Diagnosing market bottoms with breadth, divergence and averages
- 1988Diagnosing index tops with breadth divergences
- 1988Record highs versus seven-day breadth and divergence
- 1989Constructing a percentage-scaled internals composite
- 1989Constructing a weekly block-tick breadth z-score
- 1989Constructing a dual-rate advance-decline oscillator
- 1989Normalize advance-decline series for a common-scale comparison
- 1990Unchanged-issue share as a narrow-breadth case study
- 1990Evaluating daily and weekly unsigned plurality breadth
- 1990Constructing paired new-high and new-low breadth indicators
- 1990Ten-day HI/LO extremes as a long-horizon breadth signal
- 1990Confirming index cycles with breadth, volume, and waves
- 1990Index cycle gates from breadth and volume
- 1990Constructing advance-decline breadth indicators
- 1990Weekly advance-decline oscillator: weight map, extremes, and spike cycle
- 1990Price-weighted construction distorts breadth, support, and trend
- 1991A peak-sequence test from the new-highs-to-advances-ratio
- 1991Fuzzy rules that turn daily market-breadth into a session consensus
- 1991From daily breadth tallies to a weighted consensus signal
- 1991Retesting market-breadth when market structure changes
- 1991Constructing TRIN as a breadth-volume ratio
- 1991Build the market clock before you read a price bar
- 1991A construction audit of the long-horizon trading index
- 1991Independent formula timers kept as a testable combination
- 1992When identical TRIN prints come from different pairings
- 1992Grade closing tick before a next-session breadth hypothesis
- 1992Noncumulative advance-decline swing confirmation
- 1992Five-day sum construction of the trading index
- 1992Daily closing-trin extremes and next-day direction
- 1992A three-layer audit: regime, breadth, and group RSI
- 1992Constructing a nine-state trend, momentum, and breadth score
- 1993Constructing a market-volume-impact rating from nested averages
- 1993When advance-decline confirmation counts the wrong universe
- 1993Constructing breadth momentum from advance-decline smoothing
- 1993Constructing a cumulative market-thrust line
- 1994Three-horizon construction of the Haurlan index
- 1994Checklist-gated session entry in 1993 index futures
- 1994Read one advance-decline pair through three windows
- 1994Constructing calibrated market-breadth summation indexes
- 1994Constructing a two-speed advance-decline oscillator and a calibrated summation
- 1995NYSE tick extremes and candlestick reversal entries
- 1995Assembling range, breadth, and a stored stop into one procedure
- 1995Restating market breadth timing rules as ratios
- 1995Constructing breadth ratio gates after lookback drawdowns
- 1995Building a short-range breadth and price oscillator
- 1996Constructing a smoothed advance-decline trend filter
- 1996Smoothed advance-decline alerts at the 1987 and 1990 turning points
- 1996Constructing breadth, RSI, and stochastic range filters
- 1996New-high and new-low counts as a breadth construction
- 1996Constructing the four-input breadth-volume ratio
- 1996Constructing the McClellan oscillator and a calibrated summation index
- 1996Declare the oscillator seed, then calibrate only the summation index
- 1997Three-gate centered strength in market-breadth construction
- 1997Daily advance-decline and new-high new-low breadth signals
- 1999Index-fund positions as a majority-vote committee
- 2000Tick, tiki and TRIN as a three-layer session confirmation stack
- 2000Constructing an advance-decline oscillator from one listed tape
- 2001Market breadth, beta, and volume-price confirmation
- 2001Regime context from relative venue volume, breadth, and intermarket spreads
- 2002When NYSE breadth misreads operating-stock participation
- 2003Two-gate breadth divergence and a trend filter for rally tops
- 2003Market internals confirm or diverge from the index
- 2004Constructing the McClellan oscillator and summation index
- 2005Intraday index-futures divergence as a three-part session hypothesis
- 2005Breadth summation levels as a short-term signal filter
- 2005Checking trend versus range with breadth and divergence
- 2011Constructing a Nasdaq hi-lo index from highs, lows, and issues traded
- 2013Cumulative advance-decline versus a one-year average
- 2013A one-year breadth average as a participation gate
- 2015Falsifying a healthy correction with breadth and support