Skip to main content
Track Price-indicator divergence
10 / 16
Search the archiveLibrary
Library

1999issue C091-7

Stochastic rules versus buy and hold

The stochastic oscillator locates the close inside a lookback range and can be read through %K versus %D, the 80 and 20 bands, or price-indicator divergence. A mechanical trading system comparison codes those readings as rule variants and scores them against buy and hold on the same window for each security.

  • The stochastic oscillator places the close inside a chosen lookback range, so downtrends pin closes near the low and uptrends near the high until those positions reverse as a trend ends.
  • Three readings are distinguished: %K versus %D, position versus the 80 and 20 bands, and price-indicator divergence. Only a divergence followed by a %K/%D cross was originally a complete signal.
  • A mechanical trading system coded raw crosses, 80/20 filters, band exits, and joint-threshold crosses on daily bars and scored long and short trades in points against buy and hold.
  • Buy and hold produced the largest net gain on the full 32-name panel, while a band-exit variant led on a six-name detailed subset and still required active management.
Entries in this reading3 entries
Mechanical trading system7 readingsStochastic oscillator9 readingsCurrent trackPrice-indicator divergence19 readings

Where the close sits in the range

The stochastic oscillator locates the close inside a chosen lookback range. Downtrends tend to pin closes near the low of that range and uptrends near the high. Those positions reverse as a trend ends.

Three reading methods

Three reading methods are distinguished: %K versus %D, position versus the 80 and 20 bands, and price-indicator divergence versus price. The original presentation treated only a divergence followed by a %K/%D cross as a complete signal. The other two readings were confirmation or warning.

Even in that original divergence setup, %D had to sit beyond 80 or 20 for a bearish or bullish implication. Those same bands later became standalone overbought and oversold markers.

How %K and %D were paired

Fast %K uses a selected range period, commonly 5 or 10 bars, and may be averaged for slowing. %D is a slower average of %K. The illustrated default pairing is a five-bar range with no slowing and a three-bar simple average of %K.

Mechanical rules against buy and hold

A mechanical trading system coded several rule variants on daily bars: raw crosses, 80/20 filters, band exits, and joint-threshold crosses. Both long and short trades were taken at the next open after a signal. A fixed $5 commission was subtracted and no slippage was applied. Results were scored in points against buy and hold.

The same-period test window ran from 4 January 1993 to 2 July 1999 across 32 named equities. Later start dates were used only for Allstate, Gateway, and Safeskin. Every method used that security's identical window.

What the panel showed

Across the 32-stock panel, buy and hold produced the largest net gain. Thirty-one names were profitable under buy and hold, versus 24 under the best-performing coded system.

On the six names given detailed statistics, the band-exit variant (method 3) beat buy and hold on those stocks. It showed expectancy of 3.64, 0.019 net per bar versus 0.016 for buy and hold, 65% winning trades, and nearly half of winners on the short side. It still required active management and delivered less total net than buy and hold on the full 32-name set.

Thirty-two stocks: net-gain count by stochastic rule versus buy and hold

Buy and hold finished profitable on 31 of the 32 names and had the single best net on 17 of them. Among the coded oscillator rules, method 3 — enter on a %K/%D cross only when %K is beyond the 20/80 bands and exit when %K recrosses those bands, with no slowing — was the only variant that came close, with a gain on 24 names and four best-net finishes. Counts are taken from the 32-stock summary in the source spreadsheet, not from a redrawn chart.
Buy and hold finished profitable on 31 of the 32 names and had the single best net on 17 of them. Among the coded oscillator rules, method 3 — enter on a %K/%D cross only when %K is beyond the 20/80 bands and exit when %K recrosses those bands, with no slowing — was the only variant that came close, with a gain on 24 names and four best-net finishes. Counts are taken from the 32-stock summary in the source spreadsheet, not from a redrawn chart.32 U.S. stocks · Daily · 1993-01-04T00:00:00.000Z to 1999-07-02T00:00:00.000Z

Daily bars, next-open fills, $5 commission and no slippage. Each method uses the same window on a given name. Methods 1–5 vary %K slowing and whether the 80 and 20 bands filter entries and exits. Test window is 4 January 1993 to 2 July 1999 except later starts for Allstate, Gateway and Safeskin.

Editorial note

This is an editorial reading, not an archive claim. The stochastic oscillator is better treated as a family of testable rule sets than as a single chart cue. Crossings, 80/20 bands, and price-indicator divergence should each be judged by whether a coded combination survives a documented out-of-sample comparison with an explicit baseline. The archive workflow compared coded variants with buy and hold on a shared same-period window for each security.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
10 of 16 in the Price-indicator divergence track
20011-4 pp.Next on Price-indicator divergenceConstructing confirmation filters for RSI overbought and oversold extremesOverbought and oversold readings are constructed extremes that an oscillator tends to traverse; they raise the chance of a pause or turn without proving that price has reversed.
All readings on this track · 16 readings
  1. 1989Volume confirmation windows and exponential average construction
  2. 1990Constructing stochastic %K and %D from range position
  3. 1990Build a weekly leading sector composite from scaled transports and financials
  4. 1990Constructing stochastic K and D lines and divergence cues
  5. 1993Relative strength index events depend on the chosen input combination
  6. 1995Constructing a dual-horizon force index
  7. 1996Building a range-normalized divergence index from relative strength index
  8. 1998Treat RSI as a testable filter rather than a trigger
  9. 1999Primary-cycle windows, then stochastic confirmation
  10. 1999Stochastic rules versus buy and hold
  11. 2001Constructing confirmation filters for RSI overbought and oversold extremes
  12. 2003Constructing divergence-equivalent relative strength index and stochastic oscillators
  13. 2003Reverse-engineered RSI as a next-close projection
  14. 2003Scoring open versus resolved relative strength divergences
  15. 2003Bull-and-bear-balance from OHLC bar patterns
  16. 2003Constructing bull and bear balance from session paths
By fieldIndicators and filters56 readingsBy year19995 readings
All 19 readings tagged Price-indicator divergence
Also on Price-indicator divergence5 readings
1990Constructing stochastic K and D lines and divergence cuesStochastic oscillator1990Constructing stochastic %K and %D from range positionStochastic oscillator1999Primary-cycle windows, then stochastic confirmationStochastic oscillator2003Constructing divergence-equivalent relative strength index and stochastic oscillatorsStochastic oscillator1994Stacking scored filters into a hierarchical stock outlook systemMechanical trading system