Skip to main content
Track Trend following
37 / 38
Search the archiveLibrary
Library

2020issue C0328-31

Moving-average construction as an objective trend filter

A moving average is built from ordered prices, a chosen lookback, and a weighting rule so a trend filter can classify direction instead of treating every short-term wiggle as a signal.

  • A moving average averages ordered prices over a chosen lookback so short-term fluctuations are filtered and a smoother path remains for reading underlying direction.
  • A simple moving average treats each period equally, while an exponential moving average weights recent prices more heavily and responds faster at the same lookback.
  • Lengthening the lookback makes the line smoother and less responsive; shortening it makes the line faster and more sensitive to changing prices.
  • Price above a rising average can be read as an objective bullish classification, and price below a falling average as an objective bearish classification.
Entries in this reading3 entries
Current trackTrend following168 readingsTrend filter184 readingsMoving average597 readings

How a moving average is constructed

A moving average is constructed by averaging ordered prices over a chosen lookback so short-term fluctuations are filtered and a smoother path remains for reading underlying direction. The lookback is the sampling length used to compute the average, and it jointly sets smoothness and lag.

A trend filter then uses price relative to that constructed line, and often the slope of the line, to classify the market as trending, not trending, or no longer trending.

Weighting and lookback set speed against smoothness

A simple moving average treats each period equally, while an exponential moving average weights recent prices more heavily and therefore responds faster to new prices at the same lookback.

Lengthening the lookback makes the constructed line smoother and less responsive. Shortening it makes the line faster and more sensitive to changing prices.

A longer simple average filters more short-term noise than a short exponential average, but the delayed cross means a larger share of the initial move has already occurred before the filter flips.

Short 21-period EMA versus longer 100-period SMA on hourly WTI

After the reversal the 21-period exponential average is crossed first and then climbs much faster than the 100-period simple average, leaving about a dollar of WTI between them—the extra confirmation the slower average demands. Readings were taken from the hourly curves on the source chart; 58.53 and 58.00 are the last printed 21-EMA and 100-SMA values.
After the reversal the 21-period exponential average is crossed first and then climbs much faster than the 100-period simple average, leaving about a dollar of WTI between them—the extra confirmation the slower average demands. Readings were taken from the hourly curves on the source chart; 58.53 and 58.00 are the last printed 21-EMA and 100-SMA values.WTI crude oil · 1 hour

Hourly WTI crude oil. The source labels only day-of-month ticks (19–27) and does not name the calendar month. Endpoints 58.53 (21-EMA) and 58.00 (100-SMA) are the chart’s last printed readings; other points are digitized from the curves to the nearest 0.05.

Confirmation keeps a short average from firing on every cross

A short exponential average used as a buy-above or sell-below rule can produce many false crosses when price oscillates without a clear trend, and more consistent signals only after a trend is already underway.

Pairing a short-average cross with an independent confirmation, such as a pattern breakout or a later opposite-side invalidation, is one construction used to cut false trend-following signals. Confirmation is an extra condition so a short average is not traded on every cross.

Objective bullish and bearish classification

Price above a rising average can be read as an objective bullish classification, and price below a falling average as an objective bearish classification, which is the filter’s main teaching use for early trend-following. Trend-following takes entries, holds, or abstains only when the constructed filter says a trend is present, rather than fading every excursion around the average.

A 200-period daily simple average is treated as a widely shared reference, so touches, crossings, and failed breaks around that line are expected to coincide with heightened participation. That construction is long enough to act as a shared trend reference rather than a short-horizon trigger.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
37 of 38 in the Trend following track
202024-27 pp.Next on Trend followingConstructing first and second cross trend entriesA moving-average-oscillator is the difference between a shorter and a longer moving average of price, and an oscillator-trend-line is a moving average of that difference used for cross signals.
All readings on this track · 38 readings
  1. 1986A Marshall Wave book is assembled one confirmed stance at a time
  2. 1987Operator gut check when two systems conflict
  3. 1988Visual trend rules need one testable procedure
  4. 1988Audit trend and breakout rules sold without tests
  5. 1992Weekly 4 percent close reversal case study
  6. 1992Write down the plan, then wait for the cross
  7. 1994Constructing a three-rule moving-average system
  8. 1995Weekly long-only trend rules with a gated retracement exit
  9. 1997Constructing rainbow charts with repeated smoothing
  10. 1997Three-line break as a one-parameter mechanical filter
  11. 2000Nested horizon trend alignment with a momentum stack
  12. 2001When certainty turns off the stop
  13. 2001Price must confirm every setup
  14. 2001Constructing a color stack from every lookback
  15. 2002Evaluating mechanical trend signals from ten-session average excursions
  16. 2004Trend pullbacks with a fifty-day average and relative strength index
  17. 2005Stochastic overbought as a trend-regime state
  18. 2005Trend-following as one testable signal procedure
  19. 2007One checklist for entry, exit and abstention after a market switch
  20. 2008Parallel listings and divergent trend behavior
  21. 2008Recovery headlines versus trend-regime rules
  22. 2008Constructing an intraday trend procedure with confirming close, retest, and planned exits
  23. 2010Mechanical trend-following system design, stops, and validation
  24. 2010A trading-plan that sets entry, stop, trail, and pause together
  25. 2010Multi-horizon cycle signals, trend filters, and stops
  26. 2012Weekly bear-market and shock signals as a mechanical procedure
  27. 2013A 2013 rates week as a structure-first exit drill
  28. 2013Trend following as one testable procedure
  29. 2013A forex veto checklist: trend, pair economy, and crowd bias
  30. 2015Relative-strength-index rules under a 20-percent trend filter
  31. 2015Trend-gated oscillator entry and exit rules
  32. 2016Trend process versus end-goal trading
  33. 2017Gold as a classroom for one midterm-trend procedure
  34. 2017Falling-knife FX longs wait for a four-hour trend turn
  35. 2018A closed procedure for quarterly sector rank slots
  36. 2018A recursive median as shared state for trend rules
  37. 2020Moving-average construction as an objective trend filter
  38. 2020Constructing first and second cross trend entries
By fieldIndicators and filters874 readingsBy year2020165 readings
All 168 readings tagged Trend following
Also on Trend following5 readings
1995Weekly long-only trend rules with a gated retracement exit2001Constructing a color stack from every lookback1992Ranking futures trend tendency against a random walk2007Constructing a veto-first trend permission stack2015Bandedge trend filter construction with inverse crossover rules