Skip to main content
Track Seasonality analysis
11 / 41
Search the archiveLibrary
Library

1997issue C061-6

Lunar phase delay as a testable seasonal regime

An archive evaluation treats lunar phases as an external cyclic regime and encodes full-moon and new-moon events as a one-parameter long/short switch. The delay is searched until the pairing is market-specific, then judged by whether a lunar lock can be separated from a one-way trend or a simple end-of-month effect.

  • The mechanical trading system buys a set number of bars after a full-moon event and sells the same number of bars after a new-moon event, with no other parameters, stops, or price filters.
  • A delay of zero aligns buys with the full moon and sells with the new moon; a delay near 14 reverses that pairing, which is why the search ran from 1 to 20.
  • Wheat selected delay 1, the NYFE composite selected delay 10, and silver selected delay 17, where entries clustered a few days after the new moon.
  • A balanced long/short sample was used in place of a formal significance test on wheat, and the lunar lock is treated as a regime that can fade when another force dominates.
Entries in this reading3 entries
Current trackSeasonality analysis65 readingsSeasonal trading90 readingsMechanical trading system172 readings

What was tested

The evaluation treats new, first-quarter, full, and last-quarter lunar phases as an external cyclic regime that might mark bottoms or tops. The idea is tested on several markets with end-of-day data.

A one-parameter switch

The mechanical trading system uses a single delay. It buys a set number of bars after a full-moon event and sells the same number of bars after a new-moon event. There are no other parameters, stops, or price filters. The mechanical trading system and the seasonal trading procedure are the same template: one delay, and no extra filters.

A delay of zero aligns buys with the full moon and sells with the new moon. A delay near 14 reverses that pairing to the opposite half of the lunar cycle.

How phase events are dated

Moon-phase dates come from an external function that takes a bar date and a phase code and returns the next date of that phase. The codes are 0 for new, 1 for first quarter, 2 for full, and 3 for last quarter. The procedure feeds the function the date from five bars earlier.

Overnight full-moon and new-moon events are flagged when the phase date falls between the current session date and the next session, including weekends.

Market-specific delays

The delay was searched from 1 to 20 so a reversed-phase solution near 14 could appear. Wheat selected delay 1, the NYFE composite selected delay 10, and silver selected delay 17.

At silver's selected delay of 17, entries clustered a few days after the new moon and exits a few days after the full moon. Buy dates drifted relative to calendar month-starts. That drift was used to argue against a simple end-of-month effect.

Wheat daily price under a one-day lunar delay

With delay set to one day the system buys after the full moon and sells after the new moon. Wheat climbed from about 455 to 585 cents a bushel in April 1996, then slid to a 373 close on 21 February 1997. Several marks sit near turns; others arrive late. Prices were read from the published daily chart axis; the final 373 close is the figure printed in the header.
With delay set to one day the system buys after the full moon and sells after the new moon. Wheat climbed from about 455 to 585 cents a bushel in April 1996, then slid to a 373 close on 21 February 1997. Several marks sit near turns; others arrive late. Prices were read from the published daily chart axis; the final 373 close is the figure printed in the header.Wheat futures · Daily · 1996-02-01T00:00:00.000Z to 1997-02-28T00:00:00.000Z

Delay L1 equals 1. Market orders only, no stops. Digitized levels are approximate to about five cents except the header close. The sample window on this figure is the last year of the 1990–1997 test, not the full run.

Competing drivers and a fading lock

Because the system alternates long and short in equal counts and similar time in market, a one-way price trend cannot by itself explain both sides of the sample. The wheat evaluation did not include formal tests of statistical significance. The case for a lunar rhythm rested on a large, balanced long/short sample rather than a significance test.

The discussion treats the lunar rhythm as one of several possible driving regimes. When another force or a strong trend dominates, the lunar lock can fade. Seasonality analysis is then an evaluation problem: detect when that regime is present.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
11 of 41 in the Seasonality analysis track
19981-8 pp.Next on Seasonality analysisSeasonal system construction without curve-fittingA trading system is a complete set of mechanical buy and sell rules based on predefined indicator or price values, plus stop-loss, position-size, and overall risk limits.
All readings on this track · 41 readings
  1. 1989Evaluating venue volume as a speculation-breadth signal
  2. 1991A thirty-name price-weighted average as a seasonal regime classroom
  3. 1991Ranked half-year rate changes as an equity signal filter
  4. 1991Demographic wave as a market-regime overlay
  5. 1994Seasonal range regimes as a futures context overlay
  6. 1996Evaluating presidential party terms as equity regimes
  7. 1996A dominant cycle is a baseline, not a reprint
  8. 1996Seasonality and presidential election cycle regimes
  9. 1997Stacking calendar regimes around election years
  10. 1997Calendar seasonality as a testable trading procedure
  11. 1997Lunar phase delay as a testable seasonal regime
  12. 1998Seasonal system construction without curve-fitting
  13. 1999Crowd life cycle as a market regime map
  14. 2001Regime-dependent cycle timing after four-year and seasonal lows
  15. 2002A 2002 case study in regime-first seasonal selection
  16. 2002Seasonal windows and dominant-cycle rules
  17. 2002Fifty-four-year wholesale cycle as an inflation-deflation regime map
  18. 2004The championship conference rule as a yearly regime case study
  19. 2004Election-year seasonality as trade regime context
  20. 2006Two-ten inversion as an intermarket regime filter
  21. 2006Midterm-to-presidential seasonal holding window
  22. 2008Two-layer equity regimes from seasonality and price history
  23. 2008Seasonal futures as a regime filter, not a calendar rule
  24. 2008Retesting seasonal rules when regimes change
  25. 2010Corn and wheat staggered calendars as dollar-neutral seasonal spreads
  26. 2011Name the S&P 500 trend regime before using weekly and monthly seasonality
  27. 2012Seasonal windows that wait for confirmation
  28. 2013Pair-sleeve rotation as a two-state sector regime-switch
  29. 2013Lunar phase as a seasonal overlay on implied volatility
  30. 2013Soybean seasonal highs in a five-year carryover regime
  31. 2014Year-end tax-loss selling as a seasonal regime
  32. 2017Calendar-window overlays that mute mechanical signals without rewriting the system
  33. 2017A four-year cycle and volume case study of a secular bear
  34. 2017Treat the valuation climate as climate and implied-volatility extremes as weather
  35. 2019Seasonal depth versus tracking for futures position sizing
  36. 2019Stacking cycle forecasts with seasonal regimes
  37. 2019Hit-rate gates for seasonal regime evaluation
  38. 2019July to October as a seasonal window, not a reason to own the name
  39. 2020A recession-regime checklist from valuation stretch and the yield curve
  40. 2020Treat a seasonal idea as a stay-or-sit holding procedure
  41. 2020A single position as a sleeve on a seasonal regime map
By fieldSeasonality and regimes131 readingsBy year199760 readings
All 65 readings tagged Seasonality analysis
Also on Seasonality analysis5 readings
1997Calendar seasonality as a testable trading procedure2017Calendar-window overlays that mute mechanical signals without rewriting the system2019Hit-rate gates for seasonal regime evaluation1989Weekday price paths are regime-dependent1990Calendar windows as testable index-futures procedures