Skip to main content
Track Momentum strategy
21 / 51
Search the archiveLibrary
Library

1996issue C041-8

Jump and hold filters for long-term Treasury yield direction

A mechanical procedure reads a gold-mining index through a short-lag percentage jump and a multi-week hold above or below an in-position extreme, then scores the locked rules on a second yield sample. The same flags invert from yield to bond and become a cash exit when they are overlaid on a fund that cannot be shorted.

  • The procedure treats a gold-mining index as a directional gauge for long-term yields because published inflation prints describe the past, while gold is used as a store-of-value proxy for perceived future inflation.
  • GMIjmp% is a one-to-four-week percentage jump that marks a sudden change in inflation perception, and AbvBlw requires a set number of weeks above the in-short low or below the in-long high before a buy or sell.
  • Five parameters are chosen jointly on a parameter-search-segment for net result, win rate, drawdown and nearby-value stability, then locked before any run on an out-of-sample-segment.
  • Buying the 30-year Treasury yield series is defined as selling the bond, and a government-securities fund overlay can treat the short side only as a cash exit.
Entries in this reading3 entries
Current trackMomentum strategy103 readingsMechanical trading system158 readingsMoving average449 readings

A gold-mining index as a directional gauge

The procedure treats a gold-mining index as a directional gauge for long-term yields because published inflation prints describe the past, while gold is used as a store-of-value proxy for perceived future inflation.

Among several commodity and mining series tested as predictors of long-term rate direction, the gold-mining index was selected on empirical grounds. A shorter gold-and-silver exchange index was described as similar but too short to treat as valid.

A short-lag jump and a multi-week hold

The first filter is GMIjmp%, the percentage change in the gold-mining index over a short lag of one to four weeks. It is intended to mark a sudden change in inflation perception. Here that short-interval percentage jump is treated as a discrete, lagged rate-of-change baseline rather than a smoothed average of prices.

A moving-average in this procedure is therefore a defined-lookback transform of ordered observations. The momentum-strategy rule treats a large, rapid change in the gold-mining proxy as a shift in inflation perception and therefore as a directional cue for long-term yields.

The second filter is AbvBlw. The state turns positive after the proxy stays above the lowest low recorded while short for a set number of weeks, and it turns negative after the proxy stays below the highest high recorded while long for a set number of weeks. The mechanical-trading-system buys after the above-low hold and sells after the below-high hold.

Together the jump thresholds and the multi-week persistence tests fully specify entry, exit and abstention. The mechanical-trading-system is applied without discretionary override.

Choosing five parameters on one sample only

Five parameters are chosen jointly: the up-jump threshold, the down-jump threshold, the jump lag, the weeks above the short-side low, and the weeks below the long-side high. They are judged together for net result, win rate, drawdown and nearby-value stability, after first reducing the search space rather than scanning all five blindly.

Development is required to use at least two separate data segments. The parameter-search-segment is the historical window used to choose jump percentages, lag length and persistence weeks. The out-of-sample-segment is a separate historical window used only after parameters are locked, to test whether the same rules still produce a usable directional result. That later window is not used to pick the rules.

Locked values on two weekly yield windows

On weekly 30-year yields from June 1985 through December 1995, locked values of a 24 percent up jump, a minus 20 percent down jump, a four-week lag and 50-week persistence on both sides produced eight trades, of which seven were profitable. Those locked values belong to that window.

The same locked parameters applied to weekly 30-year yields from January 1975 through January 1986 produced six trades, of which five were profitable.

Weekly 30-year Treasury yields, mid-1990 through 1995

Yields fall from the late-1990 spike near 9.2 percent to a late-1993 low near 5.8 percent, reverse in the 1994 backup to about 8.2 percent, then grind down to the 5.94 percent weekly close printed for 29 December 1995. A trader watching the locked jump-and-hold rules would see that long decline, the 1994 backup, and the final slide as the test-window path the system had to navigate. Coordinates were read off the weekly curve in the source pane; the last print is taken from the chart header.
Yields fall from the late-1990 spike near 9.2 percent to a late-1993 low near 5.8 percent, reverse in the 1994 backup to about 8.2 percent, then grind down to the 5.94 percent weekly close printed for 29 December 1995. A trader watching the locked jump-and-hold rules would see that long decline, the 1994 backup, and the final slide as the test-window path the system had to navigate. Coordinates were read off the weekly curve in the source pane; the last print is taken from the chart header.US 30-year Treasury yield · Weekly · 1990-07-01T00:00:00.000Z to 1995-12-31T00:00:00.000Z

Weekly closes were digitized from the published raster, so turning points are approximate to about five basis points. The 29 December 1995 close of 5.94 percent is the printed header value, not a pixel interpolation. Numbered signal arrows on the source pane are omitted.

The same signals on a fund that cannot be shorted

When the same signals were overlaid on a government-securities fund from September 1982 through December 1995, short-side results were treated as a cash-exit teaching case because the fund cannot be shorted.

The author notes that the fund window is only a few years longer than the yield test, so the overlay can inherit curve-fitting from the yield series.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
21 of 51 in the Momentum strategy track
19971-7 pp.Next on Momentum strategyConstructing extendedness from a 10 percent swing filterA percentage swing filter declares a new trend only after price moves at least the chosen threshold from the last swing high or low, and it must keep direction, the swing-pivot, and current extent as state.
All readings on this track · 51 readings
  1. 1984Half-cycle differencing for momentum signals
  2. 1987Relative strength evaluation under competing optimization criteria
  3. 1988Weekly MACD as a two-clock momentum confirmation stack
  4. 1989Equal-weight zero-cross from smoothed spreads
  5. 1989Testing relative-strength-index reversal rules against trend continuation
  6. 1989Smoothed three-day futures filter for index option bounces
  7. 1989Cycle-length windows for momentum, Relative Strength Index, and stochastic construction
  8. 1991Filtered rally magnitude as a bull-regime breakout
  9. 1992Constructing true strength from double-smoothed momentum
  10. 1992Constructing a double-smoothed true strength index
  11. 1993When momentum structure and breadth break together
  12. 1993Constructing double-smoothed range and momentum oscillators
  13. 1993Constructing a two-parameter relative momentum index
  14. 1993Building a bounded momentum oscillator with RSI smoothing
  15. 1993Two-speed oscillators with divergence and trendline gates
  16. 1993Constructing a relative momentum index from the relative strength index
  17. 1994Constructing daily advance-decline breadth tools
  18. 1994Building a composite regime score from monetary climate and weekly trend
  19. 1994Averaging Relative Strength Index and the stochastic oscillator into one reversal oscillator
  20. 1995Dividend-yield regression as a hold versus momentum gate
  21. 1996Jump and hold filters for long-term Treasury yield direction
  22. 1997Constructing extendedness from a 10 percent swing filter
  23. 1997A range-expansion oscillator that can refuse its own stretch
  24. 1997RSI trend permission and Fibonacci pullback rules
  25. 1998Nested midpoint construction for a range-normalized oscillator
  26. 1999Evaluating Relative Strength Index momentum with zero-line and threshold rules
  27. 1999Treat RSI and momentum as three mechanical procedures
  28. 2000Thrust strength figure from moving-average swings
  29. 2001Constructing a non-range-bound balance of market power score
  30. 2004Cleaned breadth oscillator and new-high divergence: a swing-market case file
  31. 2004Evaluating advance-issues-momentum on a fixed-symbol-basket
  32. 2004Constructing a trend filter from two adjacent high-low windows
  33. 2006A dollar-versus-commodity extreme as a regime case
  34. 2008Zero-centered stochastic bands and bracket stops
  35. 2012Evaluating engulfing momentum across hold windows
  36. 2012Staged stops as one mechanical entry and exit procedure
  37. 2012Stacking a relative-strength-index forecast, a trend filter, and long-only momentum
  38. 2013Constructing fair-value filters from averages and momentum
  39. 2014Constructing a multi-window slope divergence entry
  40. 2015Bandedge trend filter construction with inverse crossover rules
  41. 2017Opposite rules for index price and volatility momentum
  42. 2018A three-state overlay that colors a trend only after the line clears the bar
  43. 2018Half-cycle relative-strength index with a Fisher map for cyclic reversals
  44. 2018Emotion as a rule input when momentum breaks
  45. 2018Two-bar body expansion as a momentum breakout construction
  46. 2019Pair a two-day high breakout with a volume-weighted exit on the same chart
  47. 2020Building reflex and trendflex cross and extreme entry rules
  48. 2020Construct a dual-series price momentum oscillator overlay
  49. 2020A multi-timeframe stochastic as a panel of weekly voters
  50. 2020Centerline crossovers that compare index momentums
  51. 2020Multi-timeframe stochastic voting as one mechanical rule
By fieldIndicators and filters682 readingsBy year199680 readings
All 103 readings tagged Momentum strategy
Also on Momentum strategy5 readings
1993Constructing a two-parameter relative momentum index1994Building a composite regime score from monetary climate and weekly trend1999Treat RSI and momentum as three mechanical procedures2004Evaluating advance-issues-momentum on a fixed-symbol-basket2004Evaluate rank rotation only where persistence already exists