Skip to main content
Track Robustness testing
41 / 51
Search the archiveLibrary
Library

2013issue C1230-33

Walk-forward filter selection for repeated-median velocity

Walk-forward analysis, an in-sample metric screen, and Robustness testing turn a three-parameter Trend filter into a repeatable keep, discard, or sit-out procedure. Filter choice, not a peak profit row, is what maps each window onto the next out-of-sample week.

  • The repeated-median velocity procedure uses three tunable inputs: lookback length N and two signed velocity thresholds that separately gate long and short signals.
  • Each of 105 walk-forward windows produced 2400 input combinations, so a reusable filter, not a single peak profit row, maps the in-sample case onto the next out-of-sample week.
  • Simple ranking by net profit or profit factor rarely carries valid average profits forward. A stricter screen drops high-activity cases, keeps the 20 rows with the fewest losing bars, then takes the highest remaining in-sample net profit.
  • The same in-sample rule can output an abstention week when no qualifying inputs fire, and session bounds are an evaluation choice rather than a given of the velocity rule.
Entries in this reading3 entries
Current trackRobustness testing58 readingsWalk-forward analysis95 readingsTrend filter137 readings

Filter choice is the evaluation object

Editorial reading: Walk-forward analysis is used here to make a Trend filter testable as one procedure. The designer is not naming a permanent lookback. The designer is choosing, in each window, which input row to keep, which rows to discard, and whether the next week should sit out.

The historical workflow applies that reading to a repeated-median velocity rule with three tunable inputs. Robustness testing then asks whether the chosen screen is distinguishable from random filter picks.

Three inputs on one-minute bars

The repeated-median velocity procedure uses three tunable inputs: lookback length N plus two signed velocity thresholds that separately gate long and short signals. Those signed gates are the Trend filter. Long and short are not required to share one velocity cutoff.

Parameter search on one-minute Russell 2000 emini bars spanned N from 20 to 70 in steps of 10 and both velocity thresholds from 0.5 to 10 in steps of 0.5.

Paired windows, not a single peak row

The evaluation used 105 paired windows of 30 in-sample days followed by seven out-of-sample days covering March 30, 2011 through May 3, 2013.

Each window produced 2400 input combinations, so filter choice, not a single peak profit row, is what maps an in-sample case onto the next out-of-sample week. Editorial reading: specifying a screen that can be reused on the next window is a different act from selecting the top profit cell in the current window.

Why net profit ranking is a weak selector

Simple in-sample ranking by net profit or profit factor is treated as a weak selector because it rarely carries valid average profits into the following out-of-sample window.

A stricter screen first drops cases with 80 or more in-sample trades per month, then keeps the 20 remaining rows with the fewest losing bars, then takes the highest in-sample net profit from that shortlist. The profit sort comes last, and only after the activity and losing-bar cuts.

One-row filter versus random picks

Applying that one-row filter to every window yields 105 out-of-sample weeks whose average profit, losing-week streak, equity smoothness, and chance probability can be compared with a bootstrap of random filter picks. That comparison is the Robustness testing step.

On the selected filter, the out-of-sample equity path tracked a second-order polynomial with an R-squared of 0.96 before costs and 0.92 after the 30-unit round-turn cost.

Abstention weeks and session bounds

Weekly activity under the same filter ranged from zero trades to 33 trades, averaging 8.9, so the in-sample rule can also output an abstention week when no qualifying inputs fire.

Restricting execution to 8:30 a.m. through 3:30 p.m. Central time omitted overnight sessions in a 22-hour contract, so session bounds are themselves an evaluation choice, not a given of the velocity rule. Editorial reading: entry, exit, and abstention are not fully specified until those session bounds are stated with the three velocity inputs.

Walk-forward OOS equity, b20tLB<80-tnp filter on TF

Net out-of-sample equity on one Russell 2000 e-mini (TF) contract climbs from break-even to about $99,000 over 105 walk-forward weeks, staying in a rising channel except for a two-week pullback around the week ended 7 December 2012. Gross equity before the $30 round trip ends near $126,000. Weekly points were read from the published Figure 2 plot; the article states the finished net figure as $99,040.
Net out-of-sample equity on one Russell 2000 e-mini (TF) contract climbs from break-even to about $99,000 over 105 walk-forward weeks, staying in a rising channel except for a two-week pullback around the week ended 7 December 2012. Gross equity before the $30 round trip ends near $126,000. Weekly points were read from the published Figure 2 plot; the article states the finished net figure as $99,040.Russell 2000 e-mini (TF) · Weekly out-of-sample equity from 1-minute bars · 2011-05-06T00:00:00.000Z to 2013-05-03T00:00:00.000Z

Y-values read from the raster to the nearest $1,000. The test used 105 windows of 30 in-sample days and seven out-of-sample days on TF one-minute bars, with N, vup and vdn re-selected each window and a $30 round-trip cost. Some weeks the filter found no qualifying inputs and the system sat out.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
41 of 51 in the Robustness testing track
201432-35 pp.Next on Robustness testingWalk-forward evaluation for fading-memory velocity systemsUntrended price meandering is treated as noise, and a velocity threshold withholds trades until estimated trend speed leaves that deadband.
All readings on this track · 51 readings
  1. 1986Degrees of freedom in trading system optimization
  2. 1988Walk-forward and neighborhood tests after optimization
  3. 1988Undisclosed rules block system robustness tests
  4. 1988Testing re-optimization calendars against random parameter controls
  5. 1989Binary search limits on multi-peak average grids
  6. 1989Parameter neighborhoods that survive a shift
  7. 1990Use profit mapping to keep a cycle and stop plateau
  8. 1990Why popular indicator optimization fails robustness
  9. 1991Retesting weighted indicator balances across horizons
  10. 1992Constructing forecast models with regression, walk-forward, and robustness
  11. 1992Diagnose regimes before you lock parameters
  12. 1992When stops change system timing
  13. 1993Walk-forward halt rules for forecast models
  14. 1994Walk-forward evaluation of genetic index rules
  15. 1995Input pruning as walk-forward system evaluation
  16. 1995Critiquing neural nets as incomplete trading systems
  17. 1996Rebuild the equity-path ratio before it ranks a designed system
  18. 1996Parameter grids can fit random walks
  19. 1996Walk-forward analysis belongs in the design of a mechanical trading system
  20. 1997When a holdout fails, discard the rule set
  21. 1997Test rewarded rule breaks before replacing the system
  22. 1997Walk-forward rules keep system research from rewriting live trades
  23. 1999Keep a channel-breakout to two lookbacks and test neighbor stability
  24. 1999Constant investment size in stock system evaluation
  25. 2000Forcing optimization maps mechanical system failure boundaries
  26. 2000Robust parameter selection with surface charts
  27. 2001A two-gate classroom test for a two-window momentum trend filter
  28. 2002How a two-sided continuation factor becomes a testable trend rule
  29. 2002Evaluating two-window trend intensity as a reversal rule
  30. 2003Discounting speculative bubbles in system robustness tests
  31. 2003Walk-forward evaluation of locked stochastic oscillator rules
  32. 2003Critiquing mechanical system design after extreme price regimes
  33. 2004Evaluating a two-window trend trigger
  34. 2005Grade backtested signals with holdouts and optimization plateaus
  35. 2006Reserved-sample evaluation of trading system design
  36. 2006Walk-forward critique of hindsight crossover systems
  37. 2008Condition-matched walk-forward evaluation for mechanical systems
  38. 2011Session-split evaluation of regular and overnight systems
  39. 2012Walk-forward evaluation as operator rehearsal
  40. 2013Two-window evaluation of mechanical trading systems
  41. 2013Walk-forward filter selection for repeated-median velocity
  42. 2014Walk-forward evaluation for fading-memory velocity systems
  43. 2015Test oscillator events before tuning rules
  44. 2016Walk-forward evaluation of a five-parameter parabolic stop-and-reversal
  45. 2016Walk-forward optimization without curve fitting
  46. 2017Optimization without overfitting in trend-system evaluation
  47. 2017Parameter stability is a better guide than a larger crossover grid
  48. 2018Point-in-time universes for system evaluation
  49. 2018Walk-forward robustness evaluation for optimized systems
  50. 2018Critiquing breakout systems through robustness tests
  51. 2018A critique of parameter fitting in system design
By fieldSystem design117 readingsBy year201398 readings
All 58 readings tagged Robustness testing
Also on Robustness testing5 readings
1986Degrees of freedom in trading system optimization1988Testing re-optimization calendars against random parameter controls1988Walk-forward and neighborhood tests after optimization1992Constructing forecast models with regression, walk-forward, and robustness1992Diagnose regimes before you lock parameters