1990issue C021-17
Bezier-curve price trend filter
A Bezier path is built from a few control vertices of sampling distance and price, with the newest observation at the far end of a fixed lookback. The finished curve is a weighted blend of those vertices, and a rudimentary trend-filter compares the current close with one chosen segment-value.
- Control vertices are ordered pairs of sampling distance and price, so the newest observation sits at the far end of a fixed lookback.
- Each point on the finished curve is a weighted blend of all control vertices, and the path begins bending toward later vertices immediately.
- Using every price as a control vertex collapses the spline into a connect-the-dots plot and fails to filter noise.
- The chosen segment index sets responsiveness: early segments lag recent prices and late segments chase them.
Control vertices along a fixed lookback
A bezier-curve is a cubic spline whose path is determined by a small set of control vertices and cannot cross itself. Each control-vertex is an ordered pair of sampling distance and price that pins the spline. The newest observation sits at the far end of a fixed lookback, and earlier prices sit at earlier distances.
The demonstration construction used seven vertices sampled at 28, 23, 18, 13, 8, 5 and 0 periods back, with the newest price as the last control point.
Weighted blending and output resolution
Each point on the finished curve is a weighted blend of all control vertices. The path begins bending toward later vertices immediately and is pulled off a straight line between the first two points.
Output resolution is a separate parameter. The demonstration spline was assembled from 30 line segments. Finer segmentation increases computation without changing the vertex set.
Using every price as a control vertex collapses the spline into a connect-the-dots plot. Too many vertices therefore fail to filter noise, because more vertices make the curve hug raw prices more closely.
Comparing the close with one segment-value
A trend-filter damps noisy price movement and compares the latest close with a smoothed baseline at a defined lookback. The archive workflow uses a rudimentary trading filter that compares the current close with one chosen segment. That segment-value is the comparison level for a long or short bias. A close above the segment is treated as an upside breakout-bias, and a close below it as a downside breakout-bias.
Early segments lag recent prices and late segments chase them, so the chosen segment index sets responsiveness in the same way a moving-average length does.
The same spline construction was applied to daily series and to half-hour series, with the close compared against segment indices 2, 7, 12, 17 and 22.
All readings on this track · 33 readings
- 1988Opening-range brackets, a two-bar trend filter, and bounded stops
- 1990Bezier-curve price trend filter
- 1992Constructing a damping-index trend filter
- 1992Building a random walk index trend filter
- 1992Phase diagrams for moving-average trend filters
- 1993Volume-weighted change smoothing and trend ranking
- 1993Concurrent highest-low filter with a largest-low-fall trigger
- 1994Unit-invariant trend filters and the c-test
- 1995Constructing cup and cap entries with a three-bar net line
- 1997Why a daily timing evaluation depends on interval, lookbacks, and the fitting objective
- 2001A volume budget clock for trend-segment construction
- 2001Keep three jobs separate when you test a composite score
- 2002Evaluating the weekly four-percent close filter as a market-state procedure
- 2003Constructing a confirmed zigzag trend filter
- 2004Decompose high, low, and close into separate forecast streams
- 2005Three-state moving-average breakout bar coloring
- 2005Constructing a volume and move-adjusted trend filter
- 2005A fifty-day average breakout as a trend permission filter
- 2005Current-bar inclusion can mute a stochastic channel break
- 2006A stochastic oscillator gated by a long-term exponential average
- 2010A construction test for a modified volume-price trend filter
- 2011Constructing a Spearman rank trend filter
- 2013Constructing a repeated-median slope as a resistant trend filter
- 2014Combining a relative-strength index and trend filters for oversold setups
- 2014Price-rooted lookbacks for a relative strength index, a moving average, and a trend filter
- 2015Evaluating next-session intermarket range forecasts
- 2018Read the intermarket weight matrix first, then the predicted moving-average filter
- 2018Constructing the stiffness trend filter from moving-average holds
- 2018The averaging kernel and the lagged trend gate are separate specifications
- 2019A trend filter is not ready to compare until portfolio constraints are written down
- 2019Lookback, threshold, and position-capacity for a stiffness trend-filter
- 2020Combining a trend filter with a moving average and a stochastic oscillator
- 2020Constructing a relative-strength oscillator with a rank-agreement trend filter