Trendlines and channels
A trendline connects consecutive lows or highs. It's useful for adding structure, but easy to manipulate because of its degrees of freedom.
Learning objective: After this lesson, you can draw a trendline and assess its limits as a tool.
You connect two lows with a line and extend it. If the price touches it again later and turns, the line counts as confirmed. Two parallel lines form a channel.
That's a useful tool for bringing structure to a picture. It has one problem worth knowing: you decide which points to connect.
With twenty candles, there are a great many possible lines. Search long enough, and you'll always find one that looks good. That's why chart analysis looks more convincing in hindsight than in real time.
A simple test helps: draw the line, then cover the right half of the chart. Would you still draw it that way? If not, you've built the outcome into the line.
A trendline is a linear approximation of selected extreme points. Since both the choice of points and the scaling are free, the number of plausible lines is large. With n extreme points, there are n choose 2 possible connections, and in hindsight the ones chosen tend to be those matching later turning points. That selection with knowledge of the outcome is the definition of overfitting.
Scale is an additional degree of freedom. A line straight on a linear scale is curved on a logarithmic one, and vice versa. Over long periods with substantial appreciation, the two displays lead to markedly different conclusions, with neither justified by the method itself.
For testable use, every parameter has to be fixed in advance: definition of an extreme point, minimum distance, number of touches, tolerance band, scale, and definition of a break. Only with these fixed does an approach become testable against historical data. Without them, it's a form of display, not a method.
Summary
- You pick the points, and that's the whole weakness.
- Linear and logarithmic scales produce different lines.
- Without predefined parameters, it's display, not method.
Did you get it?
Why does chart analysis look more convincing in hindsight?
Because the lines get chosen with knowledge of the outcome. That's overfitting.
What role does scaling play?
A line straight on a linear scale is curved on a logarithmic one. The two displays lead to different conclusions.
What turns a display into a testable method?
Predefined parameters for extreme points, touches, tolerance, scale, and break definition.
Check your understanding
Related
- Support and resistanceStage 3
- Fundamental vs. technical analysisStage 3
- Reading a chartStage 3
- Matching questions for this stageQuestions