Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation
Read the paperopens doi.org in a new tab
What they found
The authors used kernel smoothing to automatically detect ten classic chart patterns (head-and-shoulders, double tops and bottoms, triangles, rectangles, and broadening formations) in daily prices of U.S. stocks from 1962 to 1996. They then asked whether the distribution of returns after a pattern differed from the unconditional distribution. For several patterns it did, especially in Nasdaq stocks, meaning the patterns carry some information. They stopped short of claiming profitability, framing the result as 'patterns are not noise' rather than 'patterns make money'.
What you can use
- Chart patterns can be defined objectively and tested; when tested, some do contain statistical information.
- Information is not profit: the paper never shows a trading rule based on the patterns beating costs.
- Effects were stronger in smaller Nasdaq stocks than large NYSE names, consistent with patterns reflecting slow information flow.
Caveats
Mathematically dense. Tests are about return distributions, not about trading profits, and the patterns' economic magnitudes are small. A free NBER working paper version exists.
Tags: technical-analysis, chart-patterns, pattern-recognition
Summaries are our own reading of the paper, not the authors' words. Educational only, not advice. Discuss it in Book Club.