Methods for Discovering and Testing Historical Chart Patterns
Article Quant Q&A · Author: CodeCamper
Summary
The discussion considers ways to search historical stock charts for user-defined or familiar patterns. Suggested approaches include time-series matching, dynamic time warping, classical technical pattern recognition, and unsupervised discovery through hierarchical clustering. It points to statistical testing as a way to assess whether a pattern carries predictive information, rather than relying on visual appeal or received wisdom.
The responses mainly recommend research methods and external resources; they do not provide a tested scanner, implementation details, or evidence that any named pattern predicts returns. Several suggestions are explicitly unverified, and one respondent questions whether a clustering study would survive bootstrap testing. Pattern searches also raise a broader risk of data-mining bias when many shapes or parameter choices are examined. Any discovered signal therefore needs careful out-of-sample evaluation before it can inform a trading strategy.
Key ideas
- Historical pattern searches can use time-series matching or dynamic time warping.
- Hierarchical clustering offers one route to unsupervised pattern discovery.
- Statistical testing can help distinguish predictive patterns from visual folklore.
- The suggested tools and studies are not validated by evidence in the discussion.
- Searching many shapes can create data-mining bias, so discovered signals need robust validation.
Tags
Full text
# Scan for chart patterns software
# Scan for chart patterns software
Is there any software out there that currently would allow you to scan historical charts and look for a specific pattern and then show you that pattern for a list of stocks. Google finance and yahoo finance have all the charting data needed to visually detect certain patterns, but I was wondering if there was software out that I could define a particular pattern and then it would show me a bunch of real world examples of that pattern by scanning a dump of symbols. Does any such software exist or something to do something similar to this? I actually have a custom pattern I want to scan (that or I do not yet know the name of it) Not really sure how to word this but I am thinking of having it scan the recent behavior of a stock and then to scan a list of symbols for that pattern but historically back in time, not the present.
## Answer by vonjd (score 7)
https://quant.stackexchange.com/a/16059
I know of two sites that promise to do that but I haven't used either of them so I cannot tell you how reliable they are:
- http://thepatternsite.com/chartpatterns.html
- http://www.priceactionlab.com/Manual/manual.html
The excellent blog systematic investor has some articles about how to do it in R:
- http://systematicinvestor.wordpress.com/2012/01/13/time-series-matching/
- http://systematicinvestor.wordpress.com/2012/01/17/time-series-matching-strategy-backtest/
- http://systematicinvestor.wordpress.com/2012/01/20/time-series-matching-with-dynamic-time-warping/
- http://systematicinvestor.wordpress.com/2012/05/22/classical-technical-patterns/
In general this is not a trivial problem. If you are interesting in some theoretical background the following papers may be for you:
- http://www.cse.ust.hk/~leichen/courses/comp630p/collection/reference-4-6.pdf
- http://cs.stanford.edu/people/saswat/research/one.pdf
For a much broader approach you might want to check out the book Evidence Based Technical Analysis by David Aronson.
In it he applies statistical techniques to determine whether certain time series patterns have any predictive power. It's an interesting read and should equip you with some ideas on how to differentiate between folklore and statistical rigor. It also gives you ample literature references.
You can find a good overview and summary of the book on CXO Advisory.
You can also find further material on the webpage of the author.
The Automated Trading System-blog has a few posts about the book with a helpful tool at the end:
- http://www.automated-trading-system.com/evidence-based-technical-analysis-aronson-book/
- http://www.automated-trading-system.com/bootstrap-test/
and here the conclusion with the free tool for the bootstrap-test:
- http://www.automated-trading-system.com/bootstrap-take-2-data-mining-bias-code-and-using-geometric-mean/
## Answer by EllenJ (score 2)
https://quant.stackexchange.com/a/16080
You can try Macroaxis. Although it does not let you define your custom patterns but it can plot and recognize more than 300 technical indicator patterns on historical data. You can see example of pattern recognition for google at http://www.macroaxis.com/invest/Pattern-Recognition/GOOG
Disclaimer: I work for Macroaxis
## Answer by Procoder (score 1)
https://quant.stackexchange.com/a/23294
Stock Analyser is a Stock Charting, Portfolio Optimization, Pairs Trade, Technical and Quantitative Analyzer. It is a free software. I am coding it.
Stock Analyst
## Answer by Xpector (score 1)
https://quant.stackexchange.com/a/37754
A relatively easy reading on unsupervised pattern discovery with hierarchical clustering ("bottom-up clustering") is by Karsten Martiny. I'm not sure if his results would stand a bootstrap test, nor I take the choice of Dow Jones daily OHLC from 1932 to 2011 as a dataset too seriously.
You asked for a software. That would be e.g. R with a package for hierarchical clustering.
## Answer by Mia Austin (score 0)
https://quant.stackexchange.com/a/80898
I am sharing a link with you where you can find chart patterns and index historical data. I hope this will be helpful to you.
https://www.stockaio.com/UK/ai/Screener/ChartPatterns/PopularChartPatternsShown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.