The Technical Analysis Challenge

The sections of this page are:
Introduction
Why the Challenge?
Results
Plots and Data


Introduction

This was a scientific study aimed at testing the efficacy of technical analysis. The task was to select the real price series out of a set that included three random series when given the preceding two years of daily prices. Submissions were accepted from the 6th of September 2003 through the 4th of October 2003.

Why the Challenge?

The management of funds affects the future livelihood of virtually everyone. The finance industry should be using the best methodologies available, and should have a scientific reason, if possible, for believing in what is done. This test is a step in that direction -- to help prove or disprove the usefulness of technical analysis procedures.

Results

The report on the results of the study are in the Working Paper entitled
The Technical Analysis Challenge (pdf).

Abstract: We report on a study of the ability of analysts to distinguish an actual price series of an equity from random alternatives. Virtually all of the statistical tests on the results support the hypothesis that no skill was exhibited in selecting the correct response. Many of the analysts were extremely over-confident about their ability to select correct answers. The one area where it seems skill might have been exhibited is in the selection of correct answers that happened to be far from the random choices.

Also available:
A tab-separated file containing the submissions that were received: submissions.txt
A comma-separated file with the correct answers and the identity of each of the series: seriesinfo.csv

Plots and Data

Plots of the price series and the actual data are both available.

The pdf of the series -- that is, one hundred plots of 500 prices each.

The pdf of the extensions -- one hundred pages with four plots per page with 50 prices in each plot.

The pdf of all the data -- one hundred pages with four plots per page with 550 prices in each plot.


File earlydata.csv contains the 100 series of 500 prices each (series are in columns). The values are separated by commas.

File latedata.csv contains the 400 extensions of 50 prices each (extensions are in rows). The values are separated by commas. To be consistent with earlydata.csv, this should have the extensions in the columns, but Excel only reads in 256 columns of data.

A good environment in which to work with such data is R. Go here for more details on R (which is free), and for some R code to manipulate the price data and produce plots of it.

Go to Burns Statistics Home.

Last Modified: 2003 October 07