Technical analysis: Novel insights on contrarian trading

• Published date: 01 September 2023
• Author: Patrick Eugster,Matthias W. Uhl
• Date: 01 September 2023
• DOI: http://doi.org/10.1111/eufm.12389

DOI: 10.1111/eufm.12389

ORIGINAL ARTICLE

Technical analysis: Novel insights on

contrarian trading

Patrick Eugster |Matthias W. Uhl

Department of Banking and Finance,

University of Zurich, Zurich, Switzerland

Correspondence

Patrick Eugster, Department of Banking

and Finance, University of Zurich,

Zurich, Switzerland.

Email: patrick@patrick-eugster.ch

Abstract

We analyze the predictive power of technical analysis

with a novel data set based on news sentiment that

allows to systematically examine a set of technical

analysis indicators over an extensive time period. We

do not find much statistically significant relationships

with the examined indicators and future asset returns,

and we almost do not find any alphas in trading

strategies based on technical analysis sentiment. We

find evidence for a contrarian‐based hypothesis: past

market returns and technical analysis sentiment are

able to predict future technical analysis sentiment with

a negative relationship.

KEYWORDS

behavioural finance, news sentiment, NLP, technical analysis

JEL CLASSIFICATION

G11, G40

1|INTRODUCTION

‘Trade your way to financial freedom using technical analysis’. This is a common advertisement

that one can regularly read on various investment blogs, trading platforms, or the like. Many

retail investors are attracted to such bold phrases with big promises. The field of technical

Eur Financ Manag. 2023;29:1160–1190.1160

|

wileyonlinelibrary.com/journal/eufm

EUROPEAN

FINANCIAL MANAGEMENT

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and

reproduction in any medium, provided the original work is properly cited.

© 2022 The Authors. European Financial Management published by John Wiley & Sons Ltd.

We are grateful to John Doukas, four anonymous referees, and participants at the Brown Bag Lunch Seminar at the

Department of Banking and Finance at the University of Zurich for their valuable feedback.

analysis is widely known within the investment community and has been examined in great

detail in the academic literature. However, the different fields that are considered to be part of

technical analysis are plentiful, which makes it difficult to examine what works and what does

not work on a broader scale, as this has been heavily debated in the past decades. With this

study, we aim to examine whether technical analysis in general is predictive of stock returns or

vice versa based on a new data set.

The performance of technical analysis is sometimes difficult to assess, as it encompasses a

wide range of techniques. Indicators, chart formations, candlestick patterns, or Elliott Waves,

among many others, are applied to different asset classes by numerous individuals around the

world. This whole range of converging and diverging ideas, opinions, and techniques subsume

under the umbrella of technical analysis, making it challenging to scientifically examine its

significance. The difficulties are further enhanced as a generally accepted definition of a certain

chart formation does not exist.

This study circumvents these problems by using a novel data set based on an extensive news

sentiment data set from Refinitiv, known as Refinitiv News Analytics (TRNA, formerly called

Thomson Reuters News Analytics). The data set automatically collects and scores news items

across different dimensions of technical analysis and thereby allows to outsource the need to

precisely define a support level, an inverted head & shoulders formation, or a third wave of a five‐

wave Elliott Wave pattern. The advantage of such an approach is that there is a direct transfer

mechanism that the sentiment algorithm can extract the suggested direction of trading based on

technical analysis articles while topic tagging them for the respective technical analysis technique.

Forexample,ifanarticleiswrittenaboutabearishhead‐and‐shoulders formation, which should

lead to falling stock prices, the sentiment algorithm will identify negative sentiment based on

keywords like ‘bearish’and ‘falling stocks’, and in combination with the respective topic tagging.

Furthermore, such an approach ensures that all kinds of technical analysis techniques are included

in the analysis that are written or ‘in focus’to investors.

1

This also ensures that changing

techniques of chartism are included in the data set over time.

We test the validity of technical analysis indicators by first examining cross‐correlations

and, second, by estimating linear regressions to test the predictability on stock returns of the

data. Additionally, simple trading strategies are constructed based on technical analysis based

signals. Applying around 200,000 news items and their sentiment from a large universe of

global news with over 10 million news items, a global equity technical analysis based sentiment

strategy cannot systematically generate alpha over the 17 years of historical data considered.

However, we identify few strategies that deliver positive information ratios, such as based on

crude oil. Our results are robust across different regions, technical terms, individual assets,

look‐back periods, and other strategy parameters. We further test whether stock returns can

predict technical analysis signals. We find statistically significant results: stock market returns

are able to predict technical analysis sentiment with a positive relationship in the short‐run and

an inverse relationship on a monthly time frame. This leads to the hypothesis that most

technical analysts look at past stock market returns and formulate momentum or contrarian

strategies depending on their investment horizon. We additionally find an even stronger

negative relationship between past and present technical analysis sentiment. This lends support

to the hypothesis that technical analysts act in a contrarian fashion and are not able to generate

alpha in most asset classes.

1

At least based on a journalist's perception, which is admittedly one of the constraints of such an approach.

EUGSTER AND UHL EUROPEAN

FINANCIAL MANAGEMENT

|

1161