What makes trading strategies based on chart pattern recognition profitable?

Oct 13, 2026

Automating chart pattern recognition supports technical-analysis trading systems by enabling signal generation from generic patterns shown to be profitable in prior history. The paper designs a trading framework that employs a fast dynamic time warping subsequence search approach (UCR suite) to perform pattern recognition within realistic timescales. Profitability is evaluated using a mean-variance perspective over 560 NYSE stocks while controlling for data-snooping and transaction costs. Results show the proposed system dominates the market index on average, and 92.5% of experiments remain profitable when aligned parameters reflect technical analysis.

LanguageEnglish

CategoryPresentations

FormatPDF

Pages19

Reading Time7 min

AccessFree to download

Summary

Keywords

1 Introduction

Document downloaded from: [http://hdl.handle.net/10251/185285](http://hdl.handle.net/10251/185285) [This paper must be cited as:](This paper must be cited as:) Tsinaslanidis, P. ; Guijarro, F. (2021) . What makes trading strategies based on chart pattern recognition profitable?. Expert Systems. 38(5):1-17 . [https://doi.org/10.1111/exsy.12596](https://doi.org/10.1111/exsy.12596) The final publication is available at [https://doi.org/10.1111/exsy.12596](https://doi.org/10.1111/exsy.12596) Copyright Blackwell Publishing Additional Information DOI: xxx/xxxx SPECIAL ISSUE PAPER What makes trading strategies based on chart pattern recognition proﬁtable? Prodromos Tsinaslanidis 1 | Francisco Guijarro2 1 Department of Economics, University of Western Macedonia, Kastoria Campus, Greece 2 Research Institute for Pure and Applied Mathematics, Universitat Politècnica de València, Valencia, Spain Correspondence Francisco Guijarro, Universitat Politècnica de València, Cami de Vera s/n, 46022 Valencia, Spain. Email: [fraguima@upvnet.upv.es](fraguima@upvnet.upv.es) Summary Automating chart pattern recognition is a relevant issue addressed by researchers and practitioners when designing a system that considers technical analysis for trading purposes. This paper proposes the design of a trading system that takes into account any generic pattern that has been proven to be proﬁtable in the past, without restricting the search to the speciﬁc technical patterns reported in the literature, hence the term generic pattern recognition. A fast version of dynamic time warping, the University College Riverside subsequence search suite (called the UCR suite), is employed for the pattern recognition task in an eﬀort to produce trading signals in realistic timescales. This paper evaluates the signiﬁcance of the relation between the system's proﬁtability and (i) the pattern length,(ii) the take-proﬁt and stop-loss levels and (iii) the performance consensus of past patterns. The trading system is assessed under the mean-variance perspective by using 560 NYSE stocks. The results obtained by the diﬀerent parameter conﬁgurations are reported, controlling for both data-snooping and transaction costs. On average the proposed system dominates the market index in the mean-variance sense. Although transaction costs reduce the proﬁtability of the proposed trading system, 92.5% of the experiments are proﬁtable if the analysis is reduced to the parameter values aligned with the technical analysis. KEYWORDS: Stock markets, technical analysis, dynamic time warping, UCR suite, generic pattern recognition 1 INTRODUCTION Automating trading decisions, submission and management of trading orders through the use of computer algorithms is a common deﬁnition of algorithmic trading (Hendershott, Jones, & Menkveld 2011) . Many algorithmic trading systems have been proposed, particularly aimed at achieving the automation of trading decisions (e.g. Feuerriegel & Prendinger 2016; Geva & Zahavi 2014; Leigh, Modani, & Hightower 2004; Leigh, Purvis,& Ragusa 2002) . The trading rules embedded in these systems generate trading signals, such as long, short and out-of-the-market (Baía & Torgo 2017), by considering various inputs, including market sentiment and technical factors (e.g. Feuerriegel & Prendinger 2016; Nguyen, Shirai, & Velcin 2015) . Regarding the latter, a rich assortment of technical tools can be found in the voluminous literature on Technical Analysis (TA) . The celebrated classes of these TA tools are technical indicators (Kim & Han 2001; Metghalchi, Marcucci, & Chang 2012; Pätäri & Vilska 2014), candlesticks formations (Lu & Shiu 2012 2016) and chart patterns (Tsinaslanidis 2018; Zapranis & Tsinaslanidis 2012a) . However, considering chart patterns in an automated trading system is a challenging task. Although technical indicators such as moving averages, MACD and RSI are easily expressed in mathematical form, the identiﬁcation an

FAQs

What problem does the paper address in technical trading systems?

It addresses how to automate chart pattern recognition for trading purposes, using a generic approach rather than restricting search to a fixed set of literature patterns.

Which method and tool are used for pattern recognition?

The system uses a fast version of dynamic time warping, implemented via the University College Riverside subsequence search suite (the UCR suite).

How is the trading system evaluated and what do the results indicate?

Evaluation follows a mean-variance perspective on 560 NYSE stocks with controls for data-snooping and transaction costs; on average the system dominates the market index, though transaction costs reduce profitability.

Oct 13, 2026

Upload by WPS_1786070896