“Taps”: A trading approach based on deterministic sign patterns☆
Introduction
We introduce a new approach for trading financial instruments that is based on a collection of deterministic sign signals, the collection of all possible such trading signs that can be constructed from an elementary sample space consideration. We develop the underlying theory based on the intuitive observation that all possible paths of buy and sell trading signals, as represented by −1 and +1, belong to an elementary sample space whose properties are well understood and easily manipulated. Using a minimum set of assumptions about the structure of these deterministic trading signals we show that their collection constitutes a zero-cost strategy – if one was willing to undertake it. Then, we proceed to consider a rotation among all possible trading signal paths based on some performance criteria. We explore the properties of this methodology via an extensive simulation experiment and then follow-through with an in depth examination of the empirical performance of the strategies using six real world series, the US-based S&P500 index, two Chinese-based indices, the SSE50 and CSI300, and three exchange trade funds (ETF), one each for Japan, Germany and France – we attempt thus to cover different instruments and different markets. By the nature of the proposed strategy, that operates on a single series at a time, we compare performance metrics with the standard buy & hold benchmark, which is the natural benchmark for comparisons at this stage of our research.
Our results are highly supportive of our theoretical arguments and the underlying structure of the signals, and show remarkable consistency on the performance enhancements that are provided by the strategies we propose – such enhancements appear both in the simulations and the real-world series. Furthermore, we explore the link between performance and entropy of the associated markets and find that the suggested methodology works far better in the two Chinese indices than the US index and, therefore, we attempt to offer an entropy-based explanation – although further research is required to fully explore the impact of entropy on this kind of strategy. In general, we can claim that there is at least one rotation that will provide a significantly higher total return than the benchmark and that this total return will be complimented with a lower maximum drawdown than the benchmark as well. What makes our approach utilitarian and useful to consider in applications are the following: (a) it has straightforward probabilistic foundations, (b) it requires nothing more than tracking some performance measures (which all strategies do anyhow) with trading signals being based on all possible sign paths, (c) it is computationally extremely expedient to backtest, (d) it has the natural buy & hold benchmark as a yardstick for performance and (e) is easily scalable to multiple assets at once and can thus lead to portfolios of strategies - although we have not pursued this last part in this current research.
The rest of the paper is structured as follows. In Section 2 we provide a brief literature review on certain papers that are relevant to our work. In Section 3 we present our methodology and within that section we provide the theoretical foundations for our approach, some empirical motivation and the results from our simulations. In Section 4 we have the description of our data along with associated statistics while in Section 5 we discuss extensively our empirical results. Section 6 offers some concluding remarks and directions for future research.
Section snippets
Literature review
Our work clearly relates to algorithmic trading (AT), and quantitative trading in general, that has increased sharply over the past decade, as more computing power was made easily and cheaply available. Hedge funds and the proprietary trading desks of investment banks have been raising capital devoted to such trading schemes and the corresponding impact of algorithmic trading, in addition to a growing body of related literature, has grown considerably and is still growing. The review that
Theoretical preliminaries
Let be the logarithm of the price of a financial asset, be the corresponding log-return, be the sign of the return and be the sign of the trading direction, the trading sign3. Our object of analysis rests with the series of trading signs . We
Data and descriptive statistics
The first Chinese index we use is the CSI300 index. It is composed of 300 stocks with the largest market capitalization and liquidity from the entire universe of listed A share companies in China. Launched on April 8, 2005, the index aims to measure the overall performance of the A shares traded on Shanghai Stock Exchange and Shenzhen Stock Exchange. The second Chinese index we use is the SSE50 index. This one consists of 50 Shanghai stocks with large market-cap and good liquidity. The index is
Empirical results
We begin the discussion of our empirical results by a quick look at the average winning ratios, as in the simulations, in Table 7. There we have the percentage of times that any taps-based strategies beat the corresponding buy & hold benchmark in terms of higher total return or lower maximum drawdown. The results tally well with our discussion on the descriptive statistics section and the simulation section. First, we note that for the S&P500 the results are rather disappointing in terms of the
Conclusions
We have introduced a new methodology for trading financial instruments based on deterministic sign sequences that are obtained from the elementary m-dimensional sample space of , the two possible signs of selling and buying a financial instrument. There are a number of interesting properties on the corresponding trading strategies, which we termed “taps” strategies, including the zero cost characteristic that we one can derive from them and the methods for which such taps can be used in
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (35)
- et al.
Gated bayesian networks for algorithmic trading
International Journal of Approximate Reasoning
(2016) - et al.
Robust technical trading strategies using gp for algorithmic portfolio selection
Expert Systems with Applications
(2016) - et al.
A stock market trading system based on foreign and domestic information
Expert Systems with Applications
(2019) - et al.
Stock market trading rule based on pattern recognition and technical analysis: Forecasting the djia index with intraday data
Expert Systems with Applications
(2015) - et al.
An intelligent pattern recognition model for supporting investment decisions in stock market
Information Sciences
(2016) - et al.
An intelligent short term stock trading fuzzy system for assisting investors in portfolio management
Expert Systems with Applications
(2016) - et al.
The price dynamics of common trading strategies
Journal of Economic Behavior Organization
(2002) - et al.
No trade and yes trade theorems for heterogeneous priors
Journal of Economic Theory
(2019) - et al.
Low-latency trading
Journal of Financial Markets
(2013) - et al.
Signaling in deterministic and stochastic settings
Journal of Economic Behavior Organization
(2012)
Evolving trading strategies using directional changes
Expert Systems with Applications
Developing a rule change trading system for the futures market using rough set analysis
Expert Systems with Applications
Candlestick technical trading strategies: Can they create value for investors?
Journal of Banking Finance
The role of algorithmic trading in stock liquidity and commonality in electronic limit order markets
Pacific-Basin Finance Journal
Portfolio modeling for an algorithmic trading based on control theory
IFAC-PapersOnLine
Experimental analysis of an online trading algorithm
Electronic Notes in Discrete Mathematics
Algorithmic financial trading with deep convolutional neural networks: Time series to image conversion approach
Applied Soft Computing
Cited by (0)
- ☆
Any possible errors appearing on the paper are ours. Computations were performed by the authors in Python and additional, detailed results that support our arguments based on the components of the SSE50 are available on request. We would like to thank the Editor-in-Chief, the Associate Editor and three anonymous reviewers for providing us with a multitude of useful comments that improved the content and quality of our paper.