Dynamic mode decomposition for financial trading strategies

We demonstrate the application of an algorithmic trading strategy based upon the recently developed dynamic mode decomposition on portfolios of financial data. The method is capable of characterizing complex dynamical systems, in this case financial market dynamics, in an equation-free manner by decomposing the state of the system into low-rank terms whose temporal coefficients in time are known. By extracting key temporal coherent structures (portfolios) in its sampling window, it provides a regression to a best fit linear dynamical system, allowing for a predictive assessment of the market dynamics and informing an investment strategy. The data-driven analytics capitalizes on stock market patterns, either real or perceived, to inform buy/sell/hold investment decisions. Critical to the method is an associated learning algorithm that optimizes the sampling and prediction windows of the algorithm by discovering trading hot-spots. The underlying mathematical structure of the algorithms is rooted in methods from nonlinear dynamical systems and shows that the decomposition is an effective mathematical tool for data-driven discovery of market patterns.     

Technical analysis: Novel insights on contrarian trading

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.     

Technical trading revisited: False discoveries,persistence tests,and transaction costs

We revisit the apparent historical success of technical trading rules on daily prices of the Dow Jones Industrial Average index from 1897 to 2011, and we use the false discovery rate (FDR) as a new approach to data snooping. The advantage of the FDR over existing methods is that it selects more outperforming rules, which allows diversifying against model uncertainty. Persistence tests show that, even with the more powerful FDR technique, an investor would never have been able to select ex ante the future best-performing rules. Moreover, even in-sample, the performance is completely offset by the introduction of low transaction costs. Overall, our results seriously call into question the economic value of technical trading rules that has been reported for early periods.     

GP algorithm versus hybrid and mixed neural networks

In the current paper, we present an integrated genetic programming (GP) environment called java GP modelling. The java GP modelling environment is an implementation of the steady-state GP algorithm. This algorithm evolves tree-based structures that represent models of inputs and outputs. The motivation of this paper is to compare the GP algorithm with neural network (NN) architectures when applied to the task of forecasting and trading the ASE 20 Greek Index (using autoregressive terms as inputs). This is done by benchmarking the forecasting performance of the GP algorithm and six different autoregressive moving average model (ARMA) NN combination designs representing a Hybrid, Mixed Higher Order Neural Network (HONN), a Hybrid, Mixed Recurrent Neural Network (RNN), a Hybrid, Mixed classic Multilayer Perceptron with some traditional techniques, either statistical such as a an ARMA or technical such as a moving average convergence/divergence model, and a naïve trading strategy. More specifically, the trading performance of all models is investigated in a forecast and trading simulation on ASE 20 time-series closing prices over the period 2001–2008, using the last one and a half years for out-of-sample testing. We use the ASE 20 daily series as many financial institutions are ready to trade at this level, and it is therefore possible to leave orders with a bank for business to be transacted on that basis. As it turns out, the GP model does remarkably well and outperforms all other models in a simple trading simulation exercise. This is also the case when more sophisticated trading strategies using confirmation filters and leverage are applied, as the GP model still produces better results and outperforms all other NN and traditional statistical models in terms of annualized return.     

How does the market variance risk premium vary over time? Evidence from S&P 500 variance swap investment returns

We explore whether the market variance risk premium (VRP) can be predicted. We measure VRP by distinguishing the investment horizon from the variance swap’s maturity. We extract VRP from actual S&P 500 variance swap quotes and we test four classes of predictive models. We find that the best performing model is the one that conditions on trading activity. This relation is also economically significant. Volatility trading strategies which condition on trading activity outperform popular benchmark strategies, even once we consider transaction costs. Our finding implies that broker dealers command a greater VRP to continue holding short positions in index options in the case where trading conditions deteriorate.     

Wikipedia pageviews as investors’ attention indicator for Nasdaq

The attempt to measure investors’ mood to find an early indicator of financial markets has evolved and developed with the advancement of technology over the years. The first attempts were based on surveys, a long and expensive process. Nowadays, big data has made it possible to measure the investor’s mood accurately and almost entirely online. This paper analyzes the explanatory and predictive capacity of Wikipedia pageviews for the Nasdaq index. For this purpose, two econometric models have been developed. In both models, the explanatory variable is the number of Wikipedia visits, and the endogenous variable is Nasdaq index return. As an alternative to this approach, an algorithmic trading system has been developed. It uses Wikipedia visits as investment signals for long and short positions to check the predictability power of this indicator. It is determined that the volume of queries about Nasdaq companies is a statistically significant variable for expressing the evolution of this index. However, it has no predictive capacity. Keeping in mind the capacity of Wikipedia to exemplify Nasdaq trends, further studies should be conducted to determine how to make this indicator profitable.     

The predictability of security returns with simple technical trading rules

Technical traders base their analysis on the premise that the patterns in market prices are assumed to recur in the future, and thus, these patterns can be used for predictive purposes. This paper uses the daily Dow Jones Industrial Average Index from 1897 to 1988 to examine the linear and nonlinear predictability of stock market returns with simple technical trading rules. The nonlinear specification of returns are modelled by single layer feedforward networks. The results indicate strong evidence of nonlinear predictability in the stock market returns by using the past buy and sell signals of the moving average rules.     

Profitability of technical stock trading: Has it moved from daily to intraday data?

This paper investigates how technical trading systems exploit the momentum and reversal effects in the S&P 500 spot and futures market. When based on daily data, the profitability of 2580 technical models has steadily declined since 1960, and has been unprofitable since the early 1990s. However, when based on 30-minutes-data the same models produce an average gross return of 7.2% per year between 1983 and 2007. These results do not change substantially when trading is tested over eight subperiods. In particular, there is no clear trend of a declining profitability of technical stock trading based on 30-minutes-data. Those 25 models which performed best over the most recent subperiod produce a significantly higher gross return over the subsequent subperiod than all models. Between 2001 and 2007 the 2580 models perform worse than over the 1980s and 1990s. This result could be due to stock markets becoming recently more efficient or to stock price trends shifting from 30-minutes-prices to prices of higher frequencies.     

Insolvent trading and voluntary administration in Australia: economic winners and losers?

Australian directors who incur debts while their companies are insolvent can be pursued by the corporate regulator for compensation when their companies fail. Under the Australian insolvent trading laws, directors no longer experience ‘true’ limited liability, and as expected, they adjust their behaviour as a result. Identifying director's rational behaviour in an insolvent trading world is difficult as there are no formal economic models of director decision-making under Australian current corporate law. In this paper, we develop such a model primarily for private companies. We incorporate the threat of insolvent trading as well as director's tactical use of voluntary administration to avoid insolvent trading litigation. We show that neither a combination of insolvent trading or voluntary administration can simultaneously ensure creditors-best outcomes, eliminate insolvent trading and reduce director underinvestment.     

Market Efficiency and Returns to Simple Technical Trading Rules: Further Evidence from U.S., U.K., Asian and Chinese Stock Markets

Tian, Wan and Guo (2002) explored the predictability and profitability of technical trading rules in markets with different efficiency levels; namely, the U.S. and China. In the case of the U.S. they found rules to have no predictability after 1975, whereas their results give support to technical trading rules having both predictability and profitability for the Chinese markets across the 1990's. The purpose of this paper is to extend the analysis of Tian et al. in two ways. First, to see if the conclusions extend to other markets – namely, the U.K., Hong Kong and Japan. Second, in the case of China, to examine whether the predictability and profitability of technical trading rules changed across the 1990's. On the basis of daily data Tian et al's results for the U.S. market are supported by the results for a number of the main developed markets where the technical trading rules had predictive ability during the 1970's that disappeared by the 1990's. Furthermore, the results suggest that while technical trading rules had short term predictive ability and profitability in the Chinese stock markets during the 1990's, this lessened as the decade progressed. JEL Classification: G14, G15     

The intra-day performance of market timing strategies and trading systems based on Japanese candlesticks

Abstract

We develop market timing strategies and trading systems to test the intra-day predictive power of Japanese candlesticks at the 5-minute interval on the 30 constituents of the DJIA index. Around a third of the candlestick rules outperform the buy-and-hold strategy at the conservative Bonferroni level. After adjusting for trading costs, however, just a few rules remain profitable. When we correct for data snooping by applying the SSPA test on double-or-out market timing strategies, no single candlestick rule beats the buy-and-hold strategy after transaction costs. We also design fully automated trading systems by combining the best-performing candlestick rules. No evidence of out-performance is found after transaction costs. Although Japanese candlesticks can somewhat predict intra-day returns on large US caps, we show that such predictive power is too limited for active portfolio management to outperform the buy-and-hold strategy when luck, risk, and trading costs are correctly measured.     

Time-varying neural network for stock return prediction

We consider the problem of neural network training in a time-varying context. Machine learning algorithms have excelled in problems that do not change over time. However, problems encountered in financial markets are often time varying. We propose the online early stopping algorithm and show that a neural network trained using this algorithm can track a function changing with unknown dynamics. We compare the proposed algorithm to current approaches on predicting monthly US stock returns and show its superiority. We also show that prominent factors (such as the size and momentum effects) and industry indicators exhibit time-varying predictive power on stock returns. We find that during market distress, industry indicators experience an increase in importance at the expense of firm level features. This indicates that industries play a role in explaining stock returns during periods of heightened risk.     

Insider Trading and the Bid-Ask Spread

This study examines the intertemporal and cross-sectional association between the bid-ask spread and insider trading. Empirical results from the cross-sectional regression analysis reveal that market makers establish larger spreads for stocks with a greater extent of insider trading. The time-series regression analysis, however, finds no evidence of spread changes on insider trading days. These results suggest that although market makers may not be able to detect insider trading when it occurs, they protect themselves by maintaining larger spreads for stocks with a greater tendency of insider trading. The results also reveal that market makers establish larger spreads when there are unusually large transactions. In addition, this study finds that spreads are positively associated with risk and negatively with trading volume, the number of exchange listings, share price, and firm size.     

从光大“乌龙指”事件看高频交易监管

随着电子化交易的发展,高频交易在证券市场中的运用越来越多。高频交易在增加流动性的同时,也因技术故障等问题造成证券市场价格剧烈动荡,这引发了人们的争论,即是否需要对高频交易进行限制。高频交易虽然带来了流动性,但是技术故障引发市场波动,量化交易的助涨助跌,更是会引发崩盘的系统性风险。因此,限制高频交易具有正当性。光大"乌龙指"事件,是我国首例因为技术故障引发的极端事件,应该借此机会反思,建立熔断机制、异常交易的处理规则和法律制度。     

Entry, exit and trading profits: A look at the trading strategies of a proprietary trading team

We examine the behavior of a 15 strong proprietary stock trading team and show how consistent intraday trading profits were generated. The team, who worked for a large US direct access trading firm, executed over 96 thousand trades in 3 months in 2000. Profitable intraday trading occurred in an anonymous dealer capacity, on both long and short positions, especially when volume and price volatility were higher. The traders rapidly entered long (short) positions when the number of dealers and size become greater on the bid (offer) side of the spread. Profits were taken early against the trend.     

Profitability of alternative methods of combining the signals from technical trading systems

Past efforts determining the profitability of technical analysis reached varied conclusions. We test the profitability of a composite prediction that uses buy and sell signals from technical indicators as inputs. Both machine learning methods, like neural networks, and statistical methods, like logistic regression, are used to get predictions. Inputs are signals from trend‐following and mean‐reversal technical indicators in addition to the variance of prices. Four representative commodities from agricultural, livestock, financial, and foreign exchange futures markets are selected to determine profitability. Special care is taken to avoid data snooping error. Both neural networks and statistical methods did not show consistent profitability.     

How do individual investors trade?

This paper examines how high-frequency trading decisions of individual investors are influenced by past price changes. Specifically, we address the question as to whether decisions to open or close a position are different when investors already hold a position compared with when they do not. Based on a unique data set from an electronic foreign exchange trading platform, OANDA FXTrade, we find that investors’ future order flow is (significantly) driven by past price movements and that these predictive patterns last up to several hours. This observation clearly shows that for high-frequency trading, investors rely on previous price movements in making future investment decisions. We provide clear evidence that market and limit orders flows are much more predictable if those orders are submitted to close an existing position than if they are used to open one. We interpret this finding as evidence for the existence of a monitoring effect, which has implications for theoretical market microstructure models and behavioral finance phenomena, such as the endowment effect.     

Robust and adaptive algorithms for online portfolio selection

We present an online approach to portfolio selection. The motivation is within the context of algorithmic trading, which demands fast and recursive updates of portfolio allocations as new data arrives. In particular, we look at two online algorithms: Robust-Exponentially Weighted Least Squares (R-EWRLS) and a regularized Online minimum Variance algorithm (O-VAR). Our methods use simple ideas from signal processing and statistics, which are sometimes overlooked in the empirical financial literature. The two approaches are evaluated against benchmark allocation techniques using four real data sets. Our methods outperform the benchmark allocation techniques in these data sets in terms of both computational demand and financial performance.     

Fuzzy logic,trading uncertainty and technical trading

From the market microstructure perspective, technical analysis can be profitable when informed traders make systematic mistakes or when uninformed traders have predictable impacts on price. However, chartists face a considerable degree of trading uncertainty because technical indicators such as moving averages are essentially imperfect filters with a nonzero phase shift. Consequently, technical trading may result in erroneous trading recommendations and substantial losses. This paper presents an uncertainty reduction approach based on fuzzy logic that addresses two problems related to the uncertainty embedded in technical trading strategies: market timing and order size. The results of our high-frequency exercises show that ‘fuzzy technical indicators’ dominate standard moving average technical indicators and filter rules for the Euro-US dollar (EUR-USD) exchange rates, especially on high-volatility days.