Quantpedia Premium Update – 22nd August

New strategies:

#903 – Technical Sentiment Index in Cryptocurrencies

Period of rebalancing: Weekly
Markets traded: cryptos
Instruments used for trading: cryptos
Complexity: Very complex strategy
Backtest period: 2017 – 2021
Indicative performance: 37.02%
Estimated volatility: 26.67%

Source paper:

Viswanath-Natraj, G. and Nguyen, My T. and Filippou, I.: Quantifying Narratives and their Impact on Financial Markets
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4402365
Abstract:
This paper investigates the cross-sectional predictive ability of text-based factors in the cryptocurrency market –an important asset class for retail and institutional investors. We employ Bidirectional Encoder Representations from Transformers (BERT) topic modeling to analyze news articles discussing the top 43 cryptocurrencies by market capitalization. We build text-based factors related to fundamentals and technical trading. We find that pessimism about technical news is positively priced in the cross-section of cryptocurrency returns, while pessimism about fundamental news is negatively priced. These factors provide information over and above existing factor models. Our results demonstrate the importance of considering text-based factors when analyzing cryptocurrency returns.

#904 – Fundamental Sentiment Index in Cryptocurrencies

Period of rebalancing: Weekly
Markets traded: cryptos
Instruments used for trading: cryptos
Complexity: Very complex strategy
Backtest period: 2017-2021
Indicative performance: 37.2%
Estimated volatility: 30.49%

Source paper:

#905 – Exploring the Factor ZOO with a Machine-Learning Portfolio

Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 1998-2016
Indicative performance: 30%
Estimated volatility: 12%

Source paper:

Sak, H. and Chang, M. T., and Huang, T.: Exploring the Factor Zoo With A Machine-Learning Portfolio
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4418633
Abstract:
Over the years, top journals have published hundreds of characteristics to explain stock return, but many have lost significance. What fundamentally affects the time-varying significance of characteristics that survive? We combine machine-learning (ML) and portfolio analysis to uncover patterns in significant characteristics. From out-of-sample portfolio analysis, we back out important characteristics that ML models uncover. The ML portfolio’s exposure alternates between investor arbitrage constraint and firm financial constraint characteristics, the timing of which aligns with credit contraction and expansion states. We explain and show how the credit cycle affects different characteristics’ ability to explain cross-sectional stock return over time.

#906 – Using ChatGPT to Forecast Stock Price Movements

Period of rebalancing: Daily
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 2021-2022
Indicative performance: 110%
Estimated volatility: 28.94%

Source paper:

Lopez-Lira, A. and Zang, Y.: Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4412788
Abstract:
We examine the potential of ChatGPT, and other large language models, in predicting stock market returns using sentiment analysis of news headlines. We use ChatGPT to indicate whether a given headline is good, bad, or irrelevant news for firms’ stock prices. We then compute a numerical score and document a positive correlation between these “ChatGPT scores” and subsequent daily stock market returns. Further, ChatGPT outperforms traditional sentiment analysis methods. We find that more basic models such as GPT-1, GPT-2, and BERT cannot accurately forecast returns, indicating return predictability is an emerging capacity of complex models. ChatGPT-4’s implied Sharpe ratios are larger than ChatGPT-3’s; however, the latter model has larger total returns. Our results suggest that incorporating advanced language models into the investment decision-making process can yield more accurate predictions and enhance the performance of quantitative trading strategies. Predictability is concentrated on smaller stocks and more prominent on firms with bad news, consistent with limits-to-arbitrage arguments rather than market inefficiencies.

New research papers related to existing strategies:

#575 – Momentum and Low Risk Effects in India
#620 – Long Term Time-series Momentum in India

Raju, Rajan: An Examination of Number of Holdings and Universe Size in Momentum Strategies: Evidence from India
https://ssrn.com/abstract=4453680
Abstract:
This paper explores the effects of the number of portfolio holdings and the size of the investment universe on momentum strategies in India. Using mixed linear models, we test hypotheses related to exposure to market, size, and momentum factors, the influence of universe selection on these exposures, and the role of idiosyncratic risk. While concentrated portfolios offer superior factor exposure, they also carry higher idiosyncratic risk. On a risk-adjusted basis, highly concentrated portfolios do not outperform. Our findings also illustrate that the choice of the investment universe and the number of holdings in the portfolio significantly affect performance momentum strategies. We propose a framework targeting craftsmanship alpha, the excess returns skilled managers can achieve beyond quantitative strategies.

#12 – Pairs Trading with Stocks

Xu, Jing and Yang, Peiquan: Pairs Trading with Costly Short-Selling
https://ssrn.com/abstract=4470496
Abstract:
We study an optimal pairs-trading model with costly short-selling, allowing for an empirically documented negative relation between the short selling costs and future asset returns. When the investor has a logarithm utility function, we derive the solution in closed form, which indicates that: (i) the optimal allocations are piecewise linear functions of the degree of mispricing; (ii) short selling costs asymmetrically reduce the optimal size of the long/short position; and (iii) for risk-hedging purpose, it can be optimal to short sell the overvalued asset even when its short selling costs dominate the abnormal return earned from short selling. When the investor exhibits a general constant relative risk aversion preference, we propose an allocation rule which has a closed form and approximates the optimal one very well. We also show that short selling costs can increase the cost of enforcing delta neutrality in pairs trading.

#685 – Boosted Trees and Cryptocurrency Return Prediction
#823 – Machine Learning and the Cross-Section of Cryptocurrency Returns
#875 – Trend-Based Machine Learning Crypto Strategy

Rizzuti, Luca and Parente, Mimmo and Trerotola, Mario: A Profitable Trading Algorithm for Cryptocurrencies Using a Neural Network Model
https://ssrn.com/abstract=4471119
Abstract:
Algorithmic trading enables the execution of orders using a set of rules determined by a computer program. Orders are submitted based on an asset’s expected price in the future, an approach well suited for high-volatility markets, such as those trading in cryptocurrencies. The goal of this study is to find a reliable and profitable model to predict the future direction of a crypto asset’s price based on publicly available historical data. We first develop a novel labeling scheme and map this problem into a Machine Learning classification problem. The model is then validated on three major cryptocurrencies through an extensive backtest over both a bull and a bear market. Finally, the contribution of each feature to the classification output is analyzed.

#7 – Low Volatility Factor Effect in Stocks

França, Luciano and de Avelar Fernandes Filho, Mario Candido and Portella Teles, Pedro Paulo: Low Volatility Asset Valuation in Brazilian Stock Market: Lower Risk with Higher Returns
https://ssrn.com/abstract=4480285
Abstract:
This work evaluates the behavior of portfolios comprised of Brazilian stocks ranked by their volatility to investigate the low volatility anomaly.
Between January 2003 and December 2021, the low volatility portfolio presented a 6% annual return above the high volatility portfolio. This result is aligned with the observation made by Blitz and Van Vliet (2007) in global markets, with an annual alpha spread of 12% over the period between 1986 and 2006.
Also, through a double sorting process, it was possible to obtain portfolios with higher returns and lower risk than those ranked by a single risk factor, although this difference was not statistically significant in most cases.

#903 – Technical Sentiment Index in Crypt
#904 – Fundamental Sentiment Index in Cryptocurrencies

Akyildirim, Erdinc and Aysan, Ahmet Faruk and Çepni, Oğuzhan and Corbet, Shaen: Exploring the Impact of News Sentiment on Defi Coin Returns
https://ssrn.com/abstract=4404802
Abstract:
This paper examines the factors that affect the returns of Decentralized Finance (DeFi) coins and emphasizes the significance of news-based sentiment in the market. Results show that sentiment has a notable impact on DeFi returns, with negative sentiment presenting greater influence than positive sentiment. Additionally, transaction volume and network security are critical drivers of DeFi coin returns, with smaller coins being more sensitive to news sentiment and exhibiting greater return volatility. The impact of news-based sentiment on DeFi returns is greater during the week, possibly due to the reduced presence of institutional investors and trading algorithms. These findings have implications for investors and policymakers, with evidence suggesting the existence of multiple pathways to manipulate DeFi pricing under certain market conditions.

#460 – ESG Level Factor Investing Strategy

Juddoo, Kumari and Malki, Issam and Mathew, Sudha and Sivaprasad, Sheeja: An Impact Investment Strategy
https://ssrn.com/abstract=4411670
Abstract:
Impact investing is based on using the ESG framework as a tool to evaluate firms that engage in generating positive impact. Most impact investors and fund managers now integrate the ESG framework in their investment and stock-picking process. However, due to lack of standardisation of ESG reporting, it remains a challenge for investors and the public to identify the truly sustainable companies. We propose an additional measure of tax avoidance to identify firms that are socially responsible. When firms indulge in excessive tax avoidance behaviour, it may be viewed as unethical or socially irresponsible. We integrate the empirical association between corporate social responsibility and tax avoidance into an investment strategy based on impact. We adopt an investment strategy based on firm‐level ESG ratings and tax avoidance practices. In a pure impact investment strategy based on ESG and tax avoidance, we find that investing in high‐ESG rated firms and low tax avoidance firms yields a buy and hold abnormal return of 3.4% per annum and 11.4% in a 3 years investment horizon. Next, if impact investors were to combine traditional investment strategies based on risk with impact measures, we find that portfolios of high‐ESG and high price‐to‐book‐ratio firms earn a buy and hold abnormal return of 21.2%, while a portfolio of low tax avoidance and high price-to-book portfolios earns 29.8% in the long run. Collectively, our results suggest that, whilst impact investing does provide investors a return, it does not necessarily outperform traditional investment strategies. Our results are robust to other risk factors and the sector of the firm.

#480 – Machine Learning-Based Financial Statement Analysis
#537 – The Positive Similarity of Company Filings and Stock Returns

Han, Henry and Wu, Yi and Li, Deqing Diane and Ren, Jie: Forecasting Stock Excess Returns with Sec 8-K Filings
https://ssrn.com/abstract=4263876
Abstract:
The stock excess return forecast with SEC 8-K filings via machine learning presents a challenge in business and AI. In this study, we model it as an imbalanced learning problem by proposing a multiclass SVM forecast with tuned Gaussian kernels to handle it. The proposed model performs better than peers from state-of-the-art deep and machine learning. We also show that the TF-IDF vectorization would demonstrate advantages over the BERT vectorization in the forecast. Unlike general assumptions, we find that dimension reduction generally lowers forecasting effectiveness compared to using the original high-dimensional vectorized data. Furthermore, inappropriate dimension reduction may increase the overfitting risk in the forecast or cause the machine learning model to lose its learning capabilities. We also find that resampling techniques cannot enhance forecasting effectiveness for high-dimensional imbalanced data. In addition, we propose a novel dimension reduction stacking method to retrieve both global and local data characteristics for high-dimensional vectorized data that outperforms other peer methods in forecasting and decreases learning complexities. The algorithms and techniques proposed in this work can help stakeholders optimize their investment decisions by exploiting the 8-K filings besides shedding light on AI innovations in accounting and finance.

And several interesting free blog posts have been published during last 2 weeks:

Technical Analysis Report Methodology + Double Bottom Country Trading Strategy

Some of the more vague terms in Technical Analysis are really hard to quantify as nearly every TA user defines and interprets them differently. We mean mainly TA patterns like supports, resistances, trend lines, double tops, double bottoms, and/or more complex patterns like head-and-shoulders. Now, what we can do with that? We tried to spend some time and fought a little with some of these TA terms, and the following article/study results from our attempts to quantify a tiny subset of the world of Technical Analysis patterns.

Avoid Equity Bear Markets with a Market Timing Strategy – Revisiting Our Research

In March, we posted a series of three articles where our goal was to construct a market timing strategy that would reliably sidestep the equity market during bear markets. In this article, we revisit our research to address the forward-looking bias in our final market timing strategy. Upon careful examination, we identified a bias in our macroeconomic trading signal based on the U.S. S&P Composite dividends. To eliminate the issue, we have replaced the signal from U.S. S&P Composite dividends with Housing Starts Growth sourced from FRED, ensuring the strategy is no longer biased.

The unbiased version of our TrendYCMacro strategy, which uses the HOUSE signal, yields an annual excess return of 6.59%, slightly below the 7.10% of the biased version with the DIVIDEND signal. Interestingly, the unbiased version experiences slightly lower annualized volatility at 11.87% compared to the 11.89% of the biased version. Both versions have suffered the same maximal drawdown of -25.13% and exhibit comparable risk-adjusted returns, with the unbiased version having a Sharpe ratio of 0.56 and the biased version having a Sharpe ratio of 0.60.

Predicting Stock Market Performance with the Global Anomaly Index

Today’s article focuses on investigating long-short anomaly portfolio return predictability in international stock markets, which often undergo mispricing due to investors’ sentiment. A paper by Jiang, Fuwei et al. (Apr 2023), suggests using the AAIG (Global Anomaly Index), and it examines the ability of the aggregate anomaly index to predict future returns in 33 stock markets. While previous research finds that a high aggregate anomaly measure predicts a low return in the U.S. market, this study further demonstrates that the global component of AAI (aggregate anomaly indices) is the key that drives international return predictability and reveals that the global anomaly index is a strong and robust predictor of equity risk premiums not just in the U.S. market but also in international markets, both in- and out-of-sample, consistently delivering significant economic values.