New strategies:
#926 – Overnight Effect during High Volatility Days in Bitcoin
Period of rebalancing: Intraday
Markets traded: cryptos
Instruments used for trading: cryptos
Complexity: Moderately complex strategy
Backtest period: 2017 – 2023
Indicative performance: 37.26%
Estimated volatility: 24.63%
Source paper:
Vojtko, Radovan and Javorská, Juliána : The Seasonality of Bitcoin
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4581124
Abstract:
The cryptocurrency market has grown significantly and is no longer considered a small or negligible sector. As the crypto market continues to evolve, researchers are increasingly interested in conducting in-depth analyses, for example, one of the most fascinating phenomena in the world – seasonality effects. We examine the distribution of the daily returns, proposing a simple seasonality strategy based on holding BTC only for two hours per day. Moreover, this paper extends research on potential seasonal patterns related to Bitcoin, focusing on whether these patterns are influenced by factors such as current market trends or the level of volatility in the market.
#927 – Mean-Reversion within Cluster Industries
Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 1984-2019
Indicative performance: 6.4%
Estimated volatility: 6.5%
Source paper:
Goodarzi, M. and Bagnara, M.: Clustering-Based Sector Investing
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4528879
Abstract:
Industry classification groups firms into finer partitions to help investments and empir-ical analysis. To overcome the well-documented limitations of existing industry defini-tions, like their stale nature and coarse categories for firms with multiple operations, we employ a clustering approach on 69 firm characteristics and allocate companies to novel economic sectors maximizing the within-group explained variation. Such sectors are dynamic yet stable, and represent a superior investment set compared to stan-dard classification schemes for portfolio optimization and for trading strategies based on within-industry mean-reversion, which give rise to a latent risk factor significantly priced in the cross-section. We provide a new metric to quantify feature importance for clustering methods, finding that size drives differences across classical industries while book-to-market and financial liquidity variables matter for clustering-based sectors.
#928 – Using Machine Learning to Predict Stock Earnings
Period of rebalancing: Daily
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 2011-2022
Indicative performance: 51%
Estimated volatility: –
Source paper:
Dichev, Huang, Lee, Zhao: You Have a Point – But a Point Is Not Enough: The Case for Distributional Forecasts of Earnings
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4547337
Abstract:
Existing forecasts of earnings are typically expressed as point estimates. However, ex-ante, the future earnings number is unknown, and is statistically represented by a probability distribution over all possible earnings outcomes. We use recent advances in statistical machine learning to estimate the ex-ante distributions of future earnings right before earnings announcements, and investigate how these distributions can help managers, analysts, and investors make better decisions along three directions. First, we show that our distributional forecasts are well calibrated to actual earnings realizations. Second, we document that management and financial analyst forecasts are substantially miscalibrated, severely underestimating the variability of future earnings. Critically, since our distributional estimates are available ex ante at the firm-quarter level, they can be proactively used to identify and correct such miscalibration. Third, we use our distributional estimates to model the probability of beating or missing the consensus analyst forecasts. Going long (short) on stocks in the top (bottom) decile probabilities of beating (missing) the consensus produces hedge returns of about 60 basis points over the three-day earnings announcement window.
#929 – Machine Learning ESG Strategy
Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 2011-2020
Indicative performance: 16.1%
Estimated volatility: 12.85%
Source paper:
Brown, W. O. and Gao, X. and Han, Y. and Huang, D. and Wang, F.: Granular Environmental Variables and Stock Returns
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4504079
Abstract:
Individual environmental variables may contain information obscured in aggregate environmental scores for return forecasting. We apply machine learning methods to granular environmental variables and find that a long-short portfolio that longs stocks with high forecasted returns and sells stocks with low forecasted returns earns about one percent per month. Stocks with high forecasted returns are associated with strong environmental operational performance. Variables related to Scope 3 emissions are neglected risks. The long-short portfolio performs better when climate concerns in the media are more intense.
#930 – Influence of Liquidity, Institutional Ownership & Lottery Effect on Stocks
Period of rebalancing: Yearly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Complex strategy
Backtest period: 1979-2013
Indicative performance: 10.56%
Estimated volatility: 9.85%
Source paper:
Hwang, Chuan-Yang and Yi, Long: Liquidity, Favorite-Longshot Bias, and the Return of Lottery-Like Stocks
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4533690
Abstract:
We document novel results that the lottery effect (lottery-like stocks earn lower returns) is much stronger in liquid stocks. We posit that noise traders dominate the trading of liquid lottery-like stocks, as described in the limits to arbitrage literature, who are attracted by the high jackpot probability of the lottery-like stocks and commit the favorite-longshot bias in which investors overweight the probability of a longshot. The overweight, rather than being driven by preference as suggested in the prospect theory, is more likely caused by overestimation error.
#931 – Forecasting Crude Oil Prices
Period of rebalancing: Monthly
Markets traded: commodities
Instruments used for trading: CDFs, ETFs, futures
Complexity: Moderately complex strategy
Backtest period: 1992-2020
Indicative performance: 4.65%
Estimated volatility: 19.38%
Source paper:
Ma, Yong and Li, Shuaibing and Zhou, Mingtao: Forecasting Crude Oil Prices: Does Global Financial Uncertainty Matter?
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4565543
Abstract:
In this study, we identify a powerful predictor — Global Financial Uncertainty (GFU) — for the prediction of crude oil returns. Our analysis unveils compelling evidence of GFU’s superior predictive power compared to other prominent uncertainty measures and macroeconomic predictors, with in-sample $R^2$ of 5.13\% and out-of-sample $R_{OS}^2$ of 3.27%. Economically, we demonstrate that the mean–variance investor can obtain considerable economic gains based on the return forecasts of GFU relative to the competing predictors. Finally, our empirical results remain significantly robust to the alternative choice of economic metrics, and oil price proxies.
New research papers related to existing strategies:
#906 – Using ChatGPT to Forecast Stock Price Movements
Cheng, Yuhan and Tang, Ke: GPT’s Idea of Stock Factors
https://ssrn.com/abstract=4560216
Abstract:
We amalgamate the capabilities of the GPT-4 computational model with the avant-garde methodology of autonomous factor generation, culminating in the synthesis of high-return factors within the equity investment milieu. Empirical outcomes elucidate that the factors conceptualized by ChatGPT attain a commendable Sharpe ratio peaking at 4.49, accompanied by an annualized return trajectory reaching 66.16\%. Notably, the superlative excess returns garnered remain unaccounted for by the quintessential five-factor model. Through the implementation of an unembellished model averaging paradigm, the ensemble of 35 factors, conceived by ChatGPT, manifests an apex long-short annualized return of 88\% and a Sharpe ratio registering at 2.46. In stark contrast to conventional data mining techniques, the temporal expenditure requisite for GPT’s factor generation is minuscule. It relies on knowledge inference without the need for data input, and it can provide a thorough economic explanation for its factors.
Gupta, Udit: GPT-InvestAR: Enhancing Stock Investment Strategies through Annual Report Analysis with Large Language Models
https://ssrn.com/abstract=4568964
Abstract:
Annual Reports of publicly listed companies contain vital information about their financial health which can help assess the potential impact on Stock price of the firm. These reports are comprehensive in nature, going up to, and sometimes exceeding, 100 pages. Analysing these reports is cumbersome even for a single firm, let alone the whole universe of firms that exist. Over the years, financial experts have become proficient in extracting valuable information from these documents relatively quickly. However, this requires years of practice and experience. This paper aims to simplify the process of assessing Annual Reports of all the firms by leveraging the capabilities of Large Language Models (LLMs). The insights generated by the LLM are compiled in a Quant styled dataset and augmented by historical stock price data. A Machine Learning model is then trained with LLM outputs as features. The walkforward test results show promising outperformance wrt S&P500 returns. This paper intends to provide a framework for future work in this direction. To facilitate this, the code has been released as open source.
#294 – Seasonality Within Trend-Following Strategy in Commodities
Li, Yan and Liu, Qingfu and Miao, Deyu and Tse, Yiuman: Return Seasonality in Commodity Futures
https://ssrn.com/abstract=4575549
Abstract:
We examine seasonality in commodity futures markets using monthly returns for 26 commodities from 1970 through 2022. There are only a few commodities in the early years that show half-month and monthly seasonalities. The same-month trading strategy proposed by Keloharju, Linnainmaa, and Nyberg (2016) outperformed the other-month strategy only for the subperiod of 1990–1999, showing that the findings of Keloharju et al. are not robust to different sample periods. We then conduct back-testing of the momentum strategy to compare its performance with that of a seasonality approach. We find significant momentum returns for most subperiods; however, when combined with the seasonality strategy, the performance is less pronounced. The overall results support market efficiency in commodity futures and seasonality has largely disappeared in recent years. This research has implications for asset pricing and provides insights into the adaptive market hypothesis.
#670 – Machine Learning Pairs Trading Strategy
Roychoudhury, Raktim and Bhagtani, Rahul and Daftari, Aditya: Pairs Trading Using Clustering and Deep Reinforcement Learning
https://ssrn.com/abstract=4504599
Abstract:
Conventional pairs trading strategies are based on concepts of mean reversion and stationary stochastic processes, where pairs are assumed to have linear relationships. However, empirical evidence suggest that asset prices in equity markets frequently exhibit non-linear dynamics. We aim to tackle this issue by exploring pairs trading from a deep learning perspective to leverage these non-linear relationships. Our proposed workflow includes two steps, first clustering equity indices and then using a Reinforcement Learning based trading strategy on pairs selected from these clusters. Using a combination of fundamental and technical signals we extract a set of 10 latent risk factors using a convolutional auto-encoder and create 10 clusters of indices using these risk factors. We test our approach on a sample of 13 pairs to train and test our trading strategy and compare our performance against the S&P 500. From the period starting from April 2017 to December 2022, our best performing strategy consisting of the NASDAQ Composite and MSCI World IT Index earns an annualized return of 21.86% with a Sharpe ratio of 1.15, while generating an alpha of 20%.
#576 – Boosted Regression Trees in Corporate Bonds
#685 – Boosted Trees and Cryptocurrency Return Prediction
Chevalier, Guillaume and Coqueret, Guillaume and Raffinot, Thomas: Interpretable Supervised Portfolios
https://ssrn.com/abstract=4230955
Abstract:
The supervised portfolios approach is an effective asset allocation strategy that engineers optimal weights before feeding them to a supervised learning algorithm. Yet, supervised learning algorithms are often seen as opaque, which undermines trust in those models, thereby limiting their adoption. To alleviate this issue, we apply an enhanced version of RuleFit, an intrinsic interpretable algorithm, which transforms a black-box non-linear predictive algorithm into a simple combination of rules. It fits a sparse linear model that includes feature interactions, derived from decision tree ensembles. Our first empirical analysis illustrates that the interpretable approach is statistically as accurate as gradient boosting on three different investment universes. Our second analysis highlights which characteristics and interactions matter for an equity portfolio manager.
And several interesting free blog posts have been published during last 2 weeks:
Time Invariant Portfolio Protection
In this article we are going to continue the discussion on portfolio insurance strategies. An exhaustive description of this methodology was already presented in the article Introduction to CPPI. This article will focus on an extension of the original model introduced by Estep and. Kritzman (1988), namely Time Invariant Portfolio Protection. Constant Proportion Portfolio Insurance (CPPI) and Time-Invariant Portfolio Protection (TIPP) are two of the most famous portfolio insurance strategies that play an important role in the realm of investment management and risk mitigation. These strategies are designed to address the fundamental challenge of balancing the pursuit of financial growth with the imperative of capital protection against market downturns. Ideally, the guaranteed protection is achieved at the lowest possible premium for the investors.
What’s the Key Factor Behind the Variation in Anomaly Returns?
In a game of poker, it is usually said that when you do not know who the patsy is, you’re the patsy. The world of finance is not different. It is good to know who your counterparties are and which investors/traders drive the return of anomalies you focus on. We discussed that a few months ago in a short blog article called “Which Investors Drive Factor Returns?“. Different sets of investors and their approaches drive different anomalies, and we have one more paper that helps uncover the motivation of investors and traders for trading and their impact on anomaly returns.
Hello ChatGPT, Can You Backtest Strategy for Me?
You may remember our blog post from the end of March, where we tested the current state-of-the-art LLM chatbot. Time flies fast. More than six months have passed since our last article, and half a year in a fast-developing field like Artificial intelligence feels like ten times more. So, we are here to revisit our article and try some new hacks! Has the OpenAI chatbot made any significant improvement? Can ChatGPT be used as a backtesting engine? We retake our risk parity asset allocation and test the limits of current AI development again!
Plus, the following trading strategies have been backtested in QuantConnect in the previous two weeks:
920 – Financial Uncertainty Explains Cryptocurrency Returns
922 – Price-Based Quantitative Strategy for Country Valuation
925 – Option Factor Momentum
926 – Overnight Effect during High Volatility Days in Bitcoin
