Quant LetterNo. 37Wednesday 14 February 2024

February 2024, Week 2

71 items across 7 sections, as sent to readers on 14 February 2024. Paper titles open their ML-Quant page; ↗ goes to the source.

Items: 71
Papers: 43
Sections: 7
Issue: No. 37

View as Markdown Open in Claude Open in ChatGPT Open in PerplexitySite index for AI (llms.txt)

arXiv

Quantitative-finance and ML-for-finance preprints from arXiv.

12 items

Finance7

Autoencoder Trading Strategies

The research shows that using Autoencoder architectures in Statistical Arbitrage simplifies strategy development and improves returns compared to traditional methods.

13 Feb 20244 shares1 citation todaySource ↗

Closed-Form AMM Arbitrage Solutions

The article introduces a new mechanism for identifying arbitrage trades on automated markets, providing better opportunities and quicker capitalization than previous methods, and enabling on-chain arbitrage bots for multi-asset pools.

9 Feb 20244 shares1 citation todaySource ↗

Causal Network Contagion VaR for Risk Management

The paper presents the Causal-NECOVaR, a new method for financial risk analysis that provides reliable risk predictions regardless of market shocks and systemic changes.

8 Feb 20244 shares2 citations todaySource ↗

Monotone Control Integration for Option Pricing

The study introduces a new solution to the Hamilton-Jacobi-Bellman equation in option pricing, proving its stability, consistency, and effectiveness against traditional methods.

9 Feb 20242 shares2 citations todaySource ↗

Statistical Arbitrages: A Optimization Approach

A Optimization Approach: A new technique is suggested for identifying statistical arbitrages, not limited to traditional pairs, through a portfolio optimization problem solvable by the convex-concave procedure.

12 Feb 20244 shares1 citation todaySource ↗

Portfolio Optimization with Costs and Preferences

The Merton investment-consumption problem is expanded to incorporate transaction costs and stochastic differential utility, using new math techniques to understand all parameter combinations and previously difficult aspects.

13 Feb 20244 shares4 citations todaySource ↗

Multivariate CRPS Learning

A new method for merging multivariate probabilistic forecasts, considering dependencies between quantiles and marginals, has shown significant improvements in predicting day-ahead electricity prices.

17 Mar 202317 shares24 citations todaySource ↗

Crypto & Blockchain5

Liquidity Provision in Crypto Markets

A study reveals that the liquidity provision premium in cryptocurrency markets can be predicted using factors like the VIX index and Tether liquidity, and is influenced by stock market premiums globally.

21 Mar 2022394 sharesSource ↗

Bundling Broadband for K-12 Schools: Cost Savings & Improved Speed

Cost Savings & Improved Speed: A study revealed that New Jersey K-12 schools experienced a decrease in broadband internet prices by a third and a sixfold increase in speed when they switched to a bundled procurement system in 2014. The schools also saved an amount equivalent to their total federal E-rate subsidy, resulting in significant welfare gains.

11 Feb 20242 sharesSource ↗

Fragmentation & Liquidity on DEX

Research indicates that high-fee trading pools on Uniswap attract more liquidity providers but have lower trading volumes, suggesting that fragmented liquidity boosts market participation and competition.

17 Nov 2022648 sharesSource ↗

Emoji Sentiments in Crypto Markets

Research using GPT4 and a BERT model shows that Twitter emoji sentiment can predict cryptocurrency market trends and help avoid major downturns.

11 Feb 20244 shares5 citations todaySource ↗

Salience Theory and Crypto Returns

The Salience theory, based on returns and trading volume, has a negative predictive power for forecasting return trends in the cryptocurrency market, with trading volume being a key factor.

9 Feb 20243 sharesSource ↗

SSRN

Working papers in finance and economics from SSRN.

18 items

Quantitative9

Asset Embeddings

The paper suggests that investors' holdings data, when analyzed with artificial intelligence and machine learning, can reveal significant company traits.

20 Jul 20232 sharesSource ↗

WTI Futures Market View-Taking

Commercial positions in WTI futures often fail, with financial commercials outperforming supply chain ones.

13 Feb 20246 sharesSource ↗

Low Turnover Portfolio vs. 50/50 Strategy

A low turnover portfolio, slowly readjusted to fixed weights, performs better than the standard equities/bonds portfolio and is a viable alternative to a simple momentum or value portfolio.

12 Feb 20242 sharesSource ↗

Forecasting US Single Family House Prices with GMDH

The article discusses the use of a machine learning technique, the group method of data handling (GMDH), for predicting house price index (HPI), leading to more accurate housing market forecasts.

14 Jun 20232 sharesSource ↗

Polynomial Factor Models: Nonlinear Relationships

Nonlinear Relationships: Polynomial Factor Models (PFM) provide a novel method for handling high-dimensional panel data, allowing for the consistent estimation of factor interactions and loadings by capturing nonlinear relationships.

31 Jan 20232 sharesSource ↗

Reinforcement Learning for Optimal Execution

A new actor-critic reinforcement learning algorithm is introduced for optimal execution problem, featuring a recalibration step for convergence and showing linear convergence under appropriate conditions.

10 Mar 2023372 sharesSource ↗

Market Model Calibration with Neural Network: Faster Efficiency

Faster Efficiency: Machine learning can streamline the calibration of market models, reducing computational time and increasing practicality, with a new approach allowing model parameters to evolve as a stochastic process for a robust training set.

5 Jan 20242 sharesSource ↗

Inflation and Directional Volatility Ratio

The article presents a new method called the directional volatility ratio for predicting inflation trends, which is more effective than traditional methods.

14 Oct 20232 shares1 citation todaySource ↗

Testing q-theory under Endogenous Truncation: Bias Correction

Bias Correction: A new methodology corrects bias in empirical investment studies that rely on truncated samples of publicly listed firms, supporting the q-theory and showing that investment-cash flow sensitivity disappears and the relation between investment and q increases fourfold.

26 Jul 20222 sharesSource ↗

Financial9

Credit Options and Economic Fluctuations

Options on the CDX index can predict short-term economic downturns by indicating changes in credit risk premia and shifts in credit market conditions.

9 Feb 20243 sharesSource ↗

Interest Rate Differential and Currency Returns

The study reveals that the prospective interest rate differential is a better predictor of currency excess returns than carry, explaining returns of various currency portfolios.

9 Feb 20243 sharesSource ↗

Treasury Market Depth Measurement

The study indicates that the market depth of U.S. Treasury securities, a key liquidity measure, remains largely consistent despite different measurement decisions in depth calculations.

12 Feb 20242 sharesSource ↗

Factor and Stock Return Optimization

A new asset pricing factor, created using optimal portfolio weights to maximize the Sharpe ratio, can explain the cross-section of stock and bond returns, even when accounting for popular factors.

9 Feb 20242 sharesSource ↗

Assaying Anomalies in Equity Returns

A new protocol and web application are proposed for testing potential predictors of equity returns, providing thorough analysis and identifying common problems in testing equity strategies.

26 Jan 2023970 sharesSource ↗

HFT & Option Market Liquidity

A study finds that increased aggressive high-frequency trading in equity markets results in wider bid-ask spreads in the options market due to sniping risk and informed trading.

18 Feb 20212 sharesSource ↗

Algorithmic Trading & CEO Turnover

Algorithmic trading leads to directors relying less on stock returns for CEO turnover decisions, instead focusing more on nonmarket measures.

13 Sep 20222 sharesSource ↗

Behavioral Influences on Contrarian Retail Trading

Retail investors' selling behavior, influenced by unrealized capital gains, impacts short-term return reversals and volatility among certain stocks.

25 Jul 20222 sharesSource ↗

Institutional Investors & Tiny Retail Trades

Institutional investors delay sales and accumulate stakes in response to expected small retail trades, leading to delayed price discovery.

15 Jan 20243 sharesSource ↗

RePEc

Economics working papers from RePEc's NEP field reports.

11 items

Finance3

Trading Behavior of Individual Investors in Mutual Funds

A study found that individual investors in mutual funds are momentum buyers and contrarian sellers, with older and larger transaction size investors more likely to be momentum buyers.

14 Feb 202414 sharesSource ↗

Trading Responses to Credit Rating Announcements

Investors respond differently to changes in credit ratings, selling during downgrades by investor-paid agencies and buying during upgrades by issuer-paid agencies.

14 Feb 202411 sharesSource ↗

Opening Price Gaps and Information Adjustments in Stocks

AI and big data research shows negative gap openings are more frequent than positive ones, and price adjustments for bad news are faster than for good news.

14 Feb 202411 sharesSource ↗

Machine Learning4

Accounting for Spatial Autocorrelation in Hedonic Models

The study proposes a spatial cross-validation strategy to correct bias in tree-based algorithms caused by spatial autocorrelation in real estate data, improving accuracy in mass appraisal and investment decisions.

14 Feb 202423 sharesSource ↗

Improved Prediction of Global Gold Prices with Hurst-Reconfiguration

The paper introduces a hybrid forecasting model for gold prices, combining Hurst-oriented reconfiguration and machine learning, outperforming traditional models in prediction errors and accuracy.

14 Feb 202416 sharesSource ↗

Active Learning for Stacking and AdaBoost Models

The research shows that active learning within ensemble learning can achieve similar predictive performance with fewer selected instances compared to using full data.

14 Feb 202411 sharesSource ↗

Machine Learning Simplifies Finance in Real Time

The article introduces a new deep learning algorithm designed to solve complex financial models. This algorithm provides fresh economic insights and lowers computational costs.

14 Feb 202411 sharesSource ↗

Historical Trending4

Uncertainty & Exchange Rate Volatility

The research indicates that exchange rate volatility is increased by economic policy and global financial market uncertainty, but reduced by US monetary policy uncertainty.

28 Jul 202329 sharesSource ↗

Adaptive Online Portfolio Selection with Transaction Costs

The research proposes a new algorithm for online portfolio selection that improves return prediction accuracy by considering peer impact.

6 Feb 202323 sharesSource ↗

Decomposing Downside Investment Risk: Centred Expected Shortfall

Centred Expected Shortfall: The article recommends using Centred Expected Shortfall as a risk measure in asset management for a more accurate portfolio risk breakdown.

18 Aug 202322 sharesSource ↗

Asset Diversification in High Volatility Markets

The article suggests optimizing the risk-return ratio of an investment portfolio by adjusting investment proportions for each asset based on the economic activity cycle.

6 May 202321 sharesSource ↗

Machine learning

The general machine-learning papers the letter carried in 2023-25.

2 items

Recently Published2

Scaling Laws for Mixture of Experts

The research introduces a new hyperparameter, granularity, to Mixture of Experts models, improving training optimization and outperforming dense Transformers.

12 Feb 2024120 shares181 citations todaySource ↗

WebLINX: Website Navigation

Website Navigation: The study presents WEBLINX, a benchmark for conversational web navigation, and a model that ranks relevant HTML elements, emphasizing the need for large multimodal models.

8 Feb 202471 shares189 citations todaySource ↗

Papers with code

Papers that shipped their code, from the Papers with Code feed (2023-25).

5 items

Trending2

Time Series Forecasting Models

The second article examines the profound influence of foundation models on machine learning, focusing on their zero-shot and few-shot generalization capabilities.

11 Feb 2024296 shares

Large Language Models & Time Series Analysis

The article highlights the crucial role of time series analysis in comprehending the intricacies of diverse real-world systems and applications.

11 Feb 2024174 shares

Rising3

Sim2Real Transfer in 18 Seconds

The new framework enables efficient control of a multirotor using a regular laptop and microcontrollers with minimal training.

11 Feb 2024133 shares

Self-Discovery in Large Language Models

SELFDISCOVER is a novel framework that helps Language Model Learning systems identify task-specific reasoning structures for complex problem-solving.

12 Feb 2024110 shares

Strong Generalization with MetaTree Decision Trees

MetaTree is a system designed to create trees that demonstrate high generalization performance.

12 Feb 202457 shares

GitHub

Repositories the letter featured.

7 items

Finance4

Orion: Signal Anomaly Detection

Signal Anomaly Detection: The piece introduces a machine learning library designed to detect anomalies in signals.

24 Jul 2018924 shares

TradingTechnicalIndicators: Trading Indicators Library

Trading Indicators Library: The article discusses a Python library used for implementing trading technical indicators.

27 Sep 202098 shares

Probatus: Classifier and Data Validation

Classifier and Data Validation: The piece focuses on the validation of data classifiers and the data used in their creation.

9 Nov 2020114 shares

TimeLLM: Large Language Model Forecasting

Large Language Model Forecasting: The article details the official implementation of TimeLLM for time series forecasting, as presented at ICLR 2024.

20 Jan 2024170 shares

Trending3

LLMs: Low Latency JSON Generation

Low Latency JSON Generation: The article explains the process of creating JSON with minimal delay using Large Language Models.

15 Nov 2023256 shares

pkl: Config as Code Language with Validation and Tooling

Config as Code Language with Validation and Tooling: The article presents a new coding language for configuration that provides comprehensive validation and advanced tooling capabilities.

19 Jan 20246,653 shares

tfcausalimpact: Python Causal Impact Implementation

Python Causal Impact Implementation: The article showcases a Python version of Causal Impact, inspired by Google's R package, utilizing TensorFlow Probability.

17 Aug 2020534 shares

X / Twitter

Posts from quant researchers on X.

16 items

Quantitative8

Low-risk diversification strategies white paper by AQR

AQR's white paper indicates that low-risk strategies are less sensitive to macro conditions and volatility than traditional assets and average hedge funds.

8 Feb 20245 shares