Hi! Here's Iván with this week's exciting newsletter, brimming with insights and discoveries on building robust investment strategies and risk models using Machine Learning.

In this edition, I am presenting the following sections:

• 🕹️ AI-Finance Insights: I summarize three must-read academic papers that mix cutting-edge ML/DL with Quant Finance:

  • Supervised Portfolios (Paper with code!)

  • Detecting intraday financial market states using temporal clustering

  • Stock Price Prediction Based on Morphological Similarity Clustering and Hierarchical Temporal Memory

• 💊 AI Essentials: The section on top AI & Quant Finance learning resources: Today, I'm excited to share a presentation by Prof. Bryan Kelly at the Swedish House of Finance on the paper "The Virtue of Complexity." This is a must-watch for anyone interested in the theoretical and empirical perspectives on how machine learning can improve stock price predictions among other key insights.

• 🥐 Asset Pricing Insights: In this edition, I recommend a paper showing that conditioning on volatility in multifactor portfolios outperforms single-factor volatility-managed strategies.

Today’s Sponsor

Missed out on Ring and Nest? Don’t let RYSE slip away!

Ring 一 Acquired by Amazon for $1.2B

Nest 一 Acquired by Google for $3.2B

If you missed out on these spectacular early investments in the Smart Home space, here’s your chance to grab hold of the next one.

RYSE is a tech firm poised to dominate the Smart Shades market (growing at an astonishing 55% annually), and their public offering of shares priced at just $1.50 has opened. 

They have generated over 20X growth in share price for early shareholders, with significant upside remaining as they just launched in over 100 Best Buy stores.

Retail distribution was the main driver behind the acquisitions of both Ring and Nest, and their exclusive deal with Best Buy puts them in pole position to dominate this burgeoning industry.

Secure your stake before the offer closes.

AI-Finance Insights

“Supervised Portfolios”

Finally, we are seeing papers that include complete code in well-known quantitative finance journals. In this paper, the authors obtain optimal weights and then use them in a supervised approach.

My takeaways from the paper (links at the end of the post):

➡ The paper proposes an innovative asset allocation strategy that first calculates optimal portfolio weights before utilizing them in a supervised learning algorithm, diverging from traditional approaches that prioritize expected returns.

➡ This strategy allows the machine to incorporate a wider array of risk measures, preferences, and constraints in a flexible, forward-looking, and non-linear framework, significantly improving upon the simple mean-variance models.

➡ Empirical analysis reveals that directly predicting optimal weights rather than expected returns not only stabilizes portfolio compositions but also enhances risk-adjusted performance metrics, resulting in better Sharpe ratios.

➡ The benefits include reduced trading costs—particularly for investors who are less risk-averse—and higher average returns, albeit with a slight increase in realized risk.

➡ These findings are robust across different sub-periods and remain valid even when simpler regression models replace more complex algorithms like boosted trees, suggesting a broad applicability and effectiveness of the direct weight prediction method.

Link to the code: https://colab.research.google.com/drive/1yrhuV5i_o2g0Ju-xJEyONR5s2PFqyk3k?usp=shari

“Detecting intraday financial market states using temporal clustering” 

In this paper, intraday temporal cluster configurations are used to identify market states, and the study of these states helps extract state signature vectors for real-time detection.

Main ideas from the paper in 2 min (links at the end of the post): 👇

➡ The paper introduces a high-speed maximum likelihood clustering algorithm to detect temporal financial market states from intraday data, diverging from traditional methods by focusing on market microstructure features rather than just price or volume data.

➡ This approach involves identifying temporal cluster configurations to understand market states and developing state signature vectors. These vectors serve as compact descriptors that facilitate real-time state detection in high-frequency trading scenarios, enhancing trading strategies.

➡ A novel technique extracts characteristic signatures of market activity from each identified state, enabling real-time state recognition by trading agents. This capability allows traders to adapt their strategies dynamically, depending on the current market state.

➡ The paper also explores the use of 1-step transition probability matrices constructed from the state signatures. These matrices are refined online and employed in optimal planning algorithms to predict and capitalize on probable future market movements effectively.

➡ The methodology not only supports immediate and accurate market state detection but also offers a systematic approach to integrating these insights into quantitative trading models, potentially improving decision-making and operational efficiency in high-frequency trading environments.

“Stock Price Prediction Based on Morphological Similarity Clustering and Hierarchical Temporal Memory”

There is a lot of useful information in correlation matrices that we often ignore. The paper proposes a clustering method that allows us to enrich our models with valuable cross-sectional information.

Here are my takeaways from the paper (link at the end of the post): 👇

➡ The paper introduces a novel clustering method that combines Morphological Similarity Distance (MSD) and k-means clustering to identify similar stocks, improving predictions by analyzing stock correlations within the market.

➡ Employing Hierarchical Temporal Memory (HTM), an online learning model, this approach, termed C-HTM, learns from clustered similar stocks to significantly enhance prediction accuracy compared to traditional models.

➡ Experiments on stock price predictions highlight that C-HTM not only outperforms the standard HTM but also all baseline models in short-term forecasting, showcasing superior performance.

➡ Key innovations include the first implementation of the KMSD clustering algorithm for mining stock similarities across the entire market and applying the HTM model to these clusters, achieving promising results in short-term price prediction tasks.

➡ This approach suggests further applications in dynamic and complex market environments, potentially improving strategies in financial markets through advanced pattern recognition and learning capabilities.

AI-Essentials

Presentation of the paper 'The Virtue of Complexity in Financial Machine Learning' at the Swedish House of Finance by Prof. Bryan Kelly. A must-watch for those of you interested in understanding, from both theoretical and empirical perspectives, why ML improves stock price predictions among other important key takeaways. 👇

Sponsor 👇

Learn how to research faster, automate tasks & simplify your life using AI with this 3 hour FREE workshop. 

Register & join now (FREE for First 100) 🎁

Asset Pricing Insights

“A Multifactor Perspective on Volatility-Managed Portfolios”

Conditioning on volatility from a portfolio perspective (multifactor portfolios) outperforms the well-known single-factor volatility-managed strategies.

So far, this is an 'easy to implement/replicate' paper, based on sound theory and well-executed by top academics. As a disclaimer, I might be a bit biased because it is closely related to my current research lines. :-)

My takeaways from the paper (links at the end of the post):

➡ The paper introduces a novel conditional multifactor portfolio strategy that dynamically adjusts its asset weights based on market volatility, challenging traditional static approaches and the classic risk-return tradeoff.

➡ The strategy incorporates a multifactor perspective to asset allocation, enhancing the portfolio's ability to adapt to changes in market conditions and outperforming both unconditional and other conditional multifactor portfolios in out-of-sample tests and net of costs (Sharpe Ratios).

➡ The results demonstrate that the proposed strategy effectively navigates various market sentiments and volatility levels, maintaining superior performance by leveraging a nuanced understanding of risk factors over simple mean-variance models.

➡ The study reveals that the gains from this volatility management strategy are not diminished by estimation errors, transaction costs, or shifts in market sentiment, suggesting a more complex yet rewarding approach to managing investment risk.

➡ The findings underscore the breakdown of the traditional risk-return relationship, offering new insights into asset management that could significantly benefit both high and low-risk investors without compromising on returns.

If you're enjoying our newsletter and want to support us, please recommend it to anyone you know who's interested in AI and Finance. Your referrals are the biggest compliment and help us grow! 🌟🤖💼