The focus of this paper is to test an effective rebalancing method that prioritizes trades with the strongest signals to capture more of the factor premium while reducing turnover and trading costs. The authors coined the term “smart rebalancing” which involves prioritizing trades based on the strength of their signals in order to minimize costs and turnover while retaining much of the factor return. Widely used factor strategies, including long-short factors and long-only factor-based strategies are analyzed using the smart rebalancing approach. The goal is to capture more risk premium or alpha associated with a factor or strategy through better portfolio trading techniques. Although the title of this piece is a bit intimidating at first, the principles of smart rebalancing described are broadly suitable for any number of investment strategies.

Uncertainty-Aware Lookahead Factor Models for Quantitative Investing

  • Lakshay Chauhan, John Alberg, Zachary C. Lipton
  • White Paper
  • A version of this paper can be found here
  • Want to read our summaries of academic finance papers? Check out our Academic Research Insight category.

What are the research questions?

  1. Can trading costs and other market frictions be managed to improve the live performance of common factor strategies?
  2. The authors develop and test several methods to reduce turnover including priority-best, proportional, and priority-worst rebalancing. Which work to reduce turnover costs without sacrificing returns?
  3. Do rebalancing approaches based on non-calendar events improve the performance of high-turnover strategies, such as momentum?
  4. What is new and interesting about the methodologies used to analyze “smart rebalancing” techniques?

What are the Academic Insights?

  1. YES. The study shows that careful and considerate portfolio trading, particularly through the priority-best rebalancing method, can significantly mitigate trading costs and other market frictions. This method involves prioritizing trades based on the strength of their signals, focusing on the most impactful transactions.
  2. YES. Smart rebalancing effectively reduces turnover and trading costs while maintaining or even enhancing portfolio performance. By prioritizing trades that have the strongest signals, investors can capture more of the factor premium. The priority-best rebalancing (prioritizes trades with the strongest signals) was first in performance, followed by the proportional (distributes trades evenly, which can lead to unnecessary trades and higher costs) and then the priority-worst methods (prioritizes the least useful trades which maximizes “noise”). The priority-best retained more of the factor premium, achieved higher net returns, and captured more alpha per unit of turnover than the other two. Truly smart rebalancing on a noncalendar basis.
  3. YES. For high-turnover strategies like momentum, non-calendar-based rebalancing (where the portfolio is monitored continuously, and rebalancing is triggered when the deviation from the target portfolio exceeds a preset threshold) seems to be more effective. It allows for more flexible and timely adjustments, potentially reducing unnecessary trades and associated costs.
  4. The authors employ various AI techniques and models to enhance portfolio rebalancing strategies and improve the performance of factor-based investment strategies. Here are short explanations of the AI techniques and models used successfully to implement “smart rebalancing”.
    • Deep Neural Networks (DNNs) are used to forecast fundamentals from past values. Two types of DNNs used include Multi-Layer Perceptrons (MLPs) to model complex relationships between inputs and outputs and are akin to regression tasks. Long Short-Term Memory Networks (LSTMs) which are capable of processing and predicting time series data.  
    • Ensemble Learning is used to improve the accuracy of the forecasts made from DNNs which helps reduce overfitting and associated biases.
    •  Uncertainty Quantification approached essentially produce a mean and variance for each forecast, therefore incorporating uncertainty into each prediction made.

Why does it matter?

The findings provide valuable insights for investors and portfolio managers on how to better implement factor strategies in real-world conditions by utilizing effective AI methods to make forecasts. Reducing trading costs, utilizing smart rebalancing methods such as turnover reduction and non-calendar rebalancing methods, practical solutions for improving net returns analyzed in this research. When applied to factor strategies, the focus on the most impactful trades and turnover not only enhanced performance, but managed risk more effectively. The performance results presented below for the LFM UQ-LSTM are presented in the chart below. LFM UQ-LSTM refers to Lookahead Factor Model with Uncertainty Quantification using Long Short-Term Memory networks.

 The most important chart from the paper

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained.  Indexes are unmanaged and do not reflect management or trading fees, and one cannot invest directly in an index.


On a periodic basis, publicly traded companies report fundamentals, financial data including revenue, earnings, debt, among others. Quantitative finance research has identified several factors and functions of the reported data that historically correlate with stock market performance. In this paper, we first show through simulation that if we could select stocks via factors calculated on future fundamentals (via oracle), that our portfolios would far outperform standard factor models. Motivated by this insight, we train deep nets to forecast future fundamentals from a trailing 5-year history. We propose lookahead factor models which plug these predicted future fundamentals into traditional factors. Finally, we incorporate uncertainty estimates from both neural heteroscedastic regression and a dropout-based heuristic, improving performance by adjusting our portfolios to avert risk. In retrospective analysis, we leverage an industry-grade portfolio simulator (backtester) to show simultaneous improvement in annualized return and Sharpe ratio. Specifically, the simulated annualized return for the uncertainty-aware model is 17.7% (vs 14.0% for a standard factor model) and the Sharpe ratio is 0.84 (vs 0.52).

Print Friendly, PDF & Email

About the Author: Tommi Johnsen, PhD

Tommi Johnsen, PhD
Tommi Johnsen is the former Director of the Reiman School of Finance and an Emeritus Professor at the Daniels College of Business at the University of Denver. She has worked extensively as a research consultant and investment advisor for institutional investors and wealth managers in quantitative methods and portfolio construction. She taught at the graduate and undergraduate levels and published research in several areas including: capital markets, portfolio management and performance analysis, financial applications of econometrics and the analysis of equity securities. In 2019, Dr. Johnsen published “Smarter Investing” with Palgrave/Macmillan, a top 10 in business book sales for the publisher.  She received her Ph.D. from the University of Colorado at Boulder, with a major field of study in Investments and a minor in Econometrics.  Currently, Dr. Johnsen is a consultant to wealthy families/individuals, asset managers, and wealth managers.

Important Disclosures

For informational and educational purposes only and should not be construed as specific investment, accounting, legal, or tax advice. Certain information is deemed to be reliable, but its accuracy and completeness cannot be guaranteed. Third party information may become outdated or otherwise superseded without notice.  Neither the Securities and Exchange Commission (SEC) nor any other federal or state agency has approved, determined the accuracy, or confirmed the adequacy of this article.

The views and opinions expressed herein are those of the author and do not necessarily reflect the views of Alpha Architect, its affiliates or its employees. Our full disclosures are available here. Definitions of common statistics used in our analysis are available here (towards the bottom).

Join thousands of other readers and subscribe to our blog.

Print Friendly, PDF & Email