Paper: arXiv 2305.17523

Authors: Jaydip Sen, Aditya Jaiswal, Anshuman Pathak, Atish Kumar Majee, Kushagra Kumar, Manas Kumar Sarkar, Soubhik Maji

Abstract

This paper presents a comparative analysis of the performances of three portfolio optimization approaches. Three approaches of portfolio optimization that are considered in this work are the mean-variance portfolio (MVP), hierarchical risk parity (HRP) portfolio, and reinforcement learning-based portfolio. The portfolios are trained and tested over several stock data and their performances are compared on their annual returns, annual risks, and Sharpe ratios. In the reinforcement learning-based portfolio design approach, the deep Q learning technique has been utilized. Due to the large number of possible states, the construction of the Q-table is done using a deep neural network. The historical prices of the 50 premier stocks from the Indian stock market, known as the NIFTY50 stocks, and several stocks from 10 important sectors of the Indian stock market are used to create the environment for training the agent.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper involves advanced mathematics including Markowitz optimization and Reinforcement Learning, but appears to be a capstone project with limited empirical details and no access to the full backtest results, placing it in the Lab Rats quadrant.

Research Flowchart

  flowchart TD
  A["Research Goal: Compare Portfolio Optimization<br>Performance on Indian Stock Market"] --> B["Data Collection: NIFTY50 &<br>10 Sector Stocks"]
  B --> C{"Methodology Implementation"}
  C --> D["Mean-Variance Portfolio MVP"]
  C --> E["Hierarchical Risk Parity HRP"]
  C --> F["Reinforcement Learning RL<br>Deep Q-Learning with DNN"]
  D & E & F --> G["Performance Evaluation<br>Annual Return, Risk, Sharpe Ratio"]
  G --> H["Key Findings: Comparative Analysis<br>Performance Rankings & Insights"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a comparative analysis of three portfolio optimization approaches—mean‑variance, hierarchical risk parity, and reinforcement‑learning—using Indian stock data.

Source sentence
This paper presents a comparative analysis of the performances of three portfolio optimization approaches. Three approaches of portfolio optimization that are considered in this work are the mean-variance portfolio (MVP), hierarchical risk parity (HRP) portfolio, and reinforcement learning-based portfolio.

Assumes 6

  • Asset returns are normally distributed statistical
    Source sentence
    Normal distribution of asset returns: The Markowitz model assumes that asset returns are normally distributed, which means that they follow a bell-shaped curve.
  • Investors are risk‑averse model
    Source sentence
    Investors are risk-averse: The Markowitz model assumes that investors are risk-averse, which means that they prefer portfolios that provide a higher expected return for a given level of risk or a lower level of risk for a given expected return.
  • Single‑period investment horizon model
    Source sentence
    Single period investment: The Markowitz model assumes that the investment period is a single period, which means that it does not consider the effects of compounding over time.
  • No taxes, transaction costs, or other frictions execution
    Source sentence
    No taxes, transaction costs, or other frictions: The Markowitz model assumes that there are no taxes, transaction costs, or other frictions associated with buying or selling assets.
Show 2 more
  • Stable correlations between assets over time market
    Source sentence
    Stable correlation between assets: The Markowitz model assumes that the correlation between assets is stable over time.
  • Short‑selling is allowed execution
    Source sentence
    No short-selling restrictions: The Markowitz model assumes that investors can sell assets short, which means they can borrow assets and sell them with the expectation of buying them back at a lower price in the future.

Data

  • Historical prices of NIFTY50 stocks and sector stocks market-prices public
    50 premier Indian stocks (NIFTY50) and several stocks from 10 important sectors
    Source sentence
    The historical prices of the 50 premier stocks from the Indian stock market, known as the NIFTY50 stocks, and several stocks from 10 important sectors of the Indian stock market are used to create the environment for training the agent.

Recognised sources: Yahoo Finance

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
No taxes, transaction costs, or other frictions: The Markowitz model assumes that there are no taxes, transaction costs, or other frictions associated with buying or selling assets.

Finds 2

  • Reinforcement‑learning portfolio achieves higher Sharpe ratios than both MVP and HRP. performance
    Source sentence
    From the Sharpe ratios, we can conclude that RL is capable of outperforming traditional portfolio optimization techniques like MVP & HRP.
  • HRP outperforms MVP on out‑of‑sample test data, delivering a higher Sharpe ratio despite slightly higher volatility. performance
    Source sentence
    the HRP portfolio has outperformed MVP portfolio on the test (i.e., out-of-sample) data producing a higher value of Sharpe ratio, with slightly higher volatility.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 42 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.