Article Details
  • Published Online:
    June  2026
  • Product Name:
    The IUP Journal of Financial Risk Management
  • Product Type:
    Article
  • Product Code:
    IJFRM010626
  • DOI:
    10.71329/IUPJFRM/2026.23.2.5-25
  • Author Name:
    Aniket Singh, Blessy Thomas, Devangi Sharma and Seshanwita Das
  • Availability:
    YES
  • Subject/Domain:
    Finance
  • Download Format:
    PDF
  • Pages:
    5-25
Volume 23, Issue 2, April-June 2026
Duration Matters: A Semi-Markovian Approach to Forecasting Short-Term Return Momentum
Abstract

The paper unveils a revised logit model designed to forecast intraday short-term momentum in Indian equity markets. A semi-Markov logistic regression model combines state switching probabilities and sojourn-time (duration) characteristics on three years of 5-min trading data of 200 active traded stocks in six intraday horizons (15-90 min). There are considerably large differences between sojourn distributions and geometric assumptions, which substantiate dependence on duration. The model provides stable prediction and mean accuracy of 75.46% and AUC-ROC of 0.832 with a 45-min interval yielding the most viable tradeoff between responsiveness and stability. The paper shows that duration-adjusted transition characteristics have actionable predictive content and provide a more interpretable and transparent alternative to data-driven intraday trading models.

Introduction

The efficient market hypothesis (EMH) is a theory of asset prices, which assumes that all available information is reflected in the price of assets, and excess returns are not predictable (Fama, 1965). Nevertheless, the body of evidence on recurring empirical data has shown that there is consistent short-term momentum, and according to the data, assets that have been improving in the previous period continue to show improvement over close time periods (Jegadeesh & Titman, 1993).