Using the Markov switching autoregressive (MSAR) model, this study examines the return behaviour of the BSE Sensex index from April 1999 to March 2025. The best model for the data was found to be the MSAR (2, 1) model, which includes a first-order autoregressive (AR) component and two different regimes. According to empirical findings, at a 1% significance level, the dependent variable’s first-order lag significantly increased returns in both Regimes 0 and 1. However, in Regime 1, the AR component demonstrated a considerable negative influence, which was not seen in Regime 0. According to the transition probability matrix (TPM), there was a 73.11% chance that the market would go from Regime 1 to Regime 0, and a 49.74% chance that it would move from Regime 0 to Regime 1. The findings imply that, although volatility is treated as being constant throughout regimes, regime transitions are mostly caused by modifications in the mean dynamics of returns. This suggests that although the variation is consistent across all regimes, the market fluctuates between distinct return patterns. This study adds to our understanding of the volatility of the Indian stock market by detaching light on the market dynamics and regime switching (RS) behaviour of the BSE Sensex.