This paper considers the problem of forecasting from Markov switching models and proposes weighting observations to obtain optimal forecasts in the MSFE sense. We derive optimal weights for one step ahead forecasts. We provide analytical expressions for optimal weights in Markov switching models conditional on the size and dates of switches. It is shown that the optimal weight is the same across observations within a given regime and differs only across regimes. In practice, where information on switches is uncertain, we derive weights that are conditional on the probability of switches. The relative performance of our proposed approach is investigated using Monte Carlo experiments and an empirical application.