Designing an investment strategy in transition economies is a difficult task because stock-markets opened through time, time series are short, and there is little guidance how to obtain expected returns and covariance matrices necessary for mean-variance portfolio allocation. Also, structural breaks are likely to occur. We develop an ad-hoc investment strategy with a flavor of Bayesian learning. An observation is that often an extreme event will herald a new state of the economy. We use this observation to re-initialize learning when unlikely returns materialize. By using a Cornell benchmark, we are able to show the usefulness of our strategy for certain types of re-initializations.