The article employs a new forecasting methodology for out-of-sample forecasting of US house prices when there might be multiple structural breaks in the time-series. For this purpose, we use a theoretical framework called 'Breaks-Unknown Forecast (BUF)', which is developed upon an already existing methodology of forecasting in the presence of breaks in time series. We, further, examine the forecasting ability of the proposed methodology against several other benchmark models, popularly used for predicting US house prices. According to our empirical findings, the Breaks-Unknown Forecast methodology outperforms other benchmark models in terms of forecasting accuracy.