Shows the value which splits the relative error distribution in half. A lower number (below 0) means that on average the model is underestimating the prices, and a higher number (above 0) means that the model is overestimating the prices. The ideal value for this measure would be approximately 0.