Global magnetohydrodynamic (MHD) modeling is a powerful tool in space physics research. There are several advanced and still developing global MHD codes that are widely used to simulate plasma processes in solar wind magnetosphere-ionosphere system. The verification of global simulation codes is an important but a difficult problem. We present an approach for systematic and quantitative testing of code performance based on statistical empirical dependencies of the key magnetospheric parameters obtained from observations. We demonstrate the applicability of the method by testing the Grand Unified Magnetosphere Ionosphere Coupling simulation (GUMICS-4) global MHD model. A large set of nearly stationary solutions (162 runs altogether) with different stationary interplanetary magnetic field (IMF) and solar wind inputs were generated for different dipole tilts and levels of solar EUV radiation. As key parameters, we use the large-scale characteristics of the magnetosphere, including the magnetopause size and shape