Accurately estimating the users' perception of adaptive tile-based Virtual Reality (VR) video streaming is fundamental and is often used to determine subsequent streaming decisions. This is due to the fact that adaptive tile-based VR video streaming solutions aim to optimize the bandwidth usage by only streaming the areas (tiles) within the field of view of the user (viewport) at the highest quality, while keeping the remainder of the 360° environment at low quality levels. Thus, understanding the quality perceived by the user in real-time determines the success of the service. Current quality assessment approaches tend to map the quality of the viewport's center (center tile) to the overall perceived quality. However, while such models enable to measure how long the user spend on each quality layer, they ignore the fact that the user's eyes also move within the viewport. In this paper, we present a novel video quality assessment metric for adaptive tile-based VR video streaming which takes into account the distribution of eye movement within the viewport, as well as, the tiling scheme. Both elements are combined by means of a probability distribution fit from real users data. The model was evaluated in a well-known VR database and using an emulation environment, benchmarking it against state-of-the-art solutions as well as objective full reference metrics.