This paper explores, from a theoretical and technical point of view, the role of head rotation and eye gaze directionality to the human perception of attention from non-verbal cues. We have annotated two different versions of the same dataset, in order to correlate the above parameters with the degree people have considered people in the dataset pay attention to a hypothetical task they have in front of them. Based on our findings, we investigate the role of eye gaze directionality in relation to head rotations and, based on previous studies, we developed an algorithm for estimating attention levels from head pose, eye gaze and facial feature spatial locations. With the help of our AI system and people's annotation, we have made a first attempt towards quantifying the role of each cue to the overall estimation of attention. One of the important properties of the technical part of this work is that all systems we used were non-intrusive and did not demand any personal training or calibration phase, constituting themselves ideal for Game playing. Knowing the behavioral state of a player can be of vital importance for adapting the game design during interaction, or building personal profiles, leading to appropriate game features aiming at maximizing player satisfaction.