Energy efficiency is one of the main benchmarks of performance in visible light communication. Achieving high energy efficiency in a link is a challenging task when high data throughput is required. A promising approach to tackling this challenge is using multiple-input-multiple-output (MIMO) systems, which use the spatial domain for information encoding. A novel modulation scheme called Flexible light emitting diode (LED) index keying (FLIK) can harness high spectral efficiency by utilising active and inactive LED states. The high spectral efficiency, together with a straightforward encoding, makes FLIK based design a promising candidate for high energy efficiency and data throughput solutions. However, the system’s performance based on FLIK depends heavily on beam selection participating in the link, which can significantly vary with the channel conditions subject to the user’s position and orientation. In a dynamic use case scenario, a fast beam selection and selection re-adjustment are vital for an optimal use case. This study examines the performance of beam selection based on a maximal signalto-noise ratio (SNR) criterion in angle diversity hemispherical transceiver systems. In this paper, a random orientation system model for FLIK is considered. The simulations are then performed considering maximal SNR and maximal Euclidean distance criteria. The performance is evaluated in terms of achievable data throughput. A selection method, based on the maximal SNR, is compared to a method based on maximizing Euclidean distance. The numerical results show that for both the fixed and random orientation cases, a beam selection based on maximal SNR performs as well as the one based on Euclidean distance. This observation is valid up to 25 degrees of beam halfintensity angle, therefore, validating the use of maximal SNR condition in such systems