Received signal strength indication fingerprinting (RSSIF) is an indoor localization technique that exploits the prevalence of wireless local area networks (WLANs). Past research into RSSIF systems has seen the development of a number of algorithmic methods that provide effective indoor positioning. A key limitation, however, is that the performance of these methods is heavily dependent on practical implementation parameters and the nature of the test-bed environment. As a result, past research has tend to only compare algorithms of the same paradigm using a specific test-bed, and thus making it difficult to judge and compare their performance objectively. There is, therefore, a critical need for a study that addresses this gap in the literature. To this end, this paper compares a range of RSSIF methods, drawn from both probabilistic and deterministic paradigms, on a common test-bed. We evaluate their localization efficiency and accuracy, and also propose a number of improvements and modifications. In particular, we report on the impact of dense and transient access points (APs) – two problems that stem from the popularity of WLANs. Our results show methods that average the distance to the k nearest neighbors in signal space perform well with reduced dimensions. Moreover, we show the benefits of using the standard deviation of RSSI values to exclude transient APs. Other than that, we outline a shortcoming of the Bayesian algorithm in uncontrolled environments with highly variable APs and RSSI values, and propose an extension that uses a mode filter to restore its accuracy with increasing samples.