Detailed insights in complex physical situations, e.g. highly interacting many body systems, is strongly supported by numerical methods like Monte-Carlo or Langevin dynamics simulations. An alternative approach is the utilization of colloidal suspensions, which consist of mesoscopic particles being suspended in fluids. In thermal equilibrium, such particles - similar to the above mentioned numerical techniques - rapidly sample their accessible configurational space. Accordingly, colloidal suspensions can be regarded as analog computers which allow real time "calculations" in complex situations such as crystal growth, glass formation etc. Consequently, colloids are versatile model systems and can substantially contribute to the understanding of processes in atomic systems but also elucidate problems in the context of statistical physics. In the first part of the talk we discuss the role of entropy in binary colloidal systems. In contrast to what is often found in textbooks, an increase of entropy does not necessarily lead to a higher degree of disorder but can also lead to crystallization due to entropic forces. Such entropic forces are crucial for the understanding of phase separation phenomena in polydisperse particle mixtures and biological systems. Secondly, we study the relationship between particle fluctuations and melting. The enhancement of particle fluctuations leads typically to a higher degree of disorder and thus to melting. Although this intuitive relation between melting and fluctuations is widely accepted among physicists, fluctuations can also enhance order in systems which are exposed to periodic substrate potentials. We studied this - at the first glance paradoxical effect - by a two-dimensional system of charged colloids and present a simple model which explains this intriguing phenomenon in terms of particle fluctuations which tend to stabilize the crystalline phase.