I here introduce a free fitness function in population biology, which monotonically increases with time and takes its maximum at the evolutionary equilibrium. By suitably defining an "index" for each state, the free fitness is expressed as the average index plus an entropy term. In many cases, the index has a biologically clear meaning, such as the logarithmic population mean fitness. The technique is applicable to any Markov process model (either continuous or discrete) with a positive steady state. I discuss four examples from various branches of population biology: (1) one-locus-two-allele system of population genetics with mutation, selection, and random genetic drift; (2) evolutionary dynamics of quantitative characters; (3) a molecular evolution model; and (4) an ecological succession model. Introducing free fitness clarifies the balance between systematic forces (e.g. natural selection or successional trend toward the climax) and disturbing processes (e.g. random drift).