I think the best way is to know a handful of benchmark pKas (e.g., HCl, HOAc, EtOH, t-BuOH, acetylene, dimethyl malonate, benzene, diisopropyl amine, butene, butane) and then extrapolate the pKa of other molecules by rationilizing relative to the ones that you know. For example, if you know that the pKa of HOAc is 4.76, you could predict that the pKa of trichloroacetic acid would be lower, which it is (pKa ~0.63).