In statistics, it depends on whether we are interested in sampling from finite populations. Usually we are not, except when we are in the survey literature. The concept of the population for finite population sampling is straightforward. It is the population from which you sample. For example, for NBA players, it would be the actual players themselves in a particular season.
More often, we assume we are sampling from an infinite population. This can be applied to the case even when the population you sample from is apparently finite. For example, with the sampling of NBA players, we can imagine a larger population of "potential" NBA players, including future players. Since this hypothetical population does not exist, we can imagine it to have infinite possibilities. In this sense, the population is more than just the sample space. It also has a population distribution associated with it. In classical statistics, this distribution is assumed "fixed but unknown", and the purpose of our sample is to make inferences (intelligent guesses) about it. We also tend to make assumptions to ensure we have a "nice" problem to work with. For example, we would typically assume height is on a continuous scale, even though measurements are of finite precision (discrete). We may even assume it follows a Normal distribution if that helps.
This is to say, a "population" is generally a hypothetical construct to allow us to do maths. It may be regarded as an approximation to the "real" population we are interested in making inference on.
