An interesting question that will have many (equally valid answers). My opinion (not saying this is correct) from a statistical point of view:

If we had enough time we could use statistics to DESCRIBE all complexities of the "real world". But it would take a very long time, and would in fact be an ongoing process as the "real world" is constantly changing.

However most statistical models are not engaged in description, what statistical models are usually interested in is INFERENCE. Here the aim is not to describe complexity, rather it is to find an underlying simplicity that can be used to infer to other samples/ situations/ worlds even. The best statistical models are parsimonious (i.e. as simple as possible to achieve inference). Very complex outcomes can be produced from simple underlying relationships between things, in statistics we are often most interested in the relationships rather than describing all of the arising complex presentations. An example of a parsimonious model (simple equation) in physics explaining a huge amount of reality would be E = M C squared.

An age-old and widely debated question: there is no univocal and precise answer.