You have to consider what Occam's Razor is about:
Take two explanations of the same human behaviour, for example altruism. They both are "reasonable" explanations. But one explanation has to claim that there are hippie-forces making people happy if they are altruistic and the other explanation does the same task without having to claim that there are hippie-forces, but stays in the theoretical framework needed for the explanation of many, many other facts.
Which one is better? Well, it is better to cut off what is not needed. That's what we do if we use the argument of Occam's Razor.
As a less polemic example from philosophy: There are theorists that claim that there are "sense data" as instances in addition to world and mind. They use them to explain how knowledge about the world is possible. Other theorists do not need these instances to explain knowledge about the world, but do only have to claim that there are world and mind. Therefore sense data theories (in making these claims) are worse than others.
It's not about the complexity of explanations in relation to facts. It is about the complexity of explanations in relation to explanations, given that all explanations are fully appropriate to the complexity of all facts considered.
That is very important in natural sciences as in philosophy, because it means, methodologically, that we should only claim additional things/forces/faculties/spins/quarks/laws etc., if the existing ones cannot explain all facts in their complexity.