For example, frictionless surfaces, point masses and ideal gases are useful starting points, even though real systems rarely behave exactly that way.

I’m interested in how people decide when a simplified model is still useful and when its assumptions become too restrictive. Is it mainly a matter of comparing predictions with experimental results, or are there other ways to recognise when a model needs to be refined?

I’d appreciate examples from different areas of physics where a simple model provides a good explanation in some conditions but becomes misleading in others.

For example for projectiles. We know we have to include a bunch of factors, Coriolis force, air resistance, or even relativistic calculations. However we also know from the problem when these factors are less important and we can choose to ignore them to see how well it fits the problem or to what level of perturbation theory we need to have a good enough fit.

Most problems in condensed matter could be solved by turning to your atomistic simulation guy to make very heavy calculations in order to validate their experiment/phenomena. However most of the times the calculation is so long or expensive (too many atoms) that this is not the usual approach. Theoretical CM physicists have become experts in taking a very big Hamiltonian that includes everything and then grinding it with approximations to obtain an effective equation that takes into account the features that are absolutely key for the experiment.

This post is very vague but the idea is that we always know (unless we are talking about quantum gravity) what the deeper theory is.

For example for projectiles. We know we have to include a bunch of factors, Coriolis force, air resistance, or even relativistic calculations. However we also know from the problem when these factors are less important and we can choose to ignore them to see how well it fits the problem or to what level of perturbation theory we need to have a good enough fit.

Most problems in condensed matter could be solved by turning to your atomistic simulation guy to make very heavy calculations in order to validate their experiment/phenomena. However most of the times the calculation is so long or expensive (too many atoms) that this is not the usual approach. Theoretical CM physicists have become experts in taking a very big Hamiltonian that includes everything and then grinding it with approximations to obtain an effective equation that takes into account the features that are absolutely key for the experiment.