Scientists, Eccentrics, Cranks and Crackpots

General scientific community
Most of the scientific community approach science in an incremental fashion, building heavily on previous work. Most scientific models start out very simple. However, the real world is more complex and as measurements and mathematical techniques improve, the model must be refined to handle some of these details.

As an example, consider the two-body gravitational problem. The two-body problem can be modified to include the gravitational effects of more 'small' objects through perturbation analysis. If you want to consider many objects of about similar masses, you must solve the N-body problem which must be computed by numerical methods as there is no simple formula solution like the two-body problem. Most scientists spend their time not developing new models, but refining existing models and solution methods to handle more realistic situations.

Eccentrics
Scientific eccentrics often examine unconventional ideas, but don't become attached to them if they don't match up with observations or experiments. I've found papers published in professional journals where researchers have examined many of the ideas claimed by the cranks but the paper were mostly demonstrations of why that particular idea doesn't work. A number of very successful scientists do this, they look at lots of ideas, but only hang onto the ones that work. The downside of this is that the ideas that failed still tend to get fans among the cranks and crackpots.

Other eccentrics will examine unconventional ideas, but become a little more attached to them even after they are found to have serious disagreements with observations and experiments. However, they continue to explore and revise their theories. They actually try to follow good scientific methodologies, dumping aspects of their theories that don't work and going back to the drawing board. Many of them support unconventional ideas because they regard the dominant paradigm as somehow 'unsatisfying'. Such sentiments generally suggest an emotional rather than intellectual or evidence-based belief.

Cranks and crackpots
The cranks and crackpots examine unconventional ideas, but become strongly attached to them. Any error in their analysis, even basic errors that can be easily demonstrated, are not acknowledged. In the rare case where they do provide a reasonably testable scientific model, the one successful prediction it might make, in their minds, outweighs the dozens of predictions where it fails.

The cranks rarely use good scientific methodologies, instead claiming 'logic', or the 'obvious' interpretation of some data-driven image or movie is the basis for their knowledge. They make extreme claims about how their theories explain every observation, yet it is difficult to impossible to get these theories into a form that others can do calculations which can be compared to actual measurements. The theories of cranks and crackpots flunk the important scientific test of reproducibility. Their theories are often only supported by a tiny set of cherry-picked, anecdotal, events extracted from a much larger body of contrary evidence.