Seriously, a model that can predict things better than a coin flip will be more scientific. Even if it utilizes something that is somewhat hypothetical / not immediately observable. Say, Newtonian mechanics talks about forces - but forces is a totally made up concept, their nature is (for the most part) not explained within the Newtonian mechanics framework at all, but it's a good model.
Also, many models of reality that are "inferior" to the "real models" are still very helpful and useful in science. Many numerical methods scientists are happy to use, say finite differences, discretize continuous equations. This transformation makes the model strictly speaking worse. But this is fine, since it allows to produce calculations that match and predict experiments. Even in Physics, people start using ML / neural networks to approximate complex calculations. Not because a neural network is a better more descriptive model of the reality (of course it's not), but because it calculates the answer close enough to the real one.
Crank ideas are crank not because they use use some made up concepts that cannot be experimentally seen. They are crank, because they ignore mainstream development, staying blissfully unaware of the subtleties and details the mainstream theories have already considered and resolved.
So the main fault of crank theories is the ignorance of their creators, who are either not willing or not able to correctly contextualize their work within the previously existing knowledge. If you want to do science, you have to do your homework: 1) explore and learn what is known already, 2) develop it further or propose an alternative, 3) contextualize your work within the existing knowledge about the subject. Skipping steps 1 and 3 is dishonest.
Seriously, a model that can predict things better than a coin flip will be more scientific. Even if it utilizes something that is somewhat hypothetical / not immediately observable. Say, Newtonian mechanics talks about forces - but forces is a totally made up concept, their nature is (for the most part) not explained within the Newtonian mechanics framework at all, but it's a good model.
Also, many models of reality that are "inferior" to the "real models" are still very helpful and useful in science. Many numerical methods scientists are happy to use, say finite differences, discretize continuous equations. This transformation makes the model strictly speaking worse. But this is fine, since it allows to produce calculations that match and predict experiments. Even in Physics, people start using ML / neural networks to approximate complex calculations. Not because a neural network is a better more descriptive model of the reality (of course it's not), but because it calculates the answer close enough to the real one.
Crank ideas are crank not because they use use some made up concepts that cannot be experimentally seen. They are crank, because they ignore mainstream development, staying blissfully unaware of the subtleties and details the mainstream theories have already considered and resolved.
So the main fault of crank theories is the ignorance of their creators, who are either not willing or not able to correctly contextualize their work within the previously existing knowledge. If you want to do science, you have to do your homework: 1) explore and learn what is known already, 2) develop it further or propose an alternative, 3) contextualize your work within the existing knowledge about the subject. Skipping steps 1 and 3 is dishonest.