If you cannot describe the physical or logical mechanism driving the inner workings of a phenomenon, you do not have a theory, you have a statistical anecdote. One must descend from the pedestal of abstraction and explain the operational “how.” Without a mechanism, a paper is merely literature. Every variable added post-hoc is a confession of error. A model that requires “magic constants” to avoid collapse is hereby refuted as a fundamental law.

If you introduce X, tell me how it is touched, how it is measured, and what instrument changes in its presence. If it is invisible and immeasurable by design, it is academic quackery.

Physics, as a science, must be rooted in the direct observation of nature. It must strive to understand the fundamental nature of the underlying phenomenon. Propose an optimal model that explains the phenomenon through real physical mechanisms, devoid of abstract mathematics or statistical patches.

The mathematical framework must be generated only after the physical mechanism is understood. Utilize engineering principles to conduct experiments with applied formulas and perform rigorous calibration.

The principle of reality (parsimony): If a model explains or approximates reality more accurately with fewer parameters, IT IS THE ONE CLOSEST TO REALITY.

Pseudo-scientific fallacies: Models based on abstract mathematics unfalsifiable, failing to explain underlying phenomena, or capable of explaining both ‘A’ and its contradiction, where every error is patched by inventing a new ad hoc formula, are nothing more than pseudoscience, “formalism cosplay,” and brute force. They are merely numerical horoscopes.

Predictive supremacy: A true physical model must not merely curve-fit historical data. It must blindly and accurately predict future states or entirely unseen datasets without requiring post-hoc recalibration. This invalidates those who overfit parameters. If it doesn’t predict the future, it is a broken odometer.

A fundamental physical mechanism must remain coherent across different orders of magnitude.

Additional Epistemological Principles

Every new entity or variable must be anchored to an observational or experimental procedure. If X is introduced, you must define how it is measured, with what instrument, and what distinct prediction it produces.

A theory must specify concrete conditions under which it would be discarded. Example: “If in dataset Y, under condition Z, the error exceeds T, the mechanism is refuted.”

Every non-observable entity carries an epistemological cost. It is only accepted if it reduces total complexity and improves out-of-sample prediction. This is not “bad math”; it is cheap ontology.

A good model does not “explain everything”, it distinguishes between scenarios. If your theory can accommodate both A and non-A, you are not predicting, you are narrating.

The Principle of Reality (Parsimony-Occam’s Razor)

If two models explain or predict reality with comparable out-of-sample accuracy, the one with fewer parameters, fewer assumptions, and fewer invented entities is closer to reality.

Anti-Inverse Modeling

Any framework that starts from abstract mathematics and only later searches for a physical interpretation, shielding itself from falsification by adding ad hoc assumptions, is not physics, it is formalism cosplay.

Refutation by Proliferation of Patches

Popper was an idealist, he claimed science is what can be refuted. The problem is that modern formalism learned to dodge Popper. If a prediction fails, they simply invent an invisible particle or an ad hoc constant. If a model needs 10 patches to avoid failure, it has already failed.