Preregistration asks researchers to commit to their hypotheses and analysis plan before collecting data. Nothing stops them from analysing the data afterwards in any way they like, and the commitment is public rather than enforced. Why does it reduce false positives anyway?

B. Because it separates a prediction from something noticed afterwards.

The problem is degrees of freedom rather than dishonesty: any dataset offers dozens of defensible choices, and reporting the one that reached significance runs many tests while presenting one. A public commitment does not remove the freedom, it removes the ability to present a discovery as a prediction, and exploratory work stays legitimate once labelled.