How do you debug a hidden test failure when the grader only says “wrong answer”?

Treat the vague verdict as a debugging lesson: if you cannot reconstruct what happened, the program probably does not expose enough useful information. Make sure you can record or inspect the relevant inputs and state locally, then reduce failing behavior to the smallest case you can reproduce. Also have another person or a code-review tool inspect the specification and solution, but use that as a second opinion rather than blindly accepting a suggested fix.

Use targeted instrumentation instead of adding prints everywhere. Log key inputs, intermediate values, loop boundaries, and the final result for small cases, then build a collection of adversarial tests designed to break each assumption. A debugger can also help you step through those cases and compare the program’s behavior with a simple, obviously correct implementation.

Start by checking the boundaries you may have overlooked: empty input, zero, negative values, duplicate items, already-sorted or reverse-sorted data, maximum sizes, and values near numeric limits. Then reread the specification carefully and compare every assumption in your code with what the problem actually guarantees. A single hidden failure often comes from input handling or an unstated assumption rather than the main algorithm.