The value is in the machinery, not in the output. A correct result you cannot reproduce leaves you where you started, while a mistake you can explain gives you a model that will be right the next four times. Marks measure the answer, learning measures the understanding, and the two come apart constantly.

After you get something wrong, write out in plain language why your reasoning went where it did. When you get something right, check that you could rebuild it from nothing. If not, treat it as unlearned and go back, even though the page says it is correct.