Yeah I noticed that too, but I think the pre-tests have some traits of the final tests, so finetuning to pre-tests might generalize to edge cases in final tests.

Also luck is 30% of the points as this is heuristics lol