You attempted every question, felt good about the coverage, and scored below the mark you would have had by leaving the hard ones blank. The score report shows a large wrong count and a penalty column you never looked at before the test. A classmate who attempted fewer questions scored higher, which reads as unfair and is not.
A machine-scored aptitude test: 40 questions, 60 minutes, four options per question, marking scheme plus one for a correct answer and minus 0.25 for a wrong one. The scheme was stated in the instructions email and in the on-screen instructions page. The student had heard "never leave a blank" from a senior and applied it as a rule with no arithmetic behind it. Linked module: 6.2 (Online assessment prep).
Reconstruct the attempt. She solved 16 questions by actual work. With about 4 minutes left she blind-marked the remaining 24 in roughly 50 seconds, one option each, no reading. Four of those 24 came out correct and 20 came out wrong.
Score the two worlds. Blind-marking: 16 + 4 - (20 x 0.25) = 16 + 4 - 5 = 15.0. Leaving all 24 blank: 16 + 0 - 0 = 16.0. The 24 guesses cost her exactly 1.0 mark against doing nothing at all.
Now the part that surprises people. Those guesses were not against the rules of arithmetic. Expected value per blind guess at four options with a 0.25 penalty is (1/4)(+1) + (3/4)(-0.25) = 0.25 - 0.1875 = +0.0625, so 24 of them are worth 24 x 0.0625 = +1.5 marks on average. Her expected score was 17.5 and she drew 15.0. The rule was right and the draw went against her, because a per-question edge of 0.0625 marks is far too thin to carry a section.
That is the real diagnosis. The failure was not "she guessed"; it was that guessing was being used as a rescue for 24 unsolved questions rather than as a mop-up on the last few. The same 4 minutes spent differently changes the expected score materially, and the scheme decides how much. Under a harsher scheme — plus one, minus one, five options — the expected value per blind guess is 0.2 - 0.8 = -0.6, so those same 24 guesses would have been worth -14.4 marks and would have destroyed the section outright.
Turn the instruction line into a threshold before the test starts.
If the scheme is minus one, do the opposite of instinct: leave them blank, and spend the endgame checking your solved answers instead.
Write the marking scheme at the top of your rough sheet in every practice round, and score every practice round with the penalty applied. Untimed, unpenalised practice teaches a scoring habit the real test punishes. Keep a one-line rule card in your notes — n < 1 + 1/m — and rehearse it until you can state the threshold for a new scheme in under ten seconds. And treat any advice of the form "always attempt everything" or "never guess" as incomplete until someone supplies m and n.
Q: "How do you make decisions when the information is incomplete?" This is an unusually good story for that question. Describe the marking scheme as a payoff rule, derive the threshold in one line, and separate the decision from the outcome: the rule was positive expected value, the single draw was poor, and the fix was not to abandon the rule but to raise the edge by eliminating options first. Candidates who can say "a good decision and a bad outcome are different things" are describing exactly the judgement a production on-call rota needs.