What I don't understand is why anti-plagiarism software they mention seems to compare test solutions in isolation. Only 2 out of 450 made this particular mistak

What I don't understand is why anti-plagiarism software they mention seems to compare test solutions in isolation. Only 2 out of 450 made this particular mistake, but where did they sit on the exam ? If they made the same mistake and they sit within 5m of each other, that would be much stronger accusation. The bottom line: software should take into account not just how similar two solutions are, but how far from each other both students were.