The fact that we see this huge peak on the histogram at 35 and that we also see local maxima at the bottoms of the other two medal tiers is suggestive of a process which is not plain “impartial and blind grading” (perhaps that part of the IMO methodology could also use some improvement).
If it resembles the International Chemistry Olympiad (which, like most I[X]Os is based on the IMO) then yeah it’s weird and adversarial. But the threshold for Gold here is exactly 5⁄6 questions fully correct, which is also a natural breakpoint. This happens since generally you either have a proof or you don’t and getting n-1/n points usually means something like missing out a single case in a proof by exhaustion, which is much less common than just failing to produce a proof. Most of the people who got 35⁄42 did so with scores of 7, 7, 7, 7, 7, 0. So there’s that factor as well.
Ah, yes, you are right. And the silver medal threshold is 28=4*7. So this is much more natural, and mostly comes from how the competition is structured (the scoring factor still looks somewhat noticeable to my eye, but is much less of a problem than I thought).
If it resembles the International Chemistry Olympiad (which, like most I[X]Os is based on the IMO) then yeah it’s weird and adversarial. But the threshold for Gold here is exactly 5⁄6 questions fully correct, which is also a natural breakpoint. This happens since generally you either have a proof or you don’t and getting n-1/n points usually means something like missing out a single case in a proof by exhaustion, which is much less common than just failing to produce a proof. Most of the people who got 35⁄42 did so with scores of 7, 7, 7, 7, 7, 0. So there’s that factor as well.
Ah, yes, you are right. And the silver medal threshold is 28=4*7. So this is much more natural, and mostly comes from how the competition is structured (the scoring factor still looks somewhat noticeable to my eye, but is much less of a problem than I thought).