Oh wow props to USMLE for cracking down on this. Still blows my mind that this method of cheating works. Doesn’t the usmle has a huge bank of active questions from which they construct each exam? You’d need a massive number of questions to really make a difference right?

The statistical answer is "no," actually, because while no one really knows how exactly the Step exams are graded, there are a lot of pieces out there strongly suggesting the difference between, say, a 250 and a 270 comes out to a few questions/block. The average student taking a standardized test like this is already getting most of the questions correct.

To use hypothetical numbers, let's say there are 300 questions on a given test, drawn from 3,000 possible questions in a question bank. You (hypothetical student) would get an average of 75% of those 3000 questions correct (2250) based on your studying. I give you a subset of 300 questions beforehand and tell you the answers, so you'd get 100% of those specific questions correct. We don't know how many of those questions you'd get on the actual test (which will be a random collection of 300 questions of the 3000, likely mostly different questions) but of the total bank, your total pre-test projected "correct" answers just went up by by 75 (the 25% of the 300 questions I gave you that you would've otherwise missed), raising your overall total from 2250 to 2325.

Sounds miniscule, but based on the estimates we have (which might be speculation) going from 75% correct to 77.5% correct would raise your score by ~8 points. Now who knows how big the actual test bank is, but based on what I've read about this, it sounds as though the cheaters had access to potentially thousands of questions beforehand (a number of people took pictures of every question they got on the test, and that was aggregated and distributed). It'd be like doing a section of UWorld you've already done before.