Something to bear in mind is that when determining the likelihood of encountering a class of a given size, it’s the larger ones that need to be weighted more.
So if in a given year Brown offers 437 classes of 2-9 students but 55 of 100+, courses in the latter range (biggest classes) are significantly more likely to be encountered than courses in the former (smallest classes).
(Take the average of each range, multiple it by the number of classes for that range, then compare for total seats occupied. Doing that predicts the 100+ classes are twice as likely if assuming the max end of the range is 100 and approaches 3x if assuming the max end is 150. If the max end is closer to 400 as I’ve read in their student paper, then the larger classes could be 6x as likely, depending on the actual distribution of counts in the range.)
Often intro STEM courses tend to be among the biggest on a campus. Experiences in those early courses can sometimes determine whether a student sticks with STEM.