“At the end of the semester, the total scores for all students will be arranged in numerical order, the score that corresponds to the 99th percentile (S99) will be determined, and then letter grades will be

assigned based on this percentile score as follows:

A: Total Score ≥ 0.90 x S99

B: 0.80 x S99 ≤ Total Score < 0.90 x S99

C: 0.70 x S99 ≤ Total Score < 0.80 x S99

D: 0.60 x S99 ≤ Total Score < 0.70 x S99

F: Total Score < 0.60 x S99 or if you fail to complete 10 of the 12 lab

projects”

This is the explanation the department of chemistry for my college gives. But I don’t understand, so please explain it to me like I’m five.

In: Mathematics

In this grading system everyone’s performance is judged relative to one person’s grade.

The person they choose is *almost* the highest scoring person. They skip the highest few people in case those people did much better than the rest of the class (e.g. because they have a tutor that classmates had no access to). For every 100 people in the class they skip one person at the top of the class when looking for the person to grade everyone against.

A popular next step that many courses would employ next is to ask what number needs to be added to that person’s grade for them to get a 100. This curve takes a similar step but instead asks what number needs to be multiplied by that person’s score to get to 100. For example, if the chosen person got an 88 then we’d choose 1.1363636… as the number, since 88 * 1.363636… = 100.

The grader then multiplies everyone’s raw score by that number and sorts into buckets. If you land in 90-100 then you get an A; 80-90 is a B, and so on.

(Note that the description you posted does the math in a different order, but this is equivalent).

You are in college and don’t understand percentile? Or the motivation for using percentiles in grading? Take 100 random humans and measure their height. Arrange them from shortest to tallest. The guy at the end (the tallest one) is in the 99th percentile – he is taller than 99 other guys. Him and the 9 guys preceding him are in the 90th percentile: they are all taller than 90% of the sample (remaining 90 guys) etc etc etc. If you want to grade them, you define grades A: 99>x>=90th percentile; B: 60>x>=89 and so on.

This seems like a weird version of curved grading. Basically, the cut points for grades depend on what the 99th percentile is- to make this easy, if there are 100 people in the class, the 99th percentile is highest grade. If there are 300 people, the 99th percentile is the average of the top three grades.

Cut points are determined from this grade. For example, if the 99th percentile is a 100, then an A= 100 *0.9= 90%, a B=100 *0.8=80%, and so on.

If the 99th percentile is a 93, then an A=93 * 0.9=83.7, a B=93 * 0.8=74.4, and so on.

Realistically, in a college class, there’s probably at least one or two people who will get 98-100 percent on everything, so it’s likely the grading scale will be the common intervals, but in case the class is super difficult, this will adjust everyone’s grades up a bit.

Rather than assigning grades based on how many points you earned out of the available points, grades are assigned based on how you rank against the rest of your class. The people with the highest 1% of points earned determines what is considered the “maximum available points”. Grading is then done as usual, just using this adjusted maximum rather than the actual maximum.

The idea is that, if the assignments were so hard that *no one* got a perfect score, then the points needed to qualify for a particular grade are shifted down to compensate. This prevents things like flunking the whole class because no one got over 59% of the points that were available. If a situation like that happens, it’s more likely the fault of the instructor than the students, so this system adjusts the grading to favor the students.

The 99th percentile is the value that is greater than 99% of the values in the set. So, if there are 100 students, the 99th percentile is the score that 99 students did worse than and 1 student did better than.