eli5: How do classes that mark on a curve work?

In: 1

Grading on a curve is an inexact term, but in general it means adjusting the scores such that students fall into a certain preset pattern of results rather than taking the raw scores as absolute values.

For example, lets say I go with a traditional scoring system as follows:

* 90-100% = A

* 80-89% = B

* 70-79% = C

* 60-69% = D

* 59% or less = F

I have my students take a test with 100 possible points, so its easy to find out what percent they have. Lets say the best student in the class only gets 70 points. Its possible the students just aren’t that good. Its also possible I made the test too hard. So to adjust for that I “curve” the results so that 70 points is the new top score. That means 35 would be 50% (and an F). This basically boosts everyones results.

However, that doesn’t mean its the correct way to grade, its simply an option. And its not the only way to “curve” a result. In fact the original meaning (and where you get the term curve) comes from a different way of adjusting results.

The original meaning uses a “normal distribution” (here normal has a mathematical meaning) sometimes called a bell curve because of its shape. A normal distribution is a formula such that values are distributed in a certain way. The exact definition uses some math thats a bit above ELI5, but the main idea is that half the results should be above the average, and half should be below. In addition most results should be close to average, with fewer results far from average (either above or below). Its also a symmetrical distribution, meaning that it looks the same on either side of the average. The shape is also called a normal curve, because of the curved line that represents it when graphed.

To grade on a curve in that case means that exactly half your students should have results above average, and half below, and very few should get the highest grade. Using a normal curve and standard deviations (don’t worry about the meaning of that) you’d end up with the following distribution of results:

* 2% of the students would get A’s

* 14% of the students would get B’s

* 68% of the students would get C’s

* 14% of the students would get D’s

* 2% of the students would get F’s.

The idea is that MOST people should be at or close to average. But what if its a small class, say 20 students and they are all very smart? Why should some of them get below average scores if they ALL know the material well?

Grading on a curve should be used appropriately depending on what is being measured and how it is being measured. There is no one “right” way to do it (if you should even do it at all!).

There are several ways classes can be curved, and the expression refers to the infamous bell curve shape of regularly distributed data. This means that there should a large clump in the middle that quickly tapers off in both directions. The average should be 80, and most people should be 70-90 with as many getting 95 as got 65. If you look at your class average and it’s clumped around 70 instead of 80, you have to shift the grades up by ten points to have the “curve” of the grades centered on a normal number

Smartest guy sets your new 100% (A grade)

Dumbest guy always gets an F

The average skill of everyone else determines how easy it is to move between letter grades

Essentially, you are graded in competition to your peers

This feels fair for hard classes where everyone eats shit, but tends to feel unfair for easier classes sometimes (especially in weeder classes)

There are different ways to do it, but basically it means grading students relative to each other.

The simplest is to just say that whatever the top score in the class is, you then say that’s 100% and grade everyone based on that. So if you have 1000 possible points in a class and the best student gets 800, you grade everyone out of 800.

There’s also more complicated ways to do it, where you have buckets and you determine how many students ought to be in each bucket based on the expected distribution of scores, so the top 10 students get an A, next 20 get a B, next 20 get a C, etc etc you can get very into the weeds with it.

If I am understanding you’re question correctly this is how a curve works:

Let’s say the highest score you can get on an exam or in a class is 100%. So the scores ranges from 0-100. Based off that let’s say that 60% or 60 out of 100 is the lowest passing score.

Now, after the class or exam is over lets say the highest score in the class was 90% or 90 out of 100. That 90 becomes the new 100 or highest score possible. That would shift everything by roughly a 10 point curve. I say roughly because the numbers won’t be perfectly by 10 when calculating percentages. Now making the lowest passing score 50% or 50 out of 90.

So if the test was out of 100 and you scored and 80% that would be a B, but if the curve brought the points down to 90; You’d now have 80 out of 90 which is 88% instead of the original 80%.