The exam comes back and the class average is a 64. Now what? Curving is one of the most misunderstood moves in teaching, partly because "curve" means four different things. This guide runs through each method with the same set of scores so you can see exactly what every approach does, then covers the fairness rules that keep students on your side. The running example: five students scoring 58, 72, 66, 44, and 80, with a mean of 64 and a standard deviation of 12.33.
This is the "real" bell-curve adjustment. You pick where the class should center and how spread out it should be, then rescale every score to match. The formula:
curved = target mean + z × target SD, where z = (raw score − class mean) ÷ class SD
With a target of 75 ± 10, each score shifts by (x − 64) × (10 ÷ 12.33). The class mean of 64 lands exactly on 75. The 44, sitting 1.62 SDs below average, becomes 58.8. The 80, at 1.30 SDs above, becomes 88.0. You control the whole distribution with two dials, which is why departments that require standardized course averages use this method.
If your class mean is already above the target, students lose points. A class averaging 82 curved down to a 75 target drops an 80 to about 73 (assuming SD 10). It's mathematically sound and politically terrible. Rule of thumb: linear curves should move the class average up or leave it, and no individual student should end below their raw score because of the curve alone.
Replace each score with 10 × √score, capped at 100. It's blunt, it's fast, and it rescues the students who need rescuing most. The gains run backwards: the lowest scores get the biggest lift.
| Raw score | 25 | 36 | 49 | 64 | 81 | 100 |
|---|---|---|---|---|---|---|
| Curved | 50 | 60 | 70 | 80 | 90 | 100 |
| Gain | +25 | +24 | +21 | +16 | +9 | 0 |
On the running example, the 44 becomes 66.3 and the 80 becomes 89.4. Notice what didn't happen: nobody's score fell, and nobody crossed the student above them. Perfect squares make the mental math easy, which is a genuine perk when a student asks "how did you get my grade?"
Add a fixed number of points to every score. It's the honest response to a broken exam: if question 7 was unfair, add back its point value and everyone is made whole in proportion to what they knew. On the example, +10 points moves the 44 to 54 and the 80 to 90. Use it when the exam itself had a specific defect, not as a mood-lightener, and watch the top: a bonus can push scores past 100, so decide in advance whether you cap.
Divide every score by the class high and multiply by 100. The top scorer gets a perfect 100 and everyone else is measured against them. On the example, the high is 80, so the 44 becomes 55 and the 64 becomes 80. It works when the strongest student genuinely demonstrated what "full mastery" looks like. If that top score is an outlier, one freak performance inflates the whole curve's difficulty, and the class pays for it.
Paste scores, pick a target mean and SD, and see linear and square-root curves side by side for every student.
Open the Grade Curve Calculator →| Raw | Linear 75±10 | √ curve | Flat +10 | Scale to max |
|---|---|---|---|---|
| 44 | 58.8 | 66.3 | 54 | 55.0 |
| 58 | 70.1 | 76.2 | 68 | 72.5 |
| 66 | 76.6 | 81.2 | 76 | 82.5 |
| 72 | 81.5 | 84.9 | 82 | 90.0 |
| 80 | 88.0 | 89.4 | 90 | 100.0 |
Same class, five different grade distributions. The linear curve is the most conservative, the square-root curve the most generous to the bottom, and scale-to-max the only one that guarantees a 100.
A curve fixes one exam; it can't fix a semester. If you're rebuilding final grades after curving several components, the GPA calculator converts the results to grade points, and the weighted GPA calculator handles the honors and AP bumps on a 5.0 scale. For standardizing scores across sections, the standard deviation calculator computes the spread you'll need for the linear method, and the GPA scale converter maps letter grades back and forth between percentage and 4.0 systems.
If the whole class bombed it, the square-root curve (10 times the square root of the score) lifts the bottom the most: a 25 becomes 50 and a 36 becomes 60, while a 90 stays near 90. If the exam was fine but one section was impossible, a flat bonus of the section's point value is fairer because it moves everyone equally.
No, not under any of the standard methods. Linear rescaling, square-root curves, flat bonuses, and scale-to-max are all strictly increasing functions, so a higher raw score always stays higher after curving. What changes is the distance between students, which is exactly what the target SD controls.
A target mean of 75 with an SD of 10 is the classic default: the class centers on a C+ and roughly two-thirds of scores land between 65 and 85. If your course historically averages 80, match that. Keep the target SD at or below the class SD to tighten grade gaps, or above it to spread a crushed distribution back out.