Grading on a Curve

Grading on a curve means turning raw test scores into grades by comparing each student with the rest of the class, instead of using fixed cutoffs like 90 for an A. A score of 78 can be a C+ on one curve and a B+ on another, because the method — not just the score — decides the grade. This guide explains all six ways teachers curve grades, works the numbers step by step, and shows when a curve helps and when it hurts. Try the numbers yourself with the Bell Curve Grading Calculator.

Short answer: the most common method is bell-curve z-rescaling: turn each score into a z-score, z = (score − class mean) / class SD, then stretch it onto a target mean and SD with curved score = z × target SD + target mean. A class with mean 70 and SD 10 curved to a target mean of 75 turns a raw 85 into a curved 90 — an A.

Grading on a curve is a norm-referenced way of assigning grades: a student’s grade depends on where their score falls in the class distribution, not on how many questions they answered correctly. If the exam was hard and the class mean was 68, a 78 can still be a B. On a fixed (criterion-referenced) scale, 78 is a C+ no matter what anyone else scored.

Teachers curve for two reasons. First, to correct a test that turned out harder — or easier — than planned. Second, to spread grades out when scores bunch together, so the best students can be told apart from the average ones. The curve never changes the order of the class: the top scorer stays on top. What it changes is the spacing between scores and the letter each score earns.

That is also why a curve feels unfair to some students: your grade depends partly on your classmates. The fairness debate is covered below, but the math is neutral — a curve is a rescaling, and its effect depends entirely on the method chosen.

Norm-Referenced vs Criterion-Referenced Grading

The split that matters most is between norm-referenced and criterion-referenced grading.

Norm-referenced grading (grading on a curve) compares students with each other. Grades are assigned from ranks, percentiles or z-scores. A fixed share of the class gets each letter, so a strong class can still produce Fs and a weak class can still produce As.

Criterion-referenced grading compares each student with a fixed standard. A driving test, a medical licensing exam or a pass mark of 70% works this way: everyone who clears the bar passes, even if the whole class passes. Nobody fails because someone else did better.

Most school and college courses mix the two: a fixed scale most of the time, with a curve applied when an exam misfires. Knowing which system is in play matters more than the raw score — the same 78% is a C+ on a fixed scale, a B on a bell curve and a B+ on a square-root curve, as the table below shows.

How to Grade on a Curve

There are six standard methods. They differ in what they preserve: the bell-curve method preserves each student’s distance from the mean; the flat scale preserves the gaps between students; the square-root curve helps low scores most; forced distribution fixes the share of each grade in advance. Pick the method first — the numbers follow from it.

1. Bell-Curve Z-Rescaling

The bell-curve method is the one our free grading calculator uses, and it is the fairest when scores are roughly normal. It converts every raw score into a z-score, then rescales the whole class onto a chosen target mean and standard deviation. First find the class mean and standard deviation — see how to work out the standard deviation for the full steps — then check any score with our z-score calculator.

  1. Step 1 — Find the class mean and standard deviation. Add up all the scores and divide by the number of students for the mean, then work out the spread of the scores around it.
  2. Step 2 — Turn each raw score into a z-score. Use z = (score − class mean) / class SD. A z-score of +1.50 means the score sits 1.5 standard deviations above the mean.
  3. Step 3 — Pick a target mean and target SD. A common choice is a target mean of 75 with a target SD of 10, which centres the class on a C+ to B− average.
  4. Step 4 — Rescale each score. Use curved score = z × target SD + target mean. A raw 85 in a class with mean 70 and SD 10 becomes z = +1.50, then 1.50 × 10 + 75 = 90.
  5. Step 5 — Read the grade from the z-cutoff table. A z-score of +1.28 or higher earns an A, +0.52 to +1.28 earns a B, and so on.

Because z-scores keep each student’s relative position, this method never swaps two students’ order. It only moves the mean and stretches or squeezes the spread.

Two bell curves showing grading on a curve: original scores with mean 70 and curved scores with mean 75, a score of 85 becomes 90

2. Flat Scale: Add Points

Add the same number of points to every score: curved score = raw score + points, capped at 100. If the class mean was 68 and the teacher adds 10, a 78 becomes 88 and a 55 becomes 65. Gaps between students stay exactly the same, so this method is simple and transparent — but it cannot fix a bunched-up distribution, and it hands the same bonus to the top scorer as to the student who needs it most.

3. Square-Root Curve

Take the square root of the score as a decimal and multiply by 100: curved score = √(score / 100) × 100. A 78 becomes √0.78 × 100 = 88.3, while a 49 becomes √0.49 × 100 = 70. Low scores gain the most — a 25 jumps to 50 — and high scores barely move, which is why this method is popular after a very hard test. Above 90 the boost almost disappears, so it rewards the students who struggled without inflating the top.

4. Linear Scale to the Top Score

Scale every score so the highest score becomes 100: curved score = raw score × (100 / top score). With a top score of 92, a 78 becomes 78 × (100 / 92) = 84.8 and a 46 becomes exactly 50. The whole class stretches by the same factor, so the order stays the same and the spread widens. The weakness is obvious: one unusually high score decides the curve for everyone else.

5. Forced Distribution

Fix the share of each grade in advance — usually 10% A, 20% B, 40% C, 20% D and 10% F — then rank the class and cut it into those bands. The raw score does not matter at all: a 78 is an A if it lands in the top tenth and a C if it lands in the middle. This is the method behind the forced-distribution table below, and it is the strictest form of grading on a curve.

6. Shift the Mean

Move the class mean to a chosen value without changing the spread: curved score = raw score + (target mean − class mean). To move a class with mean 68 to a target mean of 75, add 7 points to every score. A 78 becomes 85 and a 60 becomes 67. It is a flat scale with the size of the shift set by the mean instead of by the teacher’s choice — simple, but like the flat scale it leaves the shape of the distribution untouched.

The grading scale below is the one the Bell Curve Grading Calculator uses. It fixes the share of each grade and the z-score that marks each boundary. The raw cutoffs are for the worked example class in this guide (mean 70, SD 10).

GradeShare of classZ-score rangeRaw score (mean 70, SD 10)
ATop 10%z ≥ +1.2882.8 or higher
BNext 20%+0.52 to +1.2875.2 to 82.8
CMiddle 40%−0.52 to +0.5264.8 to 75.2
DNext 20%−1.28 to −0.5257.2 to 64.8
FBottom 10%z < −1.28Below 57.2

The cutoffs come from the standard normal distribution: about 10% of a normal curve lies above z = +1.28, and about 70% lies below z = +0.52. You can check either value by reading a z-table or with the normal distribution calculator.

Forced distribution uses the same shares, but applies them to ranks instead of z-scores:

GradeShareStudents in a class of 30Students in a class of 100
A10%310
B20%620
C40%1240
D20%620
F10%310

In a class of 30, the top 3 students get As and the bottom 3 get Fs, whatever their raw scores. Small classes make this lumpy — with 12 students, 10% is a single student — which is one reason not to force a distribution on a small group.

Grading on a Curve Example

Here is the bell-curve method worked in full, using the same numbers the Bell Curve Grading Calculator shows on its page: class mean 70, SD 10, target mean 75, target SD 10.

Raw scoreZ-scoreCurved scorePercentileGrade
85+1.5090.0093.32ndA
700.0075.0050thC
60−1.0065.0015.87thD
55−1.5060.006.68thF

Take the 85: z = (85 − 70) / 10 = +1.50, so the curved score is 1.50 × 10 + 75 = 90.00. A z-score of +1.50 sits at the 93.32nd percentile, and because +1.50 is above the +1.28 cutoff, it earns an A. The class mean itself, 70, curves to exactly 75 and a C — the centre of the new scale. A 60 lands one standard deviation below the mean, curves to 65 and a D; a 55 curves to 60 and an F. To see the same steps on any numbers, plot the z-scores on a bell curve.

A second example comes from a class with mean 64 and SD 11, curved to a target mean of 76 and SD 10. A score of 75 has z = (75 − 64) / 11 = +1.00, so it curves to 76 + 1.00 × 10 = 86. A score of 53 has z = −1.00 and curves to 66, and the mean score of 64 curves to exactly 76. The spread shrinks slightly (from 11 to 10) while the mean rises by 12 points — the signature of a gentle upward curve.

Same Student, Five Methods

One student, one score of 78%, in a class with mean 68, SD 10 and a top score of 92. Watch what each method does with the same 78:

MethodWorkingCurved scoreGrade*
Bell-curve z-rescaling (target 75, SD 10)z = +1.00 → 75 + 1.00 × 1085B
Flat scale (+10 points)78 + 1088B+
Square-root curve√0.78 × 10088.3B+
Linear scale to top (92 → 100)78 × (100 / 92)84.8B
Forced distribution (top 10% A)Rank, not raw scoreDepends on rankDepends on rank

*Grades use a standard 90–100 A, 80–89 B scale. The same student lands between a B and a B+ depending on the method — a 3.5-point swing from one choice alone. Forced distribution is the odd one out: it ignores the 78 entirely. In a class of 25, a student ranked 6th sits in the top 24% and earns a B, even with the same raw score that a flat scale would lift to 88.

Grading on a Curve Formula

The bell-curve formula has two steps. First the z-score: z = (score − class mean) / class SD. Then the rescale: curved score = z × target SD + target mean. Together: curved score = ((score − class mean) / class SD) × target SD + target mean.

The other methods each have one line: flat scale, curved = score + points (capped at 100); square-root, curved = √(score / 100) × 100; linear scale, curved = score × (100 / top score); shift the mean, curved = score + (target mean − class mean); forced distribution, grade from rank and fixed percentages.

If you only need one number, start with the z-score: it tells you the student’s position in the class before any target is chosen. Our guide to how to calculate a z-score works the same 85, 70 and 10 example from the start.

Excel (and Google Sheets) can curve a whole class in one column. Put the raw scores in column A, the class mean in B1, the class SD in B2, the target mean in B3 and the target SD in B4.

MethodExcel formula (score in A2)
Bell-curve z-rescaling=(A2-$B$1)/$B$2*$B$4+$B$3
Flat scale (+10)=MIN(A2+10, 100)
Square-root curve=SQRT(A2/100)*100
Linear scale to top (top in B5)=A2*(100/$B$5)
Shift the mean=A2+($B$3-$B$1)

Check the bell-curve formula against the worked example: =(85−70)/10*10+75 returns 90, exactly the curved score for the 85. For the class statistics themselves, use =AVERAGE(A:A) for the mean and =STDEV.P(A:A) for the population standard deviation — or =STDEV.S(A:A) if you treat the class as a sample. The standard deviation calculator will give you both, and our guide to the standard normal distribution explains the z-values behind the cutoffs.

Is Grading on a Curve Fair?

It depends on the method and the class. A curve is fair when it corrects a bad test: if an exam was too hard and the class mean came out at 62, rescaling to a target mean of 75 simply repairs a mis-calibrated test, and nobody’s relative position changes. It is also fair when the class mean is low because the material was new and hard — the curve recognises that the raw numbers understate what students learned.

It feels unfair when the method caps grades. Under a forced distribution, the bottom 10% must fail even if every student passed the raw test, and the top 10% get As even in a weak class. It is unfair when a curve pushes scores down: if the class mean is already 85, curving to a target mean of 75 lowers almost everyone. And in a tiny class, one or two students decide the mean and the spread, so the curve is more luck than measurement.

Most of the controversy is about forced distribution, not about rescaling. Adding a few points to a class that sat a hard exam rarely draws complaints; guaranteeing failures does.

When NOT to Curve Grades

Do not curve a small class. With fewer than 15 students, the mean and standard deviation are unstable — one strong or weak student moves the whole curve — and a forced distribution becomes meaningless: in a class of 12, 10% is a single student.

Do not curve criterion-referenced exams. A driving test or a medical licensing exam measures whether each student meets a fixed standard; curving it would pass unsafe drivers or fail competent ones for no reason. If everyone clears the bar, everyone passes.

Do not curve when the class mean is already high. Applying a bell curve to a class averaging 85 forces the mean down to the target, so students lose points for the class doing well.

Do not curve a bimodal distribution. If the scores cluster in two humps — for example, students who attended a review session and students who did not — a single mean and SD describe nobody, and the curve will misplace the whole middle. Fix the test, offer a re-take, or use a small flat scale instead.

Grading on a Curve vs Flat Scale

A flat scale adds the same number of points to every score: 78 becomes 88, 55 becomes 65. The order stays the same and the gaps between students stay exactly the same, so it is simple, transparent and easy to explain. Its weakness is that it cannot fix a bunched-up distribution or change the spread.

Bell-curve rescaling re-centres the whole class on a target mean and standard deviation: an 85 becomes 90 when the class mean is 70 and the target is 75, while a 55 becomes 60. It can stretch or squeeze the spread, so it is the fairer choice when the test was mis-calibrated — but the result depends on the target mean and SD, and a bad target makes things worse. In practice, teachers reach for the flat scale after a slightly hard test and for the bell curve after a badly misjudged one.

Can a Curve Lower Your Grade?

Yes. If the class mean is already high, curving down lowers scores: a class averaging 85 that is curved to a target mean of 75 loses 10 points on average. Under a forced distribution, the bottom 10% must fail even if they passed the raw test. Always check whether your teacher curves up only, or applies the full bell curve in both directions.

What Is a Good Score After Curving?

In the example class in this guide (mean 70, SD 10, target 75/10), a curved score of 75.2 or above earns a B, and 82.8 or above earns an A. More generally, a good curved score is one at least half a standard deviation above the target mean — a z-score of +0.52 or higher. Anything above the target mean is above average; anything below it is below average, whatever the raw score was.

Do Colleges Use Grading on a Curve?

Many do, especially in large introductory courses such as chemistry, economics and law, where exams are hard and enrolments are big enough for the distribution to be stable. Law schools are the best-known example: many publish a target median grade for first-year courses. Policies vary by department, though, and some schools ban forced curves outright, so the only reliable answer is the one in your course syllabus.

Frequently Asked Questions

What is grading on a curve?

Grading on a curve adjusts raw scores against the class distribution instead of a fixed scale. A common method rescales scores to a target mean and standard deviation using z-scores.

How do you calculate a curved grade?

Convert the score to a z-score (score − class mean) ÷ class SD, then curved = z × target SD + target mean. Example: (85 − 70) ÷ 10 = +1.50, so 1.50 × 10 + 75 = 90.

What is a bell curve grading scale?

Grades are assigned by rank: top 10% A (z ≥ +1.28), next 20% B (+0.52 to +1.28), middle 40% C (−0.52 to +0.52), next 20% D (−1.28 to −0.52), bottom 10% F (z < −1.28).

How do you curve grades in Excel?

Use =(score − class mean) / class SD * target SD + target mean; e.g. =(85−70)/10*10+75 = 90.

Can grading on a curve hurt your grade?

Yes. If the class mean is high, the curve can push scores down, and a forced distribution caps how many students can earn each grade.

What is the difference between norm-referenced and criterion-referenced grading?

Norm-referenced grading compares students to each other (the curve); criterion-referenced grading compares each student to a fixed standard (for example, 90% = A) regardless of classmates.

Is grading on a curve the same as a flat scale?

No. A flat scale adds the same points to everyone (78 → 88); a bell-curve rescale re-centres the whole distribution on a target mean and SD (85 → 90) and can change the gaps between students.

How do I find my curved grade?

Use the Bell Curve Grading Calculator on this site: enter your score, the class mean and SD, and your target mean and SD.

For the standard normal distribution tables used in these calculations, see the NIST/SEMATECH e-Handbook.

Check Tools

Advertisement
Advertisement

Check Articles

Advertisement