Bell Curve Grade Calculator
Paste a class and this bell curve grade calculator works out the score cutoff for every letter, how many students land in each band, and how the real class compares with the normal curve. If you want to change the scores themselves rather than sort them into letters, that is curving the scores instead of banding them.
Free to use, no sign-up, and every calculation happens in your browser.
Paste a class and the bands are worked out from the mean and the standard deviation. You can paste names alongside the scores; only the numbers are read. Nothing here is a rule: your school decides whether grades are banded this way at all.
The cutoff, count and share for each letter
The model column is what a perfectly normal class of this size would give. A real class never matches it exactly.
The class against the normal curve
Every student, with the working
Runs in your browser. Nothing is sent or stored. The bands come from the standard deviation, so the same scores can produce very different letters depending on how wide the bands are set. Whether your course is graded this way is set by your instructor or your department, so treat any letter here as an estimate until the grade is published.
How to use this bell curve grade calculator
Paste the class in, pick how wide you want the bands, and read the cutoffs. If you are a teacher deciding what a banded grading scale would do to a set of results, this is the whole job in one screen. The calculator opens on a class of twenty-four, so there is a worked result on screen before you type anything.
- Put the scores in. Paste them straight from your roster or gradebook, separated by commas, spaces or new lines. Anything that is not a number is ignored.
- Pick a band scheme. Classic and Half SD set their own boundaries. Custom lets you move each one.
- Read the cutoffs. The table gives the score at which each letter grade starts, how many students landed there, and what a perfectly normal class of that size would have given instead.
Underneath, the chart draws your class as a histogram against the normal curve, and the last table shows every student with the z-score and percentile behind their letter.
What a bell curve actually decides
A bell curve does not rank students on points. It ranks them on distance from the class average, measured in standard deviations. That distance is the z-score, also called a standard score, and it is the only thing the letters are built from.
A student on the class average has a z-score of zero. One standard deviation above it is z = 1, and one below is z = −1. The letters are then assigned by drawing lines across that scale: everything above a certain z gets an A, everything below another gets an F.
This is why two classes with identical scores can produce different letters. If one class is tightly bunched, a score five points above the average might be two standard deviations out and earn an A. In a class where the scores are spread wide, the same five points might be a quarter of a standard deviation and earn a C. The raw number has not changed. The room around it has.
It is also why the calculator asks for the whole class rather than one score. Without the spread there is no z-score, and without a z-score there is no band.
The two band schemes, and why they disagree
Almost every page about bell curve grading picks one set of boundaries and calls it the bell curve. There is more than one, and they are not close. The grade distribution that comes out the other end depends entirely on which one you pick. Both of the schemes below are normal distributions. The difference is only how wide the middle band is.
The classic scheme: 2, 14, 68, 14, 2
The familiar version puts C between one standard deviation either side of the mean, B and D in the next band out, and A and F in the tails.
| Letter | Range | Share of the class | In a class of 100 |
|---|---|---|---|
| A | above +2 SD | 2.28% | 2 students |
| B | +1 to +2 SD | 13.59% | 14 students |
| C | −1 to +1 SD | 68.27% | 68 students |
| D | −2 to −1 SD | 13.59% | 14 students |
| F | below −2 SD | 2.28% | 2 students |
Those five figures total 100% and round to the 2 / 14 / 68 / 14 / 2 split that gets quoted as the bell curve. Two students in a hundred fail, and two get an A.
The half standard deviation scheme
Dave Richeson, a professor of mathematics at Dickinson College, set out a narrower version: make the mean a C, then the mean plus/minus a half standard deviation would be the C-/C/C+ scores, one more standard deviation out would give the B’s and D’s, and the tails would give the A’s and F’s. Half the width, and the shape of the class changes completely.
| Letter | Range | Classic scheme | Half SD scheme |
|---|---|---|---|
| A | top tail | 2.28% | 6.68% |
| B | upper band | 13.59% | 24.17% |
| C | middle | 68.27% | 38.29% |
| D | lower band | 13.59% | 24.17% |
| F | bottom tail | 2.28% | 6.68% |
Moving the middle band from one standard deviation to half of one nearly triples the A’s, from 2.3% to 6.7%, and cuts the C’s from roughly two thirds of the class to under two fifths. Nothing about the students changed. The curve did not decide these grades. The band width did. That is worth knowing before anyone tells you a result came from the bell curve, because on its own that sentence does not say much.
The 68-95-99.7 rule
The band percentages are not conventions. They come from a property of the normal distribution often called the empirical rule, and the calculator uses the exact figures rather than the rounded ones.
| Within | Exact share | Usually quoted as |
|---|---|---|
| ±1 standard deviation | 68.27% | 68% |
| ±2 standard deviations | 95.45% | 95% |
| ±3 standard deviations | 99.73% | 99.7% |
Read the first row as the classic C band and the second as everything between an F and an A. The third row is the reason scores more than three standard deviations from the average are treated as remarkable: in a perfectly normal group, fewer than three students in a thousand land out there.
The rule is symmetric, so each figure splits evenly either side of the average. The tails are what the rule leaves out, and they are small. That is the arithmetic behind a scheme where only two students in a hundred earn the top grade.
A real class of twenty-four, banded
These are the scores the calculator starts with: 41, 52, 55, 58, 61, 63, 64, 66, 67, 68, 69, 70, 71, 72, 73, 74, 76, 77, 79, 81, 84, 87, 91 and 96. The class average is 70.63 and the standard deviation is 12.31.
| Letter | Score from | Students | Share | Normal model |
|---|---|---|---|---|
| A | 95.3 and up | 1 | 4.2% | 2.3% |
| B | 82.9 and up | 3 | 12.5% | 13.6% |
| C | 58.3 and up | 16 | 66.7% | 68.3% |
| D | 46.0 and up | 3 | 12.5% | 13.6% |
| F | below 46.0 | 1 | 4.2% | 2.3% |
The middle of the table is close to the model. The tails are not: one A is 4.2% of twenty-four students where the model asks for 2.3%, and the same goes for the single F. That is not an error. In a class of twenty-four, one student is worth 4.2 percentage points, so the tails can only ever land on a multiple of that. A real class cannot match the curve exactly, and the smaller it is, the further off it will be.
Switch the same twenty-four scores to the half standard deviation scheme, and the picture changes again: two A’s instead of one, five B’s instead of three, and ten C’s instead of sixteen. Same students, same test, different boundaries.
When a bell curve is meaningless
Every band on this page hangs off the standard deviation, so anything that makes the standard deviation unstable makes the letters unstable with it. Four situations do that.
The class is small. The standard deviation is an average of how far scores sit from the mean, so in a small group a single outlier drags it a long way. That is a sample size problem: the fewer scores there are, the more each one moves the spread. Here is the same experiment at different class sizes: one student scores 30 points higher than they did, and nothing else changes.
| Students | Standard deviation before | After | Change |
|---|---|---|---|
| 8 | 5.95 | 8.26 | +38.8% |
| 10 | 6.43 | 8.27 | +28.5% |
| 20 | 6.10 | 7.01 | +15.0% |
| 30 | 6.11 | 6.89 | +12.8% |
| 100 | 6.05 | 6.29 | +3.9% |
In a class of eight, one changed score moves the spread by nearly forty percent, and every cutoff moves with it. In a class of a hundred the same change is under four percent. It is also why Richeson puts this kind of grading in large classes and multiple sections rather than small ones, and why the calculator says so on screen when there are fewer than twenty scores. A class with one very low score is skewed rather than normal, and a bell curve laid over it will put boundaries where almost nobody is sitting.
Everyone scored alike. If the spread is zero, there is no distance from the mean to rank on and no band can be worked out. The calculator stops and says so rather than inventing an answer.
The class splits in two. A class can divide into a group that understood the material and a group that did not. A split like that is bimodal, not bell shaped, and the average falls in the gap between the two groups where almost no one actually scored.
Zeros and absences are in the data. A missed test entered as a zero is not a measurement of what that student knows, but it pulls the mean down and pushes the spread up, which moves every boundary for everyone else. Take them out before banding, then decide what to do about them separately.
One quick check covers all four. Compare the class average with the median, the middle score, which barely moves when a single result is extreme. If the two sit far apart the class is not symmetric, and the bands are being set by the shape of one tail rather than by the middle of the class.
Is grading on a bell curve fair?
It depends on what you think a grade is for, and the two views do not reconcile.
Under a bell curve your grade is norm-referenced: it reports where you finished relative to the people who took the same test. Under the more common alternative it is criterion-referenced: it reports how much of the material you showed, measured against a fixed standard, and everyone in the room could in principle earn an A. Put plainly, a criterion-referenced grade is absolute and a norm-referenced grade is relative: one measures you against the material, the other against the room.
Richeson lists the case for it in one line: grades end up with a very predictable distribution, which is exactly what a department wants when several sections of the same course have to line up. He lists the case against just as briefly, calling it ruthless, students competing against classmates, and placing it in large classes or multiple sections when there must be a fixed distribution. He also notes he does not know whether professors still use it, at least in small classes.
When a department fixes the share of each letter in advance, that is a forced distribution, and it is the version most of the criticism is aimed at. The bands on this page are worked out from your class, not imposed on it. That is the tension in a sentence. A curve makes results comparable across sections and caps the damage from an unusually hard exam. It also means a strong student in a strong cohort can be marked down for company they did not choose. Which of those matters more is a policy question for your instructor and your department, not a mathematical one, so the place to settle it is your syllabus. If it describes a fixed distribution or a set number of each grade, you are in a norm-referenced course and this page is modeling something real. If it lists percentage cutoffs, you are not.
Bell curve grading in Excel or Google Sheets
With the scores in column A of a spreadsheet starting at A2, four formulas do everything on this page.
| What you want | Formula |
|---|---|
| Class average | =AVERAGE($A$2:$A$25) |
| Standard deviation | =STDEVP($A$2:$A$25) |
| z-score for one student | =(A2-AVERAGE($A$2:$A$25))/STDEVP($A$2:$A$25) |
| Score at which a band starts | =AVERAGE($A$2:$A$25)+2*STDEVP($A$2:$A$25) |
| Share the model predicts above a z | =1-NORM.S.DIST(2,TRUE) |
Two things to watch. Use STDEVP rather than STDEV when the scores are the whole class rather than a sample of it. That is the population standard deviation, which is what the calculator above uses, and the two return different numbers. And change the 2 in the last two formulas to move a boundary: 1.5 gives the half standard deviation scheme, and any value you like gives your own.
Where to go next
Banding a class and curving it are different jobs, and they usually happen in that order.
- The grade curve calculator, linked at the top of this page, applies seven methods including this one and shows every score before and after.
- The scores you paste here have to come from somewhere, and turning right and wrong answers into a score is the step before this one.
- Every cutoff here is built off one number, and working out the class average is how you get it.
Frequently asked questions
What is bell curve grading?
It is grading on a curve in its strictest form: a way of assigning letter grades by position rather than by points. Each score is converted to a z-score, which is how many standard deviations it sits from the class average, and the letters are given to bands of that scale. A student is not measured against a fixed pass mark but against the rest of the room, which is why the same score can be a B in one class and a D in another.
What percentage of students get each grade on a bell curve?
It depends entirely on where the boundaries are set, which is the part usually left out. Under the classic scheme, with C running one standard deviation either side of the average, the split is 2.28% A, 13.59% B, 68.27% C, 13.59% D and 2.28% F. Under a scheme using half a standard deviation for C it becomes 6.68% A, 24.17% B, 38.29% C, 24.17% D and 6.68% F. Both are bell curves.
Is grading on a bell curve fair?
It is predictable, which is its main defense, and it makes several sections of one course comparable. The objection is that it sets students against each other: Richeson calls it ruthless, with students competing against classmates. A strong student in a strong cohort can be marked down for company they did not choose. Whether that trade is acceptable is a policy question for your department rather than a mathematical one.
What is the 68-95-99.7 rule?
It describes how a normal distribution spreads out. About 68% of values fall within one standard deviation of the mean, about 95% within two, and about 99.7% within three. The exact figures are 68.27%, 95.45% and 99.73%. On this page the first of those is the classic C band, and the gap between the second and 100% is what the A and F tails share.
How many students do you need for a bell curve to work?
There is no official minimum, but the arithmetic shows the problem clearly. In a class of eight, one student scoring 30 points differently moves the standard deviation by 38.8%, and every cutoff moves with it. The same change in a class of 100 moves it 3.9%. Small groups also cannot match the model: with twenty-four students, one student is 4.2% of the class, so a band the model puts at 2.3% has to round to nothing or to double. With fewer than twenty scores, the calculator warns you.
Can a bell curve lower my grade?
Yes. Banding is not the same as adding points. If you scored above the pass mark on a fixed scale but finished below the class average, a norm-referenced band can still place you in a low letter, and a passing score on that fixed scale can end up in a failing band, because your position in the room is what is being graded. It works the other way too: on a hard test where the average is low, a middling raw score can land in the C band or above.
How do I do bell curve grading in Excel?
You need two figures and one formula. Get the class average with AVERAGE and the spread with STDEVP, then compute each z-score as the score minus the average divided by the standard deviation. The score where a band begins is the average plus the boundary times the standard deviation, so the A line under the classic scheme is the average plus two standard deviations. Use STDEVP and not STDEV when you have the whole class.
What is the difference between a bell curve and a regular grade curve?
A regular curve changes the scores. Adding five points to everyone or taking the square root of each score produces new numbers that still get read against your usual scale. A bell curve leaves the scores alone and changes what they mean, sorting students into letters by how far they sit from the class average. One rewrites the scores, the other rewrites the boundaries.