Grade curve calculator

What "curving" a test can mean

Teachers use the word for several different things, and the calculator above offers six. Some raise every score by the same amount, some stretch the scores so the best result becomes full marks, some move the whole class to a chosen average and spread, and some replace each score with the one that has the same rank in a bell-shaped distribution. All are only arithmetic on your list of scores; none is a standard, and your syllabus or department rules decide what is allowed. Paste the raw scores, set the highest possible score, choose a method, and compare the class average, the spread and the letter counts before and after.

The methods and their formulas

Add points: new score = score + p. Everyone moves by the same amount, so the order and the spread do not change. Top score to maximum: the same idea with p = maximum − highest score, so the best result becomes full marks. Scale to maximum: new score = score × maximum ÷ highest score; the best result becomes full marks and everyone else rises in proportion, so weaker scores gain fewer points than a flat shift would give them, and the spread grows by the same factor.

Square root: new score = maximum × √(score ÷ maximum), for a 100-point test 10 × √score. It lifts low scores most (a 64 becomes 80 and a 36 becomes 60) and leaves 0 and the maximum where they are. New mean and standard deviation: new score = target mean + (score − class mean) × target SD ÷ class SD. This is a z-score rescaling; it keeps the shape of the distribution and the order, and it lowers the scores of anyone far above the class if the target mean is below theirs. Normal curve by rank: each score is replaced by target mean + target SD × z, where z is the point of the standard normal curve at the cumulative fraction (rank − 0.5) ÷ n; tied scores share the average rank. It forces a bell shape, so a skewed class becomes symmetric; the order of the scores does not change, but the gaps between neighbours become artificial.

The class mean, median and standard deviation shown are plain descriptive statistics of your list; the standard deviation divides by the number of scores, because the class is the whole group being graded. Results above the maximum or below zero are held at the limit and counted, so you can see when a method over-reaches.

A worked example with five scores

Take five scores on a 100-point test: 48, 55, 62, 71, 88 (class average 64.8, standard deviation 13.9). The table shows what each method does to them; every number comes from the same functions the calculator uses.

Adding 5 points keeps the gaps; scaling to 100 stretches them, so the weakest score gains 6.5 points and the best 12; the square root gives the weakest score 21.3 points and the best 5.8. The new-mean-and-SD rescale lands on an average of 75 and a spread of 10 by construction. The normal-curve version also averages 75, but with only five scores its spread is 8.8, and it replaces the real gaps between the five scores with evenly spaced positions on the bell curve.

Pros, cons and limits

A curve is a form of norm-referenced grading, where a grade says how a student ranks in the class, as opposed to criterion-referenced grading, where it says how much of a defined body of material the student has mastered. The University of Michigan's Center for Research on Learning and Teaching (Piontek, 2008) lists the arguments for the norm-referenced approach: it allows flexibility with groups that are unusually strong or weak, it lets an instructor correct for an exam that turned out much harder or easier than intended, it can restrain grade inflation, and it identifies students who stand out. The same paper lists the costs: students know that some of them must get low grades, which can hurt motivation and cooperation, and it is hard for a student to judge their standing during the term because the final grade depends on everyone else's results. It also notes that in practice instructors rarely use a strict bell curve, for example one that gives 35% of the class a D or F.

Some limits are mathematical. A statistical curve needs a reasonable number of scores: with ten students one unusual result changes the class mean and spread for everyone. Methods based on the top score depend on one student; a single outlier can make a flat shift or a scaling nearly worthless or very generous. Letter cutoffs are your own choice, here 90, 80, 70 and 60 percent of the maximum as a common default; a different scale, or plus and minus grades, will give different counts. Finally, a curve cannot fix a test that did not measure what you taught; if most of the class missed the same question, the question is the first thing to check.

How do I curve grades?

Pick the idea first. If the test was simply too hard, adding points or lifting the top score to full marks is the easiest to explain to students. If you want a class average of, say, 75 with a spread you choose, rescale to a new mean and standard deviation. If you want letter counts to follow a bell shape, use the normal-curve method, but expect the strongest objections from students there. In every case, check the before-and-after table and tell students the rule before you apply it.

What is the square root curve?

The score is replaced by the maximum times the square root of the fraction earned: 10 × √score on a 100-point test. With the example above, 48 becomes 69.3 and 88 becomes 93.8. It helps low scores most, never lowers a score that is within the maximum, and keeps 0 and 100 fixed. It is a custom among instructors, not a published standard.

Can a curve lower my grade?

Adding a positive number of points, lifting the top score, scaling to the maximum and the square root can only raise or keep a score (as long as no score is above the maximum). A rescale to a new mean and standard deviation, and the normal-curve method, can lower scores of students who were well above the target average, because they place a student by position, not by points earned. Check which method your instructor uses.

Is grading on a curve fair?

There is no single answer, and teaching centers describe both sides. Norm-referenced grading, which includes curves, makes sense when you must rank students or correct a badly calibrated exam; critics point out its effect on cooperation and that a student cannot tell where they stand until others are graded. Many courses now state fixed grade cutoffs in the syllabus instead and curve only to repair a test that was clearly off.

What if the highest score is already 100?

Then "top score to maximum" and "scale to maximum" change nothing, because the top score is already full marks. Use a flat addition, the square root, or a rescale to a new mean and standard deviation instead.

Why do some scores stop at 100?

Most grading policies cap a test at the maximum, so the calculator holds any result above it at the maximum and tells you how many scores were held. If your course allows scores above the maximum, enter a higher maximum and divide later.

Sources, checked October 27, 2026: M. E. Piontek, "Best Practices for Designing and Grading Exams", CRLT Occasional Paper No. 24, University of Michigan Center for Research on Learning and Teaching, 2008, section "Normative and Criterion Grading Systems". The formulas for the flat, scaled and square root curves are common classroom practice with no official standard. Related: the grade calculator, final grade calculator, GPA calculator, average calculator and normal distribution calculator. How we check our calculators.