How a Grading Curve Changes Your Score

"Curved" means at least four different things, and on the same raw score they can differ by fifteen points. Here's what each one does to your mark.

"Curved" means at least four different things

Students talk about a curve as if it were one procedure. It isn't, and the methods disagree with each other by a lot.

A raw 72%, in a class averaging 64% with a top score of 88%:

Method	Your score
Flat +5 points	77.0
Scaled to the top score	81.8
Square-root curve	84.9
Mean shifted to 75	83.0

Nearly eight points between the least and most generous, on identical inputs. Which one you get is a decision your instructor makes, and often makes after marking.

Your details

Your raw score(%)

Highest score in the class(%)

For top-score scaling.

Class average(%)

For the mean shift.

Target average(%)

Where the instructor wants the average to land.

Flat points added(points)

Result

With a flat adjustment

The most common method — everyone gets the same points.

Scaled to the top scoreThe highest mark becomes 100 and everything scales.: 81.8
Square-root curveLifts low scores much more than high ones.: 84.9
Shifted to the target averageEvery score moves by the same amount.: 83
Points above the class averageWhere you sit in the distribution.: 8
Best of the fourThe most favourable of the four — not a prediction.: 84.9

Open the Exam Curve Calculator on its own page to bookmark or share it.

Why the square-root curve is so uneven

It takes the square root of your percentage and multiplies by ten. The square root grows fastest near zero, so the lift is dramatic at the bottom and modest at the top.

Raw	Curved	Lift
36	60.0	+24
49	70.0	+21
64	80.0	+16
81	90.0	+9

That's why instructors reach for it after an exam that went badly across the whole class — it rescues the bottom without pushing strong students past 100.

The flip side: if you scored well, it does comparatively little, and a modest flat adjustment can beat it.

Top-score scaling depends entirely on one person

Scaling to the top score treats the highest mark as 100 and scales everything proportionally.

Its behaviour is odd because it hangs on a single student. If the top score is 88, everyone gets a 13.6% relative lift. If one person scored 100, the method does nothing at all — for anybody.

So it's generous exactly when the exam defeated everyone, and inert the moment one student cracks it. Worth knowing before you hope for it.

A mean shift can move you down

The other three methods here never reduce a score. A mean shift can.

If the class average came in at 78 and the instructor targets 75, everyone loses three points. That's unusual, but it happens on an exam that turned out easier than intended, and it's why the mean shift is shown with its sign preserved rather than floored at your raw mark.

Everything is capped at 100

A mark above full marks isn't a mark, and almost no institution records one.

That cap is also why top-score scaling and the square-root curve often show the same figure for students near the top — both would push past 100, and both get clipped there.

What this can't tell you

Which method you'll get. There's no way to infer it from your score. Many instructors decide after seeing the distribution; some curve the final course grade rather than the exam; plenty don't curve at all.

True norming. If your course forces set percentages into each letter grade, your result depends on your rank among every student, not your percentage. That needs the whole distribution — a mean and a top score can't produce it, so it isn't modelled here.

The useful output is the range. Four plausible methods spanning eight points tells you how much genuine uncertainty sits between your raw mark and your final one — and that a study decision made on an assumed curve is a decision made on a guess.

Find out whether a curve is even promised

Before any of this matters, check the syllabus for what it actually commits to.

Most syllabi say nothing about curving, which means the instructor may curve, may not, and owes you neither. Some state that final grades "may be adjusted at the instructor's discretion" — that's a reservation of the right, not a promise. A minority commit to something specific, and those are the only ones you can plan around.

Also check what gets curved. An exam curve and a course-grade curve are different operations with different effects: a generous exam curve on a component worth 20% moves your final mark by a fifth of whatever it gave you.

Why your position in the distribution matters

Three of the four methods here treat everyone identically. One doesn't.

A flat adjustment adds the same points to every student, preserving the gaps exactly. Top-score scaling multiplies everyone by the same factor. The square-root curve helps low scores far more than high ones.

A mean shift moves everyone equally too — but whether that's good news depends on where the class average landed relative to the target. That's why the tool shows your points above or below the class mean: it's the single number that tells you whether a distribution-based adjustment is likely to work in your favour.

If you're well above the mean, the square-root curve is probably your worst outcome and a flat adjustment your best. Well below, and it reverses.

Don't make study decisions on an assumed curve

This is the practical point.

Deciding how much to revise for a final on the assumption that a curve will cover a weak midterm is a decision made on a number nobody has produced yet. The four methods here span eight points on a single set of inputs — and that's assuming a curve happens at all.

Use the spread as a picture of the uncertainty. If passing requires the most generous of four hypothetical curves, you don't have a curve problem, you have a margin problem, and that's worth knowing while there's still time to do something about it.