Test Curve Calculator

Raw scores:

Paste one score per line or CSV with label,score, e.g. Avery,84.

Maximum raw score:

Enter possible points before any curve, e.g. 100 or 50.

points

Curve method:

Choose fixed points, target average, scale to top, square root, rescale, or custom.

Points to add:

Enter positive or negative points; 5 adds five raw points to each score.

points

Percentage points to add:

Example: 5 pp adds 5 points on a 100-point test or 2.5 on a 50-point test.

Target class average:

Enter 0-100; values below the current average may lower scores.

Target curved range:

Enter low and high curved scores in points, such as 60 to 100.

to points

Multiplier and offset:

Example: 1.05 + 2 raises an 80 to 86 before caps.

x +

Score cap:

Usually match maximum raw score; lower it to test a stricter ceiling.

points

Rounding mode:

Choose exact for review, or a whole/tenth/hundredth rule for posting.

{{ err }}

{{ warning }}

Grade bands:

One band per line as Label,Minimum%, e.g. A,90 then B,80.

Out-of-range scores:

Exclude removes invalid rows; clamp keeps them at 0 or the maximum.

Display decimals:

Enter 0-3 decimal places for score summaries and table values.

Student	Raw	Curved	Change	Band	Copy
{{ row.label }}	{{ formatScore(row.raw) }}	{{ formatScore(row.curved) }}	{{ formatDelta(row.delta) }}	{{ row.rawBand }} → {{ row.curvedBand }}

Band	Before	After	Net shift	Copy
{{ band.label }}	{{ band.before }}	{{ band.after }}	{{ formatSignedInteger(band.change) }}

Check	Value	Readout	Copy
{{ row.label }}	{{ row.value }}	{{ row.note }}

Embed:

Customize

Include current inputs

Size

Advanced

Width

Height

Aspect ratio

Max height

Collapsible embed

Allow fullscreen

Referrer policy

Sandbox tokens

A hard test can leave an instructor with two different decisions. One is mathematical: what happens to the score distribution if a rule raises, rescales, or reshapes raw points? The other is educational: whether changing the numbers is a fair response to the assessment, the instruction that preceded it, and the grading policy students were given. A curve can make the arithmetic consequences visible, but it cannot justify the policy choice by itself.

Grade curves sit between raw performance and the grading scale used for a class. Raw points record what was earned on the assessment as given. Grade bands turn percentages into labels such as A, B, C, D, and F. A curve changes the score before that band comparison, so the same raw mark can end in a different band depending on the rule, cap, rounding choice, and cutoff definitions.

Common grading curve choices and their practical effects
Curve choice	Typical use	Practical caution
Flat point lift	A known scoring issue affected everyone by about the same amount.	Students near a cutoff may move bands while high scores may hit the cap.
Scale to the top	The highest raw score is treated as the effective full score.	All scores depend on one observed maximum, so a single outlier can change the lift.
Target average	The class mean is moved to a planned percentage.	The rule can lower scores when the target is below the current mean.
Square-root curve	Lower percentages need a larger relative lift than higher percentages.	The gain is non-linear, so the average change does not describe every student well.

Two terms often get mixed together. Criterion-referenced grading compares work with a preset standard, such as 90% for an A. Norm-referenced grading compares students with the rest of the class. Many classroom curves are not full norm-referenced systems, but they still borrow the idea that the score distribution matters. That is why a proposed curve should be checked against both the class average and the individual rows that change most.

Raw scores lifted by a curve, compared with grade bands, and checked against a cap

Fairness depends on the reason for the adjustment. A curve for a miskeyed item, a curve for an unusually difficult assessment, and a curve meant to force a distribution are different decisions. The arithmetic should make the consequences visible before a gradebook is changed, especially for students close to a cutoff or already near the maximum score.

A useful curve review keeps the grading rule, score cap, rounding, and band definitions stable while comparing alternatives. If any one of those pieces changes at the same time as the curve, the final band movement may reflect the changed grading policy rather than the curve formula alone. Clear records also make it easier to explain why one student crossed a cutoff while another did not.

How to Use This Tool:

Start with the score scale and the raw rows, then compare one curve method at a time so the ledger and band counts stay easy to audit.

Paste Raw scores as one score per line or as CSV rows such as Student,Score. If a header names a score-like column, the parser uses that column and keeps the student label beside the row.
Set Maximum raw score. This value defines the denominator for raw percentages, curved percentages, grade-band lookup, and warnings for scores outside the valid range.
Choose Curve method. Use Points to add or Percentage points to add for a uniform lift, Target class average for a mean-based adjustment, Scale highest score to max when the top raw score should anchor the scale, Square-root curve for a non-linear lift, Target curved range for min-to-max rescaling, or Multiplier and offset for a custom rule.
Set Score cap and Rounding mode. The formula runs first, the cap is applied next, and the rounded score is what the result uses for band comparison.
Open Advanced when the grading scale is not the default A/B/C/D/F pattern. Enter one Grade bands row per line as a label and minimum percentage, such as B,80.
Use Out-of-range scores to decide how rows below 0 or above the maximum are handled. Exclude them when they are data-entry mistakes; clamp them only when the source intentionally needs bounds.
Review validation messages before copying results. Skipped rows, clamped rows, invalid grade-band lines, and warnings about lowered scores change how the curve should be explained.

After the inputs look right, read Score Ledger for student-level changes, Band Shift for before-and-after band counts, Curve Shape for how the rule behaves across the scale, and Curve Audit for the formula, class size, range, cap, and rounding record.

Interpreting Results:

The New average tells you where the class mean landed after cap and rounding, not whether the proposed curve is justified. Pair it with Avg, scores changed, band moves, and cap hits before deciding whether the change is modest, uneven, or likely to need explanation.

For individual grades, check Score Ledger before relying on the summary. A curve that raises the average by only a few points can still move several students over a cutoff, and a cap can prevent top scores from receiving the same proportional lift as lower scores. If Curve Audit reports downward movement, confirm that the selected method is meant to lower scores before using the result.

Test curve outputs and interpretation checks
Output	What to read	Verification cue
Score Ledger	Raw score, curved score, point change, and raw-to-curved band for each row.	Inspect rows near configured cutoffs and any row with a negative change.
Band Shift	Before and after counts for each grade band.	Confirm that the band cutoffs match the class policy before treating moves as grade changes.
Band Shift Map	Visual before-and-after movement across the configured bands.	Use it to catch large distribution changes that the average may hide.
Curve Audit	Formula, valid score count, skipped rows, raw range, curved range, movement, cap, and rounding.	Use this readout as the record of which rule produced the ledger.

A higher score is not proof that the assessment was fair or unfair. Use the output to examine consequences, then review the answer key, item difficulty, rubric, student instructions, and grading policy when the curve changes final grades.

Technical Details:

A test curve is a deterministic transformation from an accepted raw point score to a final curved point score. The order of operations matters. Invalid rows are removed or clamped first, the selected curve function then produces an unrounded score, the cap limits that score to the allowed range, rounding sets the final value, and grade-band lookup compares the final percentage with the configured minimum cutoffs.

Class-dependent methods use statistics from the accepted rows. Scale-to-top depends on the highest raw score. Linear rescale depends on the observed low and high scores. Target-average mode depends on the raw class mean. Changing the out-of-range policy can therefore change more than one row: it can also change the class quantities that drive the curve rule itself.

Formula Core:

For each valid row, the final score is the selected curve function followed by a cap, rounding rule, and percentage conversion.

\begin{array}{lcl} u_{i} & = & f (s_{i}) \\ c_{i} & = & round (min (C, max (0, u_{i}))) \\ p_{i} & = & \frac{c_{i} \times 100}{M} \end{array}

Here s is the accepted raw score, f(s) is the selected curve rule, u is the unrounded curved score, C is the score cap, c is the final curved score, M is the maximum raw score, and p is the curved percentage used for grade-band lookup.

Curve methods and calculation rules
Curve method	Rule before cap and rounding	What changes the result
Add fixed points	`raw + points`	The same point amount is added to every accepted score.
Add percentage points	`raw + (percentage points / 100) × maximum raw score`	The same percentage-point lift becomes a different point lift when the test total changes.
Target class average	`raw + target average points - current raw average`	The class mean controls the lift; a lower target can reduce scores.
Scale highest score to max	`raw × maximum raw score / highest raw score`	At least one raw score must be above zero, and the highest accepted score anchors the rule.
Square-root curve	`sqrt(raw / maximum raw score) × maximum raw score`	Lower percentages usually receive larger point gains than high percentages.
Linear min/max rescale	`target low + (raw - observed low) × target span / observed span`	The observed low maps to target low, and the observed high maps to target high.
Custom multiplier plus offset	`raw × multiplier + offset`	The multiplier changes proportional gain; the offset changes every score by the same point amount.

As a substitution example, a 100-point test with a highest accepted raw score of 90 gives scale-to-top a multiplier of 100 / 90. A raw score of 84 becomes 93.333 before cap and rounding. With a cap of 100 and nearest-tenth rounding, the final curved score is 93.3, so the band lookup uses 93.3%.

Validation and boundary rules for test curve inputs
Item	Boundary	Effect
Maximum raw score	Must be greater than 0	Sets the denominator for percentages and prevents invalid scale formulas.
Raw score row	Accepted from 0 to maximum raw score unless clamped	Excluded rows do not affect averages, high/low scores, or charts.
Target class average	0% to 100%	Converted to target points before the mean-based lift is computed.
Linear target range	Target high must be greater than or equal to target low	A one-score class maps to the target high because no observed span exists.
Grade band cutoff	Label plus minimum percentage from 0% to 100%	The first band whose minimum is less than or equal to the final percentage is assigned.
Display decimals	0 to 3	Controls visible score formatting; the selected rounding mode controls the final curved score.

Boundary effects are most visible near grade cutoffs. Because cap and rounding happen before curved percentage lookup, a value just below 90% can move into the A band after nearest-tenth rounding, while a stricter cap can prevent the same movement. Keep the cap, rounding mode, out-of-range policy, and band definitions unchanged when comparing curve methods.

Privacy Notes:

The calculation runs in the browser, and the page does not need a server-side grading service to curve the scores. Treat student names and scores as sensitive classroom data anyway.

Use labels that fit your privacy rules, such as initials or anonymous IDs, when a full name is not needed.
Be careful with shared links, copied tables, downloaded files, and printed reports because they can contain the rows you entered.
Remove sample or test data before using the output as a grading record.

Worked Examples:

Hard test with a strong top score. On a 100-point test, suppose the accepted raw scores include 84, 71, 90, 63, 78, 55, 88, and 69. With Scale highest score to max, the highest raw score of 90 anchors the scale. The 84 becomes 93.3 after nearest-tenth rounding, and Score Ledger shows that row moving from B to A when A starts at 90%.

Average target that lowers scores. If the raw class average is 82 and Target class average is set to 78%, the rule subtracts four points from each valid score on a 100-point test. The warning about lowered scores is expected in this case, and Curve Audit should show downward movement before you decide whether that target is appropriate.

Score just below a cutoff. A raw 79.96 on a 100-point test may behave differently from a raw 79.9 when the curve, cap, and rounding are applied. Check the rounded Curved value in Score Ledger, then confirm the resulting Band rather than assuming the unrounded value controls the grade label.

Skipped row after paste cleanup. If one row contains 105 on a 100-point test and Out-of-range scores is set to exclude, the row is skipped and listed in the warning message. Correct the score, change Maximum raw score if the test really had more than 100 points, or switch to clamp only when the source data should be bounded instead of removed.

FAQ:

Which curve method should I try first?

Pick the rule that matches the grading reason. Use fixed points for a known item correction, target average for a planned mean, scale-to-top when the highest score should become full credit, and square root when lower percentages should receive larger gains.

Can the calculator lower scores?

Yes. Target class average can lower scores when the target is below the raw average, and Target curved range can lower some rows when the chosen low or high is below the observed raw range.

Why did a row disappear from the results?

Rows are skipped when no numeric score is found, the maximum raw score is invalid, or the score is outside 0 through the maximum and Out-of-range scores is set to exclude.

Do grade bands change the curved score?

No. Grade bands label raw and curved percentages after the score calculation. They affect Band Shift and the band text in Score Ledger, not the curved point value itself.

Should I use student names in the pasted rows?

Use the minimum label you need. The calculator keeps labels with rows so the ledger is readable, but shared links, copied tables, downloads, and screenshots may expose those labels.

Glossary:

Curve rule: The selected mathematical change applied to each accepted raw score before the cap and rounding are applied.
Raw score: The original point score from the assessment before any curve is applied.
Score cap: The upper point limit that final curved scores cannot exceed after the formula runs.
Grade band: A label assigned when a raw or curved percentage meets a configured minimum cutoff.
Target average: A planned class mean used to compute a uniform lift or reduction in target-average mode.
Band move: A change between the raw grade band and the curved grade band for the same row.

References:

Grading Scale, Carnegie Mellon University Eberly Center.
Use criterion-referenced grading, Carnegie Mellon University Eberly Center.
Students complain about grades, Carnegie Mellon University Eberly Center.
Grading Equitably, University of Iowa Center for Teaching.

Test Curve Calculator

{{ summaryHeading }}

Current result

How to Use This Tool:

Interpreting Results:

Technical Details:

Formula Core:

Privacy Notes:

Worked Examples:

FAQ:

Glossary:

References:

Test Curve Calculator

{{ summaryHeading }}

Current result

Introduction

How to Use This Tool:

Interpreting Results:

Technical Details:

Formula Core:

Privacy Notes:

Worked Examples:

FAQ:

Glossary:

References: