Methodology

The one rule

If a number is not in a document published by College Board or ACT, it does not ship as data. Not interpolated, not reconstructed from memory, not borrowed from a forum thread and relabelled as official. Where we could not source something, the site says we could not source it.

Where the SAT conversion tables come from

College Board publishes a scoring guide PDF for each of the 8 linear paper practice tests, numbered 4 through 11. Each contains a raw score conversion table for Reading and Writing and for Math.

Those PDFs were downloaded and parsed programmatically. The parser reads the text layer with its x and y coordinates, clusters cells into rows and columns by position, and then validates the result before anything is written to disk. The checks are:

• The raw score column must run 0 to the section maximum with no gaps.
• Lower and upper bounds must each be non-decreasing as the raw score rises.
• The lower bound must never exceed the upper bound.
• Every value must sit inside that exam's published section scale.
• The page must contain exactly one conversion table. This one matters: page 2 of each guide embeds thumbnail previews of pages 4 and 5, so it carries two table headers and a mixture of cells from both. A scraper that takes the first table it finds can silently serve one test's curve as another's.

The scripts that do this are in the repository alongside the data, so the extraction can be re-run and diffed whenever College Board republishes.

How a score is calculated

Add the two module raw scores together to get a section raw score. Look that up once in the table for the test you took. Read off the lower and upper scaled values. Do the same for the other section. Then add the two lower bounds together and the two upper bounds together.

The order matters. Converting each module separately and summing the scaled results produces incoherent output, in which a student who scored evenly across both modules ends up below one who was lopsided at the same total raw score. College Board's guide specifies summing the raw scores first, and that is what this site does.

Why every result is a range

Because the source is a range. College Board's tables give a lower and an upper scaled score for every raw score, and the guide explains why:

Section and total scores for the paper version of this digital SAT practice test are expressed as ranges. That's because the scoring method described in this guide is a simplified (and therefore slightly less precise) version of the one used in the actual test.

Collapsing that into one number would throw away information the source deliberately included.

Why the site asks which test you took

The eight official tables disagree with each other. At a Reading and Writing raw score of 40, the lower bound ranges from 470 to 540, a 70-point spread. A calculator that applies one generic curve is wrong by up to that much on one section alone.

The two raw score scales

	Adaptive Bluebook	Linear paper
Reading and Writing	54 asked, 50 scored	66
Math	44 asked, 40 scored	54
Official conversion table	None published	Yes, one per test

These are not interchangeable. A raw score out of 54 cannot be looked up in a table built for 66. This is the most common error in the category, and it is invisible to the person it affects.

What the estimate mode does

For a Bluebook test there is no official table, so the calculator does the following and labels the whole thing an estimate:

1. Express the raw score as a proportion of the section's question count, 54 or 44. The four unscored pretest questions per section are indistinguishable to the student and cancel out of the proportion.
2. Map that proportion onto the linear tables' raw axis, 66 or 54.
3. Look that position up in all eight official tables and return the lowest lower bound and the highest upper bound found.

Step three is what makes the band defensible. Its width is the measured disagreement between official tables at that position, not a number we chose. The estimate is still an estimate: proportional mapping assumes an even difficulty distribution, which an adaptive test does not have. The UI says so in every state, and the band is drawn with a hatched fill rather than a solid one so that a screenshot carries the distinction.

Anomalies in the source tables

Three rows in College Board's published tables are internally inconsistent. Each was re-checked against the raw PDF text coordinates; the extraction is correct and the oddity is in the source.

Table	Section	Raw	What the source says
sat-6	Reading and Writing	41	lower 530 < previous 540
psat89-1	Reading and Writing	57	upper 630 < previous 660
psat89-1	Reading and Writing	66	lower 700 < previous 710

They are reproduced unchanged and flagged where they appear. Silently correcting a source would make this site's numbers unverifiable against the document they claim to come from, which would defeat the point of citing it.

Percentiles

Two tables are published by College Board Research and both are shown, always labelled: the Nationally Representative Sample and the SAT User Group. They were extracted from the published page and validated for row count, 10-point stepping and monotonicity. The censored bounds 99+ and 1- are reproduced as printed rather than converted to numbers.

A third percentile exists and cannot be reproduced. Your actual score report shows an All Tester Percentile, comparing you with the past three cohorts of 12th graders worldwide. College Board publishes no lookup table for it. Any site that claims to give you the number from your report is showing you something else.

Cohort figures come from the 2025 SAT Suite of Assessments Annual Report, Total Group: mean total 1029 across 2,004,965 test takers. Benchmarks are 480 for Reading and Writing and 530 for Math.

ACT data

• Scale conversion comes from official practice form 25MC1. ACT tables are form-specific, so this is labelled as one form's table rather than as the ACT curve.
• Composite is (English + Math + Reading) / 3, round half up. Science is excluded, per ACT's own statement that science and writing do not affect the Composite score.
• Percentiles come from the ACT Score National Ranks for the 2025-2026 reporting year, based on graduates of 2023 to 2025. That table states its own validity window, which ends in August 2026, so it carries a date wherever it is shown.
• Concordance is the 2018 joint ACT and College Board table, the only official one. It is always displayed with the published standard errors of plus or minus 2 ACT points and plus or minus 80 SAT points, and with the caveat that it concords the retired four-section ACT against the pre-digital paper SAT.

College score ranges

The 1009 college pages join two federal datasets on UNITID: College Scorecard for score ranges and admission rates, and IPEDS Admissions for the percentage of enrolled students who submitted a score. The join exists because Scorecard does not carry that percentage at all, and without it a published range is misleading.

Institutions are included only if they are currently operating, four-year, predominantly bachelor's-degree granting, and report a complete SAT or ACT range. That is 1009 institutions, of which 928 have SAT data and 896 have ACT data.

Three rules apply to every one of those pages:

• No composite SAT range. The 25th percentile of a total is not the sum of the two section 25th percentiles, because the same student is rarely at the 25th percentile in both. Neither source publishes a composite, so the sections are shown separately and the summed figure is described as the rough orientation number it is.
• The submission percentage leads. Across these institutions the median SAT submission rate is 21 percent. A range built from a fifth of a class, self-selected for having scored well, is a different kind of number from one built from all of it.
• No test-required flag. The federal policy field is self-reported and lags real changes, sometimes by a full admissions cycle. Each page sends you to the institution's own admissions site instead.

The two datasets are one year apart in vintage. That is stated on every page rather than smoothed over.

AP exams

The AP tools are the clearest illustration of the rule at the top of this page. College Board publishes exam structures and score distributions, so those are here for all 15 subjects. It publishes no composite cut points for any current exam year, so there is no cut point anywhere on this site.

The AP calculator therefore computes your weighted percentage of the exam score, using College Board's own published section weights, and stops there. It shows the published distribution beside it and does not draw a line between the two.

There is a second reason beyond non-publication. College Board sets the boundaries each year so that a 3 represents the same achievement as last year's 3, which means the composite required moves with exam difficulty. A genuine cut point from a released exam does not carry forward even within one subject.

Two vintages sit on those pages and are labelled separately: the structures describe the May 2027 administration, which is what College Board currently publishes, while the distributions are from May 2026.

What this site will not publish

• A raw to scaled conversion table for the adaptive Bluebook SAT. None exists.
• A module 1 routing threshold, or any claim about how many questions unlock the harder second module. Not published, not inferable.
• A score ceiling for the lower-difficulty route. The only official figure is College Board's simulation of a deliberate underperformer, which gave at most 580 in Math specifically. That is not a general cap, and the 590, 620 and 630 figures circulating elsewhere are unsourced and contradict each other.
• AP composite cut points presented as fact. College Board publishes none.
• National Merit state cutoffs described as official. NMSC publishes none; every table in circulation is crowd-sourced.
• An SAT composite 25th to 75th percentile range summed from section percentiles. The 25th percentile of a sum is not the sum of the 25th percentiles.

Corrections

If you find a number here that does not match its cited source, that is a bug and worth reporting. Point at the page and the figure and it will be checked against the source document. See contact, and the full source list is here.

Percentile data retrieved 2026-08-26. Conversion tables retrieved 2026-08-26. ACT data retrieved 2026-08-27.