Why a raw score is not a score
Your raw score is a count: one point for every question you answer correctly, no deduction for a wrong answer, no partial credit. Your scaled score is what colleges see, and it runs from 200 to 800 per section. The two are not the same number and they are not even in a fixed relationship, because the College Board equates each test form separately.
Equating exists to solve a fairness problem. Two students who sit different SAT forms on different Saturdays face different questions, and no two sets of questions are exactly equally hard. If the raw score were reported directly, the student who happened to draw the easier form would look stronger. Equating measures how hard each form turned out to be, then stretches or compresses the raw-to-scaled mapping so that a 650 means the same level of ability whichever form produced it. On a slightly harder form, 40 correct might scale to 660; on an easier one, the same 40 correct might scale to 640.
That is why no single published chart converts raw SAT scores for all tests. Each practice test released by the College Board carries its own conversion table in its scoring guide, and the digital adaptive SAT adds a second layer: your performance on the first module determines the difficulty of the second, and the scoring engine weighs correct answers on the harder module more heavily. This calculator gives you a defensible estimate rather than a fiction of precision, and it lets you feed in the one anchor that matters most if you have the official table in front of you.
The three-anchor interpolation this calculator uses
Every SAT conversion table shares three fixed properties, and the model here is built from exactly those. A raw score of zero scales to 200, the floor. A perfect raw score scales to 800, the ceiling. In between, the mapping is monotonic — one more correct answer never lowers your scaled score — and reported in 10-point steps.
Two straight lines are enough to capture the shape. The first runs from the point (0 raw, 200 scaled) to the anchor point at half the questions correct. The second runs from that anchor to (all questions correct, 800 scaled). You control the anchor. Leaving it at 500 produces one unbroken straight line across the whole range, which is the neutral assumption when you have no table.
If you do have your form's official conversion table, find the raw score closest to half the section length, read off the scaled score beside it, and type that number into the anchor field. Real SAT curves bow: on most forms the middle of the range is worth slightly more than a straight line predicts, so the anchor sits a little above 500, and the model then reproduces the published table to within roughly a step or two across most of the range. Where it will always be least accurate is at the very top, where a single missed question can cost 20 or 30 points on a generous form and only 10 on a strict one.
The percentile is a separate calculation and a deliberately simple one. It treats total scores as normally distributed with the mean and standard deviation you supply, so z = (total − mean) ÷ SD and the percentile is the area of the normal curve to the left of z. Real SAT score distributions are not perfectly normal — they are slightly skewed and they pile up at the reporting ceiling — so treat this as an order-of-magnitude reading, and replace the mean and standard deviation with the published values for your cohort when you have them.
Worked example: 40 of 54 in Reading and Writing, 33 of 44 in Math
You have just finished a practice test. You count 40 correct in Reading and Writing out of 54 questions, and 33 correct in Math out of 44. Your answer key shows that 27 correct in Reading and Writing converts to 520, and 22 correct in Math converts to 530, so you enter those as the anchors.
- Find the halfway raw score in Reading and Writing. 54 ÷ 2 = 27 questions. Your 40 correct is above it, so use the upper branch.
- Measure how far above the midpoint you are. 40 − 27 = 13 questions, out of the 27 questions that separate the midpoint from a perfect score. That is 13 ÷ 27 = 0.4815 of the upper half.
- Interpolate. The upper half spans 800 − 520 = 280 points. So the scaled score is 520 + 280 × 0.4815 = 520 + 134.8 = 654.8, which rounds to 650.
- Repeat for Math. Half of 44 is 22, and you are 33 − 22 = 11 questions above it, out of 22. That is exactly half of the upper half. The upper span is 800 − 530 = 270 points, so the scaled score is 530 + 270 × 0.5 = 530 + 135 = 665, which rounds to 670.
- Add the sections. 650 + 670 = 1320 total.
- Estimate the percentile. Against a mean of 1030 and a standard deviation of 200, z = (1320 − 1030) ÷ 200 = 1.45. The normal curve puts 92.6% of the area below z = 1.45, so roughly the 93rd percentile.
- Measure the gap to a 1400 target. 1400 − 1320 = 80 points. On this form each additional Math question is worth about 270 ÷ 22 = 12.3 scaled points, so roughly seven more correct answers spread across the two sections closes it.
That last step is the one worth internalising. Dividing the span of a half by the number of questions in it tells you the marginal value of one more right answer, and it is usually between 10 and 15 points near the middle of the scale.
How to read the number you get
Read the section scores before the total. A 1320 built from 700 Math and 620 Reading and Writing is a different application from a 1320 built from 660 and 660, and selective programmes in engineering, computer science and economics look hard at the Math score on its own. Colleges that superscore take your best section from any sitting, which means the balanced 1320 and the lopsided one can end up in very different places after two attempts.
Treat any single practice-test score as an estimate with real width around it. The College Board publishes a standard error of measurement for SAT section scores that is around 30 points, so two forms taken the same week can differ by 40 or 50 total points from measurement noise alone. Three practice tests that read 1290, 1320 and 1350 describe one student around 1320, not a rising trend.
For the percentile, remember that the SAT-taking population is not the general population of teenagers. It is self-selected, and in states with universal school-day testing it is close to the whole cohort, which pulls the mean down. That is why a percentile computed against a national mean will read lower than the percentile you would get against your own school's testing population, and higher than the percentile you would get inside the applicant pool at a highly selective college.
Set your target from the college, not from a round number. Published common data sets give the 25th and 75th percentile SAT scores of enrolled students. Landing at or above the 75th percentile puts your score in the part of the file that helps you; landing below the 25th means the rest of the application has to carry it. If you want to see how the same performance reads on the other admissions test, run the number through the SAT to ACT score conversion calculator.
Marginal value of one more correct answer
| Section length | Midpoint anchor | Lower half: points per question | Upper half: points per question |
|---|---|---|---|
| 54 questions (R and W) | 500 | 11.1 | 11.1 |
| 54 questions (R and W) | 520 | 11.9 | 10.4 |
| 54 questions (R and W) | 540 | 12.6 | 9.6 |
| 44 questions (Math) | 500 | 13.6 | 13.6 |
| 44 questions (Math) | 530 | 15.0 | 12.3 |
| 44 questions (Math) | 560 | 16.4 | 10.9 |
Lower half = (anchor − 200) ÷ (Q/2). Upper half = (800 − anchor) ÷ (Q/2). A higher anchor makes early questions worth more and late questions worth less, which is exactly the shape of a generous curve.
Mistakes that make a practice score wrong
- Using another test's conversion table. The single largest source of error. A table from Practice Test 4 applied to Practice Test 8 can be off by 40 total points, because the two forms were equated to different difficulties.
- Scoring an adaptive digital test as if it were linear. On the digital SAT, routing to the easier second module caps the attainable scaled score even with every remaining question correct. A raw-count conversion cannot see that, so it will overstate a score that came from the lower-difficulty path.
- Counting unanswered questions as wrong in a way that changes the raw score. It does not change anything: the SAT has had no guessing penalty since 2016, so a blank and a wrong answer score identically. Answer everything.
- Rounding the total instead of the sections. Section scores are reported in 10-point steps and the total is their sum, so a total always ends in 0. A total ending in 5 means something has been averaged that should have been added.
- Treating the percentile as official. The percentile here comes from a normal-curve model of the mean and spread you typed in. The College Board's published percentiles come from actual score frequencies and differ by several points in the tails.
- Reading a 30-point improvement as progress. It sits inside the measurement error of the test. Look for a change of 60 points or more, sustained over three sittings, before you conclude that the preparation is working.
Where this sits among the other scoring tools
Raw-to-scaled conversion is one link in a chain. If you want to compare an SAT total against an ACT composite — because your state administers one and your colleges accept both — the official concordance is the right instrument, and the SAT to ACT conversion calculator applies it. If you want to understand what a percentile means in terms of population rarity, the same normal-curve machinery drives the IQ percentile and rarity calculator, where standard scores are built with a fixed mean and standard deviation instead of being equated form by form.
On the preparation side, the constraint that usually binds is time rather than technique. Estimating how long an assigned reading list will take at your own pace is the job of the reading time calculator, and if reading speed itself is the bottleneck on the Reading and Writing section, measure it properly with the reading speed calculator before deciding it is a problem. Students who have to pay for score reports and testing alongside application fees can fold those into the wider budget with the cost of attendance calculator.
One last framing point. The SAT is designed so that the scaled score, not the raw score, is the stable quantity. When you practise, track scaled scores across forms and raw percentages within a form. Raw counts across different forms tell you almost nothing, because the forms were never meant to be compared that way.
Key terms
- Raw score
- The plain count of questions answered correctly in a section, with no penalty for wrong answers.
- Scaled score
- The reported 200-800 section score produced by applying the form's equating table to the raw score.
- Equating
- The statistical adjustment that makes scores from different test forms interchangeable, by correcting for how hard each form turned out to be.
- Superscoring
- A college policy of combining your highest Reading and Writing score and your highest Math score across separate sittings into one total.
- Standard error of measurement
- The typical amount by which an observed score differs from the true ability it estimates; for SAT sections it is roughly 30 points.