How the digital SAT is actually scored

Your digital SAT score is not a count of right answers passed through a curve. College Board runs your full answer pattern through a statistical model called Item Response Theory (IRT), which accounts for which questions you answered correctly, not only how many. The adaptive second module changes which questions you saw in the first place. Together those two facts explain almost every confusing thing about SAT scoring.

Here is the whole mechanism, in order.

The test you actually sit

The digital SAT has two sections, and each section is split into two separately timed modules. You cannot go back to module 1 once you have moved on, though you can move freely among questions inside the module you are currently in.

Section	Questions	Modules	Time per module
Reading and Writing	54	2 x 27 questions	32 minutes
Math	44	2 x 22 questions	35 minutes
Total	98	4	2 hours 14 minutes

There is a 10-minute break between the Reading and Writing section and the Math section. These counts come from College Board’s test structure page.

Only 90 of those 98 questions are scored

This is the part most explanations skip. Every module contains exactly 2 pretest questions that do not count toward your score. College Board is using your answers to calibrate those items for future tests.

In the Assessment Specifications Overview, the Reading and Writing module is described as “1st module: 25 operational questions and 2 pretest questions”, and the Math module as “1st module: 20 operational questions and 2 pretest questions”.

Section	Administered	Scored (operational)	Pretest
Reading and Writing	54	50	4
Math	44	40	4
Total	98	90	8

Pretest items are placed randomly and are indistinguishable from scored ones. You cannot identify them, and you should not try. A question that feels strange or unusually hard is not evidence that it is unscored.

Module 1 decides which module 2 you see

The digital SAT uses multistage adaptive testing, specifically a 1-2 design. Module 1 is the routing module and contains a broad mix of difficulty. Based on your performance there, module 2 is delivered as either a lower-difficulty form or a higher-difficulty form.

Adaptivity happens within a section only. Reading and Writing module 1 routes to Reading and Writing module 2. Math routes separately. A weak Reading and Writing performance has no effect on which Math module you receive.

Both second-stage modules contain easy, medium and hard questions. They differ in the proportions, not in kind. College Board has never published the routing threshold, so nobody outside the organization knows how many questions you need in module 1 to reach the harder form. Any site quoting a specific number is guessing. The adaptive modules guide covers this in detail, including why the widely repeated score ceilings for the easier route are unsourced.

What IRT does with your answers

Once you finish, the model has two kinds of information: which specific questions you saw, and which of them you answered correctly. Each scored question carries calibrated statistical properties, difficulty being the most intuitive one. College Board uses a 3-parameter logistic model for multiple-choice questions and Samejima’s graded response model for student-produced responses.

College Board states it directly on its scoring page:

The digital SAT’s multistage adaptive test design and use of Item Response Theory (IRT) methodology allow for precise measurement of students’ knowledge and skills with fewer questions in less time than possible with traditional paper and pencil tests.

And on what feeds the score:

The scores students receive are a product of several factors, characteristics of the questions they answered right or wrong (e.g., the questions’ difficulty levels), and the probability that the pattern of answers suggests they were guessing.

That second sentence contains a detail worth sitting with. The model considers the probability that your answer pattern looks like guessing. A run of correct answers on hard items after misses on easy ones is statistically different from steady performance, and the model treats it differently.

The output is a section score on a 200-800 scale in 10-point steps, and a total of 400-1600. Each section is scored independently and then added.

Why there is no SAT curve

“Curve” implies your score depends on how other test takers did that day. It does not. IRT scoring puts every form on the same scale, which is the entire point of calibrating items in advance. A 1300 in March means what a 1300 in October means.

What people usually notice and call a curve is difficulty compensation. If you saw a slightly harder set of questions, the model already knows those questions were harder and accounts for it. That adjustment is baked into the item parameters, not applied afterward based on the cohort.

The consequence is the sentence in College Board’s Technical Manual that surprises students most: two students may have the same number of correct answers and receive different reported scale scores. Why that happens is the single most common question after students compare with a friend.

Where conversion tables still exist

College Board does publish official raw-to-scaled conversion tables, but only for the linear paper versions of the practice tests, available at satsuite.collegeboard.org/practice/practice-tests/paper. Eight SAT tables exist, covering Practice Tests 4 through 11.

Two things about those tables are constantly reported wrong online. First, the paper tests have 66 Reading and Writing questions and 54 Math questions, a different scale from the adaptive test’s 54 and 44. A raw score from one cannot be looked up in the other. Second, the tables return a lower and an upper scaled score, never a single number. Every scoring guide says so on page 3:

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.

There is no conversion table for the adaptive Bluebook tests, because Bluebook scores them itself using the same IRT machinery as the real test. Our raw score conversion reference reproduces the official tables per test, and the practice test calculator applies the official procedure: sum both module raw scores first, then do one section lookup.

What your score report shows

Digital score reports no longer include subscores or cross-test scores. Those were eliminated with the format change. You get section scores, a total, performance across four Reading and Writing content domains and four Math content domains, and percentile information.

Percentiles are a separate system from scoring and are worth understanding on their own, since the number printed on your report is not the one most calculators produce. That is covered in the SAT percentiles reference.

What follows from all this in practice

Three things are true and useful at once. You cannot compute your own adaptive score from a raw count, so estimates from any calculator, including ours, are estimates. Nothing is gained by trying to identify pretest questions or to steer the routing. And because there is no wrong-answer penalty, an answer left blank and an answer marked wrong cost you exactly the same, which makes guessing on a question you would otherwise leave empty strictly free.

For how we build our estimates and where each number comes from, see the methodology page.