How is a raw SAT score converted into a scaled score?

A raw score is the simplest number on the test: how many questions you got right. There is no deduction for wrong answers, so your raw score is just a count.

Everything interesting happens in the step after that.

Why it is a table and not a formula#Link to the section: Why it is a table and not a formula

The obvious approach would be a percentage — 80% correct becomes 80% of the way from 200 to 800. No real test does this, for a reason that becomes clear once you look at what scores are supposed to mean.

A scaled score is meant to represent the same level of ability across every edition of the test. But no two editions are exactly equally hard. If March's Math section happened to be slightly harder than October's, a percentage would punish the March cohort for the accident of which questions they drew. Equating exists to remove that: the conversion is adjusted per edition so the same scaled score means the same ability regardless of which form you sat.

That adjustment cannot be a formula, because the relationship it encodes is empirical. It is a table, produced from data about how test-takers of known ability performed on these specific questions.

Why the curve is not a straight line#Link to the section: Why the curve is not a straight line

Plot raw score against scaled score and you get an S, not a diagonal.

In the middle of the range, questions are doing the most discriminating work — that is where most test-takers are, and where a small difference in ability shows up as a difference in correct answers. Two more right answers there can move your section score by 30 or 40 points.

Near the top the curve flattens, then steepens sharply at the very end. Most of the hardest questions separate very few people, so getting one more of them right moves you less than it did in the middle — until you reach the last few, where the difference between 2 wrong and 0 wrong is the difference between a very good score and a perfect one.

Near the bottom it flattens for the opposite reason: below a certain point the test is not measuring much, and the floor is 200 regardless.

The practical consequence is that "how many more questions do I need to get right" has a different answer depending on where you are. A student at 600 gains more per question than a student at 750.

Why nobody outside the College Board has the real table#Link to the section: Why nobody outside the College Board has the real table

Equating tables are produced from operational data — how a representative sample of real test-takers performed on real questions, including the unscored pretest questions embedded in each form for exactly this purpose. That data belongs to the College Board and is not published.

This is the honest limit on every third-party practice score, including ours. What a practice platform can do is calibrate its own table: write questions, measure how people perform on them, and build a conversion from that. What it cannot do is claim the result matches an official conversion it has never seen.

We build one table per second module — separately for the easier and harder paths, since a raw count means something different on each — and then report a range around the result rather than the result alone, because a table calibrated on our own data has error in it and pretending otherwise would be the dishonest part.

What to do with this#Link to the section: What to do with this

Stop treating a raw score as a percentage. "I got 80%" is not a score, and it does not convert to one in your head.

And when a practice platform gives you a scaled score, ask what it converted from. If the test was not adaptive, there was no module path, so there was no correct table to use — which means the number came from somewhere, but not from anything resembling how the real test works.