Loading...
Loading...
We believe in transparency. Here's exactly how well our system predicts your SAT score.
Our prediction engine is built on the same psychometric methods used by the College Board and ETS.
We use the same two-parameter logistic Item Response Theory model the SAT itself uses. Each question has calibrated difficulty and selectivity parameters.
Each question's difficulty and selectivity are calibrated to produce a Test Characteristic Curve that maps ability to expected performance.
Your raw performance maps to the official SAT scale using equipercentile linking — the same method used to equate different SAT forms.
We maintain calibrated confidence estimates for each of 120 tracked skills (SK001-SK120), updating after every response so your score prediction improves with each question you answer.
These numbers update in real-time as students report their official SAT scores. No cherry-picking, no adjustments after the fact.
Our question taxonomy maps directly to the College Board's official SAT Blueprint. Every skill we track corresponds to real test content.
| College Board Domain | Our Skills | Coverage |
|---|---|---|
| Information and Ideas | SK001 - SK030 | Reading comprehension, central ideas, command of evidence, inferences |
| Craft and Structure | SK031 - SK050 | Words in context, text structure, purpose, cross-text analysis |
| Expression of Ideas | SK051 - SK060 | Rhetorical synthesis, transitions, organization |
| Standard English Conventions | SK061 - SK075 | Grammar, punctuation, sentence structure, subject-verb agreement |
| Algebra | SK076 - SK090 | Linear equations, systems, inequalities, functions, word problems |
| Advanced Math | SK091 - SK105 | Quadratics, polynomials, exponentials, radicals, rational expressions |
| Problem-Solving and Data Analysis | SK106 - SK115 | Ratios, percentages, probability, statistics, data interpretation |
| Geometry and Trigonometry | SK116 - SK120 | Area, volume, angles, triangles, circles, trigonometry |
Most prep tools tell you a score. We tell you how confident we are.
We don't just say "your predicted score is 1320." We say "1320 ±35 points (95% CI: 1250-1390)."
For every one of 120 skills, we track not just your estimated ability but the uncertainty of that estimate. Skills with high uncertainty get prioritized for more practice to sharpen the measurement.
We show you which skill estimates have wide confidence intervals and which are nailed down. This helps you focus practice where it matters most — and where our model needs more data from you.
1. Student takes practice tests: Our adaptive engine produces a predicted SAT score based on IRT ability estimation and equipercentile linking.
2. Student takes the real SAT: After the official test, they report their actual score.
3. We compare honestly: We take our last prediction before their test date and compare it to their actual score. No post-hoc adjustments, no filtering out bad predictions.
4. We publish everything: MAE (average error magnitude), RMSE (penalizes large errors), bias (systematic over/under prediction), and band accuracy (% within 30, 60, 100 points).
Every official score you report makes our predictions better for all students. It takes 30 seconds.