How to read your child's SHSAT diagnostic report
The 20-question diagnostic produces four things: an estimated composite score with a margin on it, the gap to your target school's last published cutoff, a label for each of the eight specialized schools, and a list of topics ordered weakest first. Each is built to answer a different question — and reading them the wrong way round is the most common way a good report goes to waste.
The big number is an estimate with a margin — on purpose
What the score is
An estimated composite out of 700 — two sections, ELA and math, each scaled to 350. How raw answers become a scaled score is worked through on the score calculator page.
What the ± means
Twenty questions is a sample, and a sample has sampling error: at 20 questions the honest band is about ±55 points. That is not hedging — a report that prints one bare number from 20 questions is claiming a precision it does not have.
How the margin shrinks
The band narrows as more questions go in — a full 100-question mock brings it down to roughly ±25. The diagnostic's job is direction, not verdict; the mock's job is a number worth quoting.
What not to do with it
Do not compare two single numbers a week apart and call it progress or decline — inside the margin, movement is noise. The trend across weeks of practice is the signal.
The gap line is the working number
Under the score, the report subtracts it from your target school's last published cutoff. That gap is the number the practice plan exists to close — and a large one at the start is normal, because the diagnostic is usually sat months before the exam.
Read the cutoff itself carefully: it is history, not a target. Cutoffs are what last year's final offer scored, they move every year, and next cohort's do not exist yet. So the goal is margin above the line, not landing on it. The cutoff page lists all eight with their source and year.
The safety / target / reach labels on the eight schools apply the same arithmetic to every cutoff at once. They are a way to see the whole field, not a prediction — a label computed from a ±55 estimate inherits the ±55.
All eight cutoffs, with sources →
The topic list is sorted weakest first — that order is the study plan
- The topics at the top of the list lost the most points. Improving a 40% topic buys far more composite than polishing a 90% one — start at the top, not with the familiar.
- Watch the pacing flag as well as the accuracy. A topic can be accurate but slow — and on a timed paper, slow-but-right still loses points, just on whatever question there was no time left for.
- On a free account the score, the margin and the school gaps are all kept; the topic-by-topic detail is part of a plan. The honest version of this page says so here, not at the paywall.
The two weeks after the report
- Spend the free practice on the top of the list. Every topic includes one free five-question set — spent on the two or three weakest topics, that is a real first fortnight of work; spent on the strong ones, it confirms what you already knew.
- Hold the free mock in reserve. The free plan includes one sitting of Mock Exam 1 — the wrong move is burning it the same week as the diagnostic. Practise the weak topics for a few weeks first, then spend the mock to see whether the work moved the score.
- If the report is weeks old, act on the trend, not the memory. The diagnostic is one per account and only counts once finished — the score does not refresh; the practice record from the weeks since is the newer evidence.
One caution before treating any of these numbers as a forecast: the official Fall 2026 SHSAT is computer-adaptive, and this estimate comes from fixed-form questions — every student answers the same twenty. The skills measured are the same; the exam-day mechanics are not, and the difference is explained on the 2026 changes page.
The official Fall 2026 SHSAT will be computer-adaptive. SHSATus currently provides fixed-form digital practice based on DOE-published content areas and question types. It does not reproduce the DOE's adaptive item-selection or scoring algorithm.