How the ATAR is calculated — and how accurate this calculator is

The four steps

Step 1 — How VCAA turns your marks into a study score

Every VCE study has (at most) three graded assessments (GAs) — usually Unit 3 SACs, Unit 4 SACs and the exam, though some studies combine Units 3–4 coursework, have a School-assessed Task (SAT), a performance exam or two exams. Each GA contributes a fixed share of the study score, set by the study design and listed in VCAA's table of contributions (e.g. English 25 / 25 / 50; Methods: coursework 40, Exam 1 20, Exam 2 40; Physics 30 / 20 / 50). ATARtools' defaults are checked against that 2026 table for all 93 studies on every test run.

VCAA's method, from the VCE Administrative Handbook:

1. Standardise each GA. standardised = (your final score − state mean) ÷ state standard deviation
2. Weight and add. Multiply each by its contribution and add them up. For English: total = 0.25 × U3 SACs + 0.25 × U4 SACs + 0.50 × exam (all standardised).
3. Rank the state. Every student's total is ranked (students who tie get the group's highest rank).
4. Normalise. The ranks go through an inverse-normal function and onto a scale with mean 30 and standard deviation 7, truncated at 0 and 50. Small studies get further adjustments.

Two consequences people often miss: a GA's spread matters, not just its weight (75% on a tightly bunched English exam is rarer than 75% on a widely spread Methods exam), and the study score is a rank: it says how your weighted total compares with everyone else's in the state that year.

What the calculator does

Exactly the same thing. The state mean and standard deviation of every GA come from VCAA's official 2025 grade distributions (browse them). The one thing VCAA doesn't publish is the full list of everyone's totals, which step 3 needs. So ATARtools rebuilds it: it simulates the state with 16,384 students whose marks on each GA follow that GA's published grade distribution (every grade band's real count and mark range, pooled over the last three years — see below), applies VCAA's formula to each simulated student, and ranks your total against theirs. Marks on different GAs are linked by realistic correlations (below), because a student who does well on SACs usually does well on the exam.

Exam difficulty changes every year

VCAA standardises with this year's state mean and SD, and those aren't known until results come out. Exams are not equally hard each year. Methods Exam 1 averaged 46.5% in 2024 and 55.9% in 2025, and across all large exams the state average moves by about 0.24 SD (≈ 4.6 percentage points) from one year to the next. So a 70% on this year's exam could rank noticeably higher or lower than a 70% did last year.

ATARtools predicts each assessment's mean and SD for this year from the official figures for 2021–2025, with each year back counting half as much. The 2019–2024 figures come from VCAA's per-study grade-distribution reports. In a backtest on every exam with 500+ students (195 exam-years), this blend predicted the next year's exam average with an error of 0.225 SD, against 0.248 SD for "same as last year". It was better in all five years tested (2021–2025). The error that remains goes into the likely range: ±0.22 SD for large written exams, ±0.30 for 100–499 students, ±0.44 for smaller ones, ±0.11 for performance exams and ±0.08 for coursework.

Difficulty is only half of what changes. The other half is the shape of the distribution: how many students pile up at the top, how skewed the marks are, and where the A+ cut lands. Rather than trust 2025's shape alone, ATARtools pools the official grade-band distributions (every band's count and mark range) of the last three years, 2023–2025, with each year back counting half as much. Each year is standardised by its own mean and SD first, so an easy or hard year contributes its shape but not its difficulty. In a backtest on every assessment with 500+ students and three earlier years, predicting each grade boundary's state percentile from the pooled shape was about 11–12% closer than using only the latest year (2025: 0.67 vs 0.76 percentile points; 2024: 0.66 vs 0.74), with the biggest gain for exams (0.77 vs 0.88 in 2025). Assessments with no comparable earlier year keep their 2025 shape.

What each grade will need

VCAA sets the mark each grade (E up to A+) starts at separately for every assessment, every year, so a 72/80 can be an A+ one year and an A the next. ATARtools forecasts next year's cut-offs from the last three years' official ones for the same assessment, with each year back counting half as much, and shows them on the calculator: the mark you need for the next grade up and for an A+, a table of every grade's forecast cut-off next to the actual cut-offs of the last three years, and a "likely ±" for each. In a backtest on every assessment with 500+ students, this beat "same as last year": for exams the error in the cut-off was 3.8 points of the maximum mark against 4.5 for 2024 and 4.2 for 2025; for school-assessed work 2.6 against 3.0 for 2024 and a tie in 2025 (2.3). The shown ± is that measured error for each grade, about ±3 points of the maximum mark for the A+ cut-off on an exam (±5 marks on a 160-mark paper) and ±1 for school-assessed work. Weighting by the forecast difficulty or by grade shares did no better, so the plain average is used. A cut-off forecast is still a forecast: a harder or easier paper can move every cut-off together.

SAC moderation

Schools set and mark their own SACs, so VCAA statistically moderates them. For each study at each school, the SAC marks are aligned to the same students' external (exam) scores. The top SAC score is matched to the top external score, and the median and quartiles are matched the same way. Your rank within the school never changes. If your school's students do well on the exam, everyone's SACs move up. Moderation depends on the whole group's exam results, never on your own exam. In a few studies the GAT is also used.

A raw SAC mark from your teacher therefore isn't yet your final score. The calculator handles this in three ways:

• Regression to the mean: your SAC mark's statewide standing is pulled 15% toward the middle (reliability 0.85), because part of a high raw mark usually reflects your school's marking scale.
• Your school (optional): a school whose median study score is above the state centre has a stronger cohort, so its SACs are moderated up. The shift is 0.7 × (school median − 29.6) ÷ 7 standard deviations, capped at ±0.7. 29.6 is the student-weighted average of every school's median in VCAA's 2025 data, ≈ 30 by construction.
• The rest is uncertainty, carried into the likely range: ±0.45 SD on your SAC standing (±0.32 once a school is chosen).

Assessments you haven't done yet

A blank isn't assumed to be average. It's projected from the marks you have entered, using the same correlations: strong SACs project a strong but slightly regressed exam. The estimate shown is the median over everything still unknown (blank GAs and moderation), and the likely range is the 16th–84th percentile (about ±1 SD). Fill in more marks and the range narrows.

Assumptions (VCAA doesn't publish these)

Sensitivity, measured over 100 typical profiles: moving every correlation by ±0.10 changes an estimate by 0.2 study-score points on average (worst case 0.8). Changing the SAC reliability to 0.80 or 0.90 has the same effect. These assumptions matter far less than the marks you enter.

How we know the estimator does what VCAA does

• With normally distributed GAs, VCAA's method has an exact closed-form answer (30 + 7 × total ÷ SD of the total), and the calculator matches it to within 0.1 points.
• Against a separate simulation of 1,000,000 English students, ranked exactly as VCAA ranks them, the calculator's total-to-study-score map agrees to within ±0.1 points from the 0.5th to the 99.5th percentile.
• Marks at the state mean on every GA give 30 ± 1 in all 18 studies tested; a higher mark never gives a lower estimate; the same marks always give the same answer.
• Simulated students' estimates come out with mean 30 and SD 7, just as real study scores do, and 40+ is the right share.

Step 2 — Scaling (VTAC)

A 30 in one study isn't the same achievement as a 30 in another, because studies attract different cohorts. VTAC compares each study's students with how the same students did in their other studies and adjusts the scores: up where competition was stronger, down where it was weaker. The result is a scaled study score (to two decimal places, up to 55). A few studies have extra rules:

• Mathematics: the four maths studies are also scaled against each other (Specialist > Methods > General > Foundation), and the higher of the two scales is used.
• Languages: government policy adds 5 to each language's average. The boost is about 5 near a study score of 30 and smaller away from it.
• Small languages (≤ 10 eligible students) use the average language adjustment of the previous three years (never below zero), published as VTAC's "All small LOTEs" row.

VTAC publishes each study's scaled scores only at raw 20, 25, 30, 35, 40, 45 and 50, rounded to whole numbers. In between, ATARtools uses monotone cubic (PCHIP) interpolation. We tested 7 methods by hiding one published point at a time and predicting it from the others, across every study in every report 2014–2025 (~5,800 cases). PCHIP had the smallest error (0.43 scaled points vs 0.55 for straight lines, 0.45 Akima, 0.48 natural spline), and it never overshoots. Below raw 20 the curve is extended along its first segment.

Which year's scaling?

VTAC's rule: "VCE studies are always scaled in the year in which you undertook them." So:

• Studies you've finished (e.g. a Year 11 Unit 3–4 study in 2025) use that year's official report exactly.
• This year's studies have no report yet: it's published in December, after exams. Most calculators reuse last year's report. ATARtools uses a forecast: a weighted average of the last six official reports, where each year back counts half as much (weights 1, ½, ¼, ⅛…). History restarts where a new study design replaced a study. General Maths (from Further), Art Creative Practice, Art Making & Exhibiting and the Music studies only use 2023 onward, and Politics only 2025.

Why not just last year's report? Because averaging cancels one-off swings and the whole-number rounding of each report. We backtested both on every VTAC report: predict year Y using only the reports published before Y, then compare with what actually happened.

Root-mean-square error in scaled-score points at the seven published raw scores. The forecast is better in all nine years. Errors are much smaller for large studies (English, Methods, Biology…) than for small languages and VET programs. The 2026 forecast for every study, next to the official curves, is on the Scaling Explorer.

Step 3 — The aggregate, and every rule

From the VTAC ATAR & Scaling Guide 2026. To get an ATAR you must qualify for the VCE and have study scores in at least four permissible Unit 3–4 studies, including one from the English group. VCE Vocational Major studies don't receive study scores and can't count.

• Primary four: your highest scaled score in English, EAL, Literature or English Language, plus the next three highest permissible scaled scores. Only VCE studies and scored VCE VET can be in it.
• If you have more than six results, the six permissible results that give the highest ATAR are used. ATARtools tries every permissible combination, so it always finds that best one.

Increments (at most two)

Truncation: VTAC's own worked examples cut increments off at two decimal places rather than rounding them: 10% of 36.86 is 3.68, not 3.69, and 10% of 17.46 is 1.74. Their aggregates are the sums of those truncated values, so ATARtools truncates too.

Study-area groupings

To keep programs broad, at most two results from one study area can be in the primary four, and at most three can count in the aggregate (the third as an increment).

Business studies, sciences and the arts have no grouping: Accounting, Business Management and Economics can all count.

Equivalent studies

Some studies overlap so much that only one can count (the one that helps most): English and EAL; the same language at more than one level, i.e. Chinese First Language, Chinese Second Language, Chinese Second Language Advanced and Chinese Language, Culture & Society; and the First/Second Language versions of Indonesian, Japanese, Korean and Vietnamese. Different languages are not equivalent.

Studies from earlier years

A Unit 3–4 study finished in Year 10 or 11 counts like any other, but is scaled with its own year's report (VTAC's example: Dance done in Year 10 scaled with that year's report). Pick its year in the calculator.

Step 4 — Aggregate → ATAR

The ATAR is a percentile rank against the whole Year 12 age group, not just VCE students: an ATAR of 75.00 means you placed above 75% of it. VTAC orders every aggregate and publishes the cut-offs as a table in 0.05 steps. The top is 99.95, and anything under 30 is reported as "less than 30". ATARtools looks your aggregate up in the official 2025 table, all 1,400 bands, each checked against VTAC's PDF.

Why not average several years' tables, as we do for scaling? Because the table drifts in one direction for years at a time. For example, the aggregate needed for an ATAR of 50 fell from 102.4 (2016) to 91.6 (2022), then stabilised. In backtests, using last year's table beat 2- and 3-year averages (ATAR error 1.10 vs 1.34 and 1.62 for ATARs 50+). At the top, recent years barely move: above an ATAR of 90, last year's table was within 0.08 ATAR on average in 2023–2025.

A full worked example

This example is computed live by the same code that runs the calculator. One study was finished in 2025, so it uses the official 2025 scaling; the rest are this year's studies, so they use the forecast.

How sure is the result?

Next to the ATAR, the calculator shows an 80% likely range. It comes from re-running your whole calculation 300 times, each time randomly varying the things that can't be known yet, by amounts measured from history:

• This year's scaling vs the forecast, by study size. Studies with 5,000+ students: ±0.30 scaled points (1 SD). 1,500+: ±0.45. 500+: ±0.55. 150+: ±0.80. Smaller: ±2.0. Studies already finished: ±0.25 (VTAC's whole-number rounding and interpolation only).
• This year's aggregate→ATAR table vs last year's: ±0.30 aggregate points above ATAR 80, ±0.65 for 70–80, ±1.1 for 50–70, ±1.35 below 50.
• Estimated study scores: each estimate's own likely range, covering SAC moderation, assessments not yet done, and how hard this year's exams turn out (see exam difficulty).

Is it honest? Over 2019–2025, 83% of real scaling outcomes fell inside the forecast's 80% band, so it's calibrated, if anything slightly cautious. With study scores entered directly, the range is usually a few tenths of an ATAR point wide at the top end and wider lower down, where the table is steeper.

How it's tested

73 automated tests run on every change (npm test), and npm run accuracy re-runs every backtest on this page:

• Every number against the source: all 579 scaling rows for 2020–2025 (means, SDs, all seven points) against the VTAC PDFs; every one of the 1,400 ATAR bands checked at both edges; the assessment weights of all 93 studies against VCAA's 2026 contribution table; and every 2025 grade distribution (110 studies, 314 assessments: every grade count, mark range, mean and SD) against VCAA's page. That last check found and fixed one wrong figure (the Dance performance exam's SD).
• VTAC's own worked examples: all 11 "ATAR snapshots" in the 2026, 2027 and 2028 ATAR & Scaling Guides reproduced to the cent. That includes higher-education increments and truncation, and each guide's ATAR via the table of its data year.
• The rules, fuzzed: 4,000 random students (4–9 studies, mixed years, VET, VE3, HES, equivalent and grouped studies) checked against a separate brute-force program that tries every primary four and every pair of increments. Every best aggregate matches.
• The estimator: the closed-form check, the million-student simulation, monotonicity, determinism and calibration described above.
• Backtests: the scaling forecast, interpolation, ATAR-table and exam-difficulty choices, each re-proven against the official data (scaling 2014–2025, grade distributions 2019–2025).

What no calculator can know

• This year's scaling and ATAR table are only set in December, from this year's results. The forecast is the best available estimate, not a promise. A subject's scaling can move by a point or more in a single year, especially in small studies.
• Your school's exact moderation depends on how your school's group performs on this year's exam.
• Special circumstances: Derived Examination Scores, interrupted studies, interstate credit, IB and other qualifications are handled by VCAA/VTAC case by case.
• Course entry uses selection ranks (ATAR plus any adjustment factors), which differ by course and institution. The calculator's course finder shows VTAC's official lowest selection ranks from the December 2025 offer round. They change every year, and some courses select by folio, audition or interview instead. Check VTAC CourseSearch.

ATARtools is independent and not affiliated with VTAC or VCAA. Your official ATAR comes only from VTAC.

Sources

• VTAC, ATAR and Scaling Guide 2026 (also the 2027 and 2028 guides): qualification, primary four, increments, study areas, equivalent studies, worked examples.
• VTAC, Study area groupings fact sheet.
• VTAC, Scaling reports 2014–2025 and Aggregate to ATAR tables 2015–2025 (archive).
• VCAA, VCE Administrative Handbook — Score aggregation (study-score formula, statistical moderation).
• VCAA, Statistical moderation.
• VCAA, VCE and VET assessment summary: each graded assessment's contribution, 2026.
• VCAA, 2025 grade distributions for VCE graded assessments, and the per-study grade-distribution reports for 2019–2024 (e.g. 2024, 2023).
• VTAC, Lowest selection ranks — December 2025 domestic offers (course finder).
• VCAA, VCE Administrative Handbook 2026: equivalent studies (languages; English and EAL), higher-education studies.
• VCAA, Senior secondary completion and achievement information 2025 (school medians).