GAMSAT percentiles

An overall score of 56 to 58 sits around the 50th percentile; 63 is roughly the top quarter; 73 and above is roughly the top 2%. Modelled The table below is calculated from the distribution parameters in the only peer-reviewed analysis of GAMSAT scores: an overall mean near 57, with a standard deviation between 7.07 and 7.53 across the ten years measured. We use the wider end of that range, which gives slightly more conservative estimates.

The model assumes a normal distribution. That assumption is ours— the paper reports no distribution shape, and one preparation company describes the curve as normal with a slight positive skew. If that is right, the estimates in the upper tail are optimistic.

Every figure here is a range for two reasons, both evidenced. Percentiles are computed within a single sitting, so the curve moves when the cohort moves. And the underlying model is a normal curve fitted to a decade-old Australian cohort — close enough to be useful, not close enough to quote to one decimal place.

Two candidates have published transcriptions of the private chart ACER gave them. They are the closest thing to official percentile data in the public domain, and they do not match our model.

A section score of 70 is close to the ceiling of Section 1’s distribution but only around the 81st–86th percentile in Section 2. ModelledThis is the single most under-reported fact about GAMSAT scoring, and applying the overall curve to a section score — which every other calculator we found does — is a material error.

The three sections have genuinely different distributions Peer-reviewed:

Three consequences worth acting on. A 55 is a below-average result in Section 1 but close to the middle of the field in Section 3. Identical section scores are not equally impressive, so “I got 65 in every section” describes three quite different performances. And because Section 3 has both the widest spread and double weight in ACER’s overall, it is where a given improvement moves your overall score furthest.

The most defensible “am I competitive?” statistic in the public domain is not a percentile at all. Mercer et al. define a typical entry score as an overall of at least 60 with at least 50 in every section, and report that 36.9% of candidates achieved one in 2014 Peer-reviewed. Note that it is a compound threshold: an overall 62 built on a 47 in Section 1 does not qualify.

We compared every public percentile table we could find. They do not agree, and the pattern of disagreement is informative.

There is a reason the numbers cluster. Four of the tables we checked publish the same 56–58 anchor, and one company’s figures fit a normal curve with a mean of about 57.5 and a standard deviation of about 7.1 almost exactly — essentially the parameters we use ourselves. They are not independent data. They appear to be the same model, published with more confidence than the method supports.

We are doing the same modelling. The difference is that we say so, publish ranges, and show you where the model loses to the real charts.

The curve differs; the score does not. ACER equates every sitting so that scaled scores are directly comparable ACER-official, but percentiles are computed within each sitting, so a smaller and differently-composed September cohort produces a different curve from the same paper difficulty. No published data quantifies the difference, and any claim that one sitting is easier is unsourced.

No, and neither can anyone else.ACER transforms raw marks onto the scaled 0–100 range using item response theory calibrated against everyone in that sitting, and a live paper also contains unscored trial questions. No public raw-to-scaled conversion exists. Our free practice test reports a raw percentage and stops there, on purpose.