Percentiles cause many misunderstandings and risks of error.

The announcement of the percentile table of high school graduation exam scores according to some traditional admission combinations in 2025 is receiving much attention from candidates, parents and universities.

This is a positive step in making admission data transparent. However, the use of percentiles to convert scores between admission groups is causing many misunderstandings, leading to the risk of bias and unfairness in university admissions.

What is percentile?

Percentile is a statistical index that shows the relative position of a candidate in the distribution of test scores of a particular combination. For example, a candidate scoring 90th percentile in combination A00 means that the candidate scored higher than 90% of the candidates taking the same combination.

However, percentiles only have internal meaning, within each combination. Percentiles cannot be used to compare different combinations such as A00, C00 or D01... because each combination has a very different way of setting questions, score distribution and candidate types.

An important but often overlooked principle: percentiles are only meaningful when the data come from candidates who took both groups. Only when the same person took all the subjects in both groups with real effort can a correlation be established between the two score distributions.

On the contrary, if using data from two independent groups of candidates, for example group A00 and group D01, then any conversion lacks scientific basis. Because the two groups may have different abilities, learning orientations and test-taking goals, direct comparison is not possible.

Risk of skewed results

A common reality today is that many candidates only focus on reviewing the main admission combination, and only take the other subjects to "enough combination", without setting a goal for admission. This leads to the situation of doing the test "just for the sake of it", causing low scores and pulling down the entire score range of that combination.

As a result, there are candidates who score average but jump to a high percentile, not because they are good, but because many others do not try hard to do the test. If these percentiles are used to convert to other combinations, it will lead to a virtual benchmark score, reflecting the wrong level of reality.