Fitting a raw-to-scaled score curve without an official table (Digital SAT)
College Board does not publish a raw-score-to-scaled-score conversion table for the Digital SAT. On the old paper test they did, one table per form. On the digital test the conversion is module-dependent, so there is no single table to publish. That is the whole problem in one sentence, and it is a genuinely interesting one to build around. This is a short write-up of how I approximated it for a…
The College Board does not provide a raw-score-to-scaled-score conversion table for the Digital SAT since there is no single, official table due to the test's adaptive, module-dependent nature. This is where the challenge lies in creating an approximation. To estimate the conversion, I developed a method that is explained below.
The Digital SAT consists of two modules per section, and a student's performance on Module 1 determines whether Module 2 is easier or harder. Additionally, unscored pretest questions in each module further complicate the situation. Despite these complexities, two sources of data are publicly available: published practice tests with raw-to-scaled score ranges per section, and the annual total-group report which shows the score distribution.
The approximation process involves three steps. First, I collected (raw, scaled) pairs from official practice tests to form a scatter plot for each section. Next, I fitted a monotone curve through these points, ensuring the curve maps more correct answers to higher scores. This was done using piecewise-linear interpolation between anchors with linear extrapolation at both ends, which avoids overshooting and maintains clarity.
Finally, I conducted a sanity check by converting the fitted section scores into a total distribution and comparing it to the published total-group distribution. If the fitted curve implies an implausible percentage of students scoring above a certain threshold, the curve is adjusted accordingly.
The resulting output provides a total score along with an error bar, which ranges from plus or minus 30 to 60 points. The fit reveals that the last 100 points on the scale require more questions to be answered correctly compared to the first 200 points, reflecting the test's design to concentrate the distribution in the middle and compress the tails. This explains some student frustration as certain questions can create plateaus, such as a 1380 that doesn't change despite additional correct answers.
There are two percentile tables: user group percentiles compare you to students in your graduating class, while nationally representative percentiles compare you to all U.S. students, including non-test takers. The same score may have different percentiles depending on the comparison group. It is crucial to use the correct percentile for the intended purpose, such as college applications, which typically use user group percentiles.
In the development of this calculator, I made two mistakes: lazy-loaded images causing issues and inaccurate percentile numbers. I have since fixed these issues, and the calculator is now a static page with inlined data tables and clear explanations, providing a transparent and reliable tool for students to understand their Digital SAT scores.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.