The ACT’s Big Change: Shorter Sections, Greater Stakes
Less than two years after the SAT update, in which the test grew shorter and became section adaptable, the ACT is set to undergo its own revision. The details of the revision can be found here, but the crucial change is that the ACT is spinning off the science section into a standalone test, leaving the core test with only three sections: English, Math, and Reading. What’s more, the sections will be shorter. While students may appreciate a shorter test with fewer questions, the ACT is creating tremendous potential for students who invest in multiple sittings to achieve higher superscores through an ACT Superscore Strategy.
For those new to the game, superscores represent the best section scores across multiple tests. So, if a student scores well on Math during the first test and performs well on Reading and English during a second test (while doing poorly in Math that day), the superscore allows the student to present the best combination of section results. Most colleges evaluate superscores rather than a student’s single best test day. This practice works best when section scores remain relatively consistent, but data released by the ACT show that the shorter section lengths are causing greater variance — sometimes referred to as “noise.” In essence, noise allows for more lucky (or unlucky) bounces.
Variance, Noise, and the Power of Lucky Bounces
The ACT doesn’t publish the variance for each section, but they do report that the combined score across all three sections has a standard error of about ±1 point. When we pull on that thread, we can infer that the standard error on each individual section is roughly ±1.73 (methodology shown below) — and that’s a big deal. It means that 68% of the time, a student’s section score will fall within 1.73 points of their true average, and 95% of the time, within about 3.5 points above or below it. In other words, single section scores can swing wildly from test to test — even when a student’s underlying ability stays the same.
This “true average” — also called a student’s expected score — represents the score they would receive if they could take the same test an infinite number of times under identical conditions. It reflects their actual ability level, separate from random factors like which questions appear, test-day headspace, or a lucky guess.
The increased variance in the new ACT can actually work in students’ favor — especially at colleges that accept superscores. Each test sitting becomes a new roll of the dice, and over multiple sittings, students are more likely to encounter a test that plays to their strengths. Superscoring captures the best of these outcomes, making it possible for a student’s final score to exceed their true average — not because they got smarter, but because they played the game often enough to hit a high roll.
Why More Attempts Lead to Higher Superscores
According to the statistical principle that the expected maximum of repeated samples increases with the number of attempts, the more times a student takes a test with built-in variance, the more likely they are to outperform their average on at least one section. We applied this principle using the ACT’s own reported standard error of measurement. Assuming a student’s ability remains stable and each test introduces natural variance (±1.73 per section), we can estimate the expected benefit of superscoring across multiple sittings.
Let’s consider a motivated student who takes the ACT five times. Statistically, their best score on each section — English, Math, and Reading — will, on average, land 2.01 points above their true ability.
But here’s the thing: even taking the test a second time provides a meaningful benefit. A student who retests once will typically gain nearly a full point on their superscore. It takes a third, fourth, and fifth sitting to earn that second point — illustrating both the opportunity and the diminishing returns of the ACT superscore strategy.
But what about schools that don’t superscore? Roughly 15% of competitive universities only accept a student’s single best test day. In those cases, the gains are more modest. Even here, however, the ACT’s reported standard error of about ±1 point on the composite score means that repeat testers can still benefit. On average, it takes four sittings to improve a composite score by one point — still worthwhile, but less dramatic than the superscore advantage.
Below is table in which students can see how much their score will improve based on number of tests.
| Number of Sittings | Average Superscore Gain (points) | Average Composite Gain (points) |
|---|---|---|
| 1 | 0.00 | 0.00 |
| 2 | 0.98 | 0.56 |
| 3 | 1.46 | 0.85 |
| 4 | 1.78 | 1.03 |
| 5 | 2.01 | 1.16 |
| 6 | 2.17 | 1.26 |
| 7 | 2.29 | 1.34 |
| 8 | 2.39 | 1.41 |
But of course, students shouldn’t rely on variance alone to boost their scores. Effective test prep drives scores much higher than the occasional lucky bounce ever could. What the new ACT is revealing, however, is just how much students can further improve their outcomes by sitting for more tests. In Part II of this series, we’ll examine this trend from the perspective of colleges. And in Part III, we’ll zoom out to explore why this may be a feature — not a bug — of the modern testing landscape and why the ACT may be more concerned with their bottom line than testing integrity.
Methodology
To calculate the expected gains from repeated ACT testing, we used statistical simulations based on the ACT’s own reporting of test score variance.
The ACT reports that a student’s composite score — the average of the English, Math, and Reading sections — has a standard error of measurement (SEM) of approximately ±1 point. Since that composite is based on the average of three sections, and assuming the errors across sections are independent (as measurement error typically is), we can work backward to estimate the SEM of a single section using the formula for the SEM of an average of independent variables:
The standard error of an average of three independent values is smaller by a factor of √3.
So, if the composite SEM is ±1, the SEM for each individual section is approximately ±1.73:
\[\text{Composite SEM} = \frac{\text{Section SEM}}{\sqrt{3}} \Rightarrow \text{Section SEM} \approx 1 \times \sqrt{3} \approx 1.73\]Using this estimate, we modeled each section score as a normally distributed variable centered around the student’s true ability, with a standard deviation (SEM) of 1.73. For each number of test sittings (from 1 to 8), we simulated 1,000,000 students taking the test that many times. For each simulated student, we recorded:
- Their best score on each section (English, Math, and Reading), across all sittings
- Their superscore composite, calculated as the average of those three best section scores
- Their best composite score from a single sitting (for comparison, representing schools that do not superscore), modeled with an SEM of ±1
