The broad mix of math on the SHSAT practice tests stayed steady between 2025 and 2026, even as the 2026 test added computer-adaptive delivery.
At Bell Curves, we analyzed 308 math questions from the six official practice tests released in 2025 and 2026. We tagged each with a primary skill and compared the two years. For 2026, we also mapped each question to New York State's math standards.
Half the questions, two categories
Across both years, algebra and proportional reasoning account for about half of the math questions.
Primary mathematical category | 2025 | 2026 |
|---|---|---|
Ratios, rates, proportions, and percentages | 23.7% | 29.4% |
Algebra, expressions, and equations | 25.4% | 21.6% |
Number sense and operations | 19.3% | 18.0% |
Probability and counting | 14.9% | 14.4% |
Geometry and measurement | 10.5% | 10.8% |
Statistics and data analysis | 6.1% | 5.7% |
Algebra + proportional reasoning | 49.1% | 51.0% |
2025: 114 questions across two practice tests. 2026: 194 classifiable questions across four practice tests. Each question was assigned one primary mathematical category.
Ratios and proportions rise from 23.7% to 29.4% while algebra falls from 25.4% to 21.6%, but their combined share barely moves.
Add number operations (18–19% in both years) and the three categories make up just under 70% of the math. Geometry, probability, and statistics split the rest.
The forms got shorter, from 57 math questions in 2025 to 50 in 2026, and the mix held steady anyway.
The math fits a fall eighth-grade exam
The SHSAT is taken in the fall of eighth grade, so we expected most of the math to come from Grade 6 and 7 standards. The data confirms it: about 90% of the 2026 questions we could tie to a standard fall in Grade 6 and 7, and fewer than 5% fall in Grade 8.
New York State's Next Generation Mathematics Learning Standards cover a much wider range of content than the test samples. Grades 6 and 7 cover rational-number operations, proportional relationships, expressions, equations, geometry, probability, and statistics. Grade 8 adds functions, transformations and congruence, irrational numbers, scientific notation, the Pythagorean theorem, and relationships between two variables.
This part of the analysis uses a subset of the data: the four 2026 practice tests (200 math questions), each tagged with a primary topic and, where one applied, a New York State standard. Some questions had incomplete source records and could not be tagged, and 22 were tagged to a topic without a specific standard code. That leaves 173 questions tied to a standard. The two 2025 tests are not yet mapped.
Standard level | Questions | Share |
|---|---|---|
Grade 6 | 70 | 40.5% |
Grade 7 | 86 | 49.7% |
Grade 8 | 8 | 4.6% |
Algebra I (high school) | 9 | 5.2% |
Questions tied to a standard | 173 | 100% |
Primary standard, 173 questions from the four 2026 practice tests.
The 8 Grade 8 questions cover linear equations, systems of equations, exponent rules, and angle relationships. None had a primary tag in functions, irrational numbers, scientific notation, or the Pythagorean theorem. Eight of the 9 Algebra I questions ask students to rearrange a formula.
The test leans heavily on what students should have mastered by the end of seventh grade. But foundational does not mean easy. A question can rest on Grade 6 ratios and still take several steps, a hidden relationship, and a tight clock to solve. Grade level says which math is on the test, not how hard each question is.
Curriculum coverage and test emphasis are different things. Students learn a broad range of math and are tested far more often on a smaller part of it.
A question's primary category doesn't tell the whole story
One category per question makes the distribution measurable, but many questions need more than one skill. A geometry question might use proportional reasoning to set up a relationship and algebra to find the unknown. A multistep word problem might need fractions, equation building, and interpretation of the result.
So the 21.6% of 2026 questions classified primarily as algebra is not the share that use algebra. Many geometry, proportion, and number questions need it too.
Our 2026 tags also note a secondary skill where a question needs one, and we haven't analyzed those yet. A fuller analysis would also separate questions that test a familiar procedure from those that ask students to recognize which procedure applies.
That last distinction is central to how we teach. Our Knowledge, Recognition, Execution (KRE™) method separates knowing the math, recognizing how a question uses it, and acting effectively under time pressure.
A miss in one category can come from any of the three, and a score by category can't tell us which.
Our read: with this much of the test built on Grade 6 and 7 math, results may depend less on advanced content than on recognizing familiar math presented in unfamiliar ways and executing it under time pressure. The practice tests don't prove that. Secondary-skill analysis and difficulty data would test it.
Digital response formats were already in place in 2025
The 2025 practice tests were already digital, and about one in five math questions asked for something other than a single choice. In 2026 that rises to about one in four.
Response format | 2025 (114 questions) | 2026 (200 questions) |
|---|---|---|
Single-choice multiple choice | 91 (79.8%) | 151 (75.5%) |
Fill-in-the-blank | 9 (7.9%) | 20 (10.0%) |
Equation editor | 4 (3.5%) | 11 (5.5%) |
Other interactive formats | 10 (8.8%) | 18 (9.0%) |
Any non-single-choice format | 23 (20.2%) | 49 (24.5%) |
Other interactive formats include drag-and-drop, multi-select, dropdowns, and number lines. By form, the non-single-choice share runs from 17.5% to 22.8% in 2025 and from 22% to 28% in 2026.
These formats ask for something the content categories don't capture. Typing an expression into an equation editor is a different task from picking it out of four choices, and a drag-and-drop question may need several classifications before a response is complete.
We looked at the possible time cost of those interactions in an earlier Bell Curves article, The Digital SHSAT's 20-Minute Problem.
Students should practice with the formats until operating the interface doesn't get in the way of solving the problem. For assessment professionals, the open issue is whether unfamiliarity with the interface creates score differences unrelated to the math being measured.
What adaptivity might change
The 2026 test added computer-adaptive delivery, which this analysis can't measure. Difficulty can change without the category changing: a unit-rate calculation becomes a multistep proportional relationship, an equation gets buried in a word problem, a geometry question chains algebra and proportions. We expect an adaptive test to draw on variations like these.
The practice tests don't say how the operational algorithm picks questions, or whether students at different performance levels will see the same mix of categories. If the concentration holds, algebra and proportional reasoning stay especially consequential. If the mix shifts by ability level, preparation based on overall frequency will need to adjust.
Students are better served by learning the variations within a skill than by memorizing one familiar question form.
What this means for instruction
Suppose a student answers 60% of algebra questions and 60% of statistics questions correctly. Algebra is about 22% of the test and statistics about 6%, so the same improvement in algebra is worth nearly four times as many points.
Now suppose a student already answers 90% of algebra questions correctly but only 50% of geometry questions. Geometry in this scenario offers more room to accumulate points, even though it appears less frequently.
Question frequency and current performance have to be read together.
Diagnostic reports should say more than "missed five algebra questions." Did the student lack the math? Did they not recognize how the question used it? Did they pick an inefficient approach, make a procedural error, or struggle with the response format?
Each answer calls for a different fix. A student who doesn't understand equations needs instruction in equations. A student who understands them but can't tell when to use them needs practice with question variations. A student who knows what to do but keeps running out of time needs better execution strategies. Treating all three as an algebra problem addresses none of them well.
For families, the takeaway is practical. Practice the question variations within algebra and proportional reasoning, and practice the response formats until the interface is routine.
The questions worth asking next
Three follow-ups would sharpen the picture:
Extend the standards mapping to the 2025 tests, so both years compare at the standards level.
Analyze the secondary skills tagged in 2026, to show how often a question depends on math outside its primary category.
Look at difficulty and response format together, to separate mathematical complexity from the demands of the interface.
The practice tests are the best public evidence available. They point the same way in both years: algebra and proportional reasoning, in more formats than before. In the 2026 questions we mapped, that math sits mostly in Grades 6 and 7.