AP Statistics Exam Format 2027: Fully Digital, 42 MCQs, and 4 FRQs

The May 2027 AP Statistics Exam is fully digital in Bluebook. Section I contains 42 multiple-choice questions with four choices each in 90 minutes. Section II contains four 10-point free-response questions in 90 minutes. Each section contributes 50% of the score.

AP Statistics exam format at a glance

The revised assessment structure
Feature	Revised May 2027 format	Preparation consequence
Delivery	Fully digital Bluebook	Type all final MCQ and FRQ responses; use scratch paper for planning and calculations
MCQ	42 questions, four choices, 90 minutes	Practice revised counts and two shared-prompt sets
FRQ	Four questions, 10 points each, 90 minutes	Practice design, analysis, inference, and multi-focus roles
Weight	50% MCQ, 50% FRQ	Balance recognition/computation with written reasoning
Calculator	Statistical calculator expected; Desmos built in; approved handheld allowed	Know one dependable workflow and current policy
Reference	Printed and digital reference information	Learn selection, conditions, and interpretation beyond formula lookup

Previous format versus the revised May 2027 format

A current-format practice test should not mix the old counts with a 2027 label
Feature	Previous exam	Revised exam
Course units	Nine commonly taught units	Five units
MCQ count	40	42
Choices	Five per MCQ	Four per MCQ
Shared-prompt sets	Not the revised stated structure	Two three-question sets: probability and regression
FRQ count	Six	Four
FRQ points	Four per question in the older design	Ten per question
2026 delivery	Hybrid digital with paper FRQ booklet	Fully digital beginning May 2027
Required topics	Included geometric, GOF, and regression-slope inference	Those topics removed from revised required scope

How the 42-question multiple-choice section is structured

The section tests all five units and all four statistical practices. Four choices reduce the old distractor count, but the questions still require method selection, design reasoning, probability, inference, interpretation, and technology-supported calculations.

Two sets contain three questions that share a prompt. One set focuses on probability, random variables, and probability distributions. The other focuses on regression analysis. A shared prompt can include a table, graph, model, study description, or output used by several items. Definitions and assumptions should remain consistent across the set.

MCQ scoring is based on correct answers; College Board’s general timing guidance says there is no extra deduction for an incorrect or unanswered item. Students should therefore make an informed selection on every question while preserving time to reach the full section.

A 42-item current-format practice blueprint

This table is a balanced original practice blueprint, not a prediction of the operational order or exact unit allocation. It shows 42 distinct ways the revised framework can become a four-choice question.

Forty-two formats for a revised practice set
Item	Question form	What the item can ask	Evidence of mastery
1	Investigative question	Select the question that defines cases, population, variables, and measurable comparison.	A statistical question anticipates variability and can be answered by a specified data process.
2	Categorical table	Read counts, marginal proportions, or conditional proportions from a one- or two-way table.	The denominator matches the named group; unequal group sizes are compared with proportions.
3	Categorical graph	Interpret or choose a bar chart, segmented bar chart, or mosaic-style comparison.	Axes and group proportions support the claim without using a quantitative display for category codes.
4	Quantitative graph	Read a dotplot, stemplot, histogram, boxplot, or cumulative distribution.	The response addresses shape, center, spread, and unusual features with units.
5	Comparative distribution	Choose the best comparison of two or more quantitative distributions.	The description uses direct comparative language and a common visual scale.
6	Center and spread	Calculate or interpret mean, median, quartiles, IQR, range, variance, or SD.	The statistic is connected to distribution shape and resistance, not treated as isolated output.
7	Outlier rule	Apply 1.5-IQR fences or evaluate the effect of an unusual value.	The calculation uses ordered quartiles and the summary choice responds to skew or outliers.
8	Transformation	Determine how adding or multiplying changes mean, median, quartiles, SD, IQR, or z-scores.	Location and spread rules are separated; negative scale factors do not create negative SD.
9	Standardized score	Compute or compare z-scores across distributions.	The result is interpreted as standard deviations above or below the relevant mean.
10	Sampling plan	Identify or evaluate SRS, stratified, cluster, systematic, or multistage sampling.	The random mechanism and represented population are explicit.
11	Survey error	Diagnose undercoverage, nonresponse, response bias, wording, or measurement error.	The proposed repair addresses the systematic source rather than only increasing n.
12	Experimental structure	Identify treatments, experimental units, factors, levels, response, random assignment, and control.	The design supports the stated causal comparison and holds competing differences constant.
13	Block or matched pair	Select an efficient design using prior similarity or repeated measures.	Random assignment occurs within blocks or order is randomized within participants.
14	Scope of inference	Choose whether a result supports population generalization, causation, both, or neither.	Random sampling and random assignment are credited for their distinct roles.
15	Simulation	Evaluate a random-digit assignment, repetition rule, or simulated estimate.	Equally likely labels reproduce the target probability and the recorded statistic answers the question.
16	Probability rule	Use complement, addition, multiplication, or a Venn/table representation.	Overlapping events are not double-counted and disjoint is not confused with independent.
17	Conditional probability	Calculate from a tree or table and interpret the conditioning group.	The restricted denominator and direction of conditioning are correct.
18	Independence	Check a conditional equality or multiplication relationship.	The response distinguishes a model assumption from empirical evidence of association.
19	Random-variable distribution	Complete or interpret values and probabilities that sum to one.	Expected value and variability are properties of repeated chance outcomes.
20	Expected value	Compute a long-run average, fair price, or net gain.	Costs are included and the noninteger expectation is not presented as a guaranteed single outcome.
21	Binomial conditions	Decide whether a count follows a binomial model.	Fixed n, two outcomes, independent trials, and constant p all hold.
22	Binomial probability	Calculate exactly, at most, at least, or a complement.	The event notation matches the calculator bounds and includes the correct endpoint.
23	Binomial center/spread	Use np and square root of np(1-p).	Mean and SD are interpreted for the count variable with the model assumptions visible.
24	Normal probability	Standardize a value and find a left, right, or middle area.	A sketch confirms the tail and the final probability is between zero and one.
25	Normal percentile	Use inverse normal reasoning and transform back to the original scale.	The percentile has the original units and the correct below/above interpretation.
26	Sampling distribution of p-hat	Find center, standard error, shape, or a probability for a sample proportion.	p, p-hat, and p0 are not interchanged, and large-counts/10% conditions are checked.
27	Proportion interval	Select, calculate, or interpret a one-proportion z interval.	Observed successes/failures and long-run confidence language are used.
28	Proportion sample size	Determine n from confidence level, margin, and a planning estimate.	The conservative 0.5 choice is used when needed and n is rounded upward.
29	Proportion test	Write hypotheses, compute z or p-value, and choose a conclusion.	The null standard error uses p0 and the conclusion states evidence rather than probability of H0.
30	Two proportions	Distinguish interval and test formulas for p1-p2.	The interval is unpooled; the equality test pools under H0; subtraction order is defined.
31	Chi-square expected counts	Compute E from marginal totals or check the approximation condition.	Expected rather than observed cell counts control the chi-square reference model.
32	Chi-square statistic/conclusion	Combine cell contributions, degrees of freedom, p-value, and association language.	Direction is read from residuals, and causation is not inferred from the test alone.
33	Sampling distribution of x-bar	Use center mu, standard error sigma/sqrt(n), and normal or CLT shape.	Population spread is separated from variability of sample means.
34	t model	Identify degrees of freedom, critical value, or the consequence of unknown sigma.	s replaces sigma and heavier tails reflect estimated standard error.
35	One-mean interval	Construct or interpret a t interval for mu.	Randomness, independence, shape/outlier checks, and contextual confidence all appear.
36	One-mean test	Calculate a t statistic or evaluate a p-value conclusion.	Hypotheses name mu and practical importance is not inferred from significance alone.
37	Paired t	Recognize before-after or matched measurements and analyze differences.	One difference per pair is used; n counts pairs; conditions are checked on differences.
38	Two-mean t	Compare independent quantitative groups.	Separate sample summaries and design-based scope are preserved without forcing equal variances.
39	Scatterplot/correlation	Describe linear association or interpret r under transformations and unusual points.	Direction, form, strength, and limitations are stated without causal language.
40	Regression line	Interpret slope/intercept or make a prediction.	Units, observed x-range, and conditional prediction wording are correct.
41	Residual/r-squared	Calculate a residual, read a residual plot, or interpret explained variation.	Residual is observed minus predicted; r-squared concerns response variability, not points on a line.
42	Influence/mixed prompt	Evaluate leverage and influence or combine regression with another statistical practice.	The fit is compared with and without a point, and valid data are not deleted automatically.

Forty-two original rehearsal designs

These are prompt designs rather than copied exam questions. Each row supplies a hypothetical context, the distractor logic, and the statistical standard a keyed option must satisfy.

Build four-choice questions that test reasoning rather than trivia
Item	Form	Original context	Plausible distractors	Keyed-response standard
1	Investigative question	A district wants to compare travel modes across grade bands.	A nonstatistical question with no variable and a causal question unsupported by the planned survey.	The keyed option must demonstrate this standard: A statistical question anticipates variability and can be answered by a specified data process.
2	Categorical table	A two-way table classifies device type and successful login.	A marginal proportion, a reversed conditional proportion, and a denominator from the full table.	The keyed option must demonstrate this standard: The denominator matches the named group; unequal group sizes are compared with proportions.
3	Categorical graph	Four course sections report different sample sizes and preferences.	Raw counts that ignore unequal section sizes and a quantitative average of category codes.	The keyed option must demonstrate this standard: Axes and group proportions support the claim without using a quantitative display for category codes.
4	Quantitative graph	Two distributions of completion time are shown on common axes.	A description that changes axis scale, omits units, or lists groups without comparison.	The keyed option must demonstrate this standard: The response addresses shape, center, spread, and unusual features with units.
5	Comparative distribution	Boxplots compare wait times for three support channels.	A claim based only on medians, only on ranges, or a visual feature not shown by boxplots.	The keyed option must demonstrate this standard: The description uses direct comparative language and a common visual scale.
6	Center and spread	An ordered sample includes a high extreme value.	A mean or SD choice pulled toward the outlier when resistant summaries are required.	The keyed option must demonstrate this standard: The statistic is connected to distribution shape and resistance, not treated as isolated output.
7	Outlier rule	Quartiles are provided for a skewed delivery-time distribution.	An IQR computed as Q1-Q3, a fence with the wrong sign, or the range mislabeled as IQR.	The keyed option must demonstrate this standard: The calculation uses ordered quartiles and the summary choice responds to skew or outliers.
8	Transformation	Temperatures are converted from Celsius to Fahrenheit.	An SD that adds 32, stays unchanged after scaling, or becomes negative under a negative multiplier.	The keyed option must demonstrate this standard: Location and spread rules are separated; negative scale factors do not create negative SD.
9	Standardized score	Two exams use different means and standard deviations.	A raw-score comparison, a z-score with reversed subtraction, or a percentile conclusion without scale context.	The keyed option must demonstrate this standard: The result is interpreted as standard deviations above or below the relevant mean.
10	Sampling plan	A district roster is separated by school level before selection.	A cluster interpretation, a convenience sample, or a design that samples only one stratum.	The keyed option must demonstrate this standard: The random mechanism and represented population are explicit.
11	Survey error	A voluntary online poll reports a very narrow margin of error.	The belief that large n removes self-selection, undercoverage, or response bias.	The keyed option must demonstrate this standard: The proposed repair addresses the systematic source rather than only increasing n.
12	Experimental structure	A study compares two review schedules with different study time.	A causal claim when study time is confounded, or random sampling described as treatment assignment.	The keyed option must demonstrate this standard: The design supports the stated causal comparison and holds competing differences constant.
13	Block or matched pair	Every participant tries two interfaces in a randomized order.	An independent-groups test that ignores pairing, or a fixed order that creates a practice effect.	The keyed option must demonstrate this standard: Random assignment occurs within blocks or order is randomized within participants.
14	Scope of inference	A random sample is observed without random treatment assignment.	Both causation and generalization claimed merely because the word random appears once.	The keyed option must demonstrate this standard: Random sampling and random assignment are credited for their distinct roles.
15	Simulation	Two-digit random numbers simulate a 23 percent event.	Twenty-two, twenty-four, or a nonuniform set of labels assigned to the 23 percent outcome.	The keyed option must demonstrate this standard: Equally likely labels reproduce the target probability and the recorded statistic answers the question.
16	Probability rule	Events overlap in a student-activity survey.	Adding probabilities without subtracting overlap, subtracting the intersection twice, or assuming disjointness.	The keyed option must demonstrate this standard: Overlapping events are not double-counted and disjoint is not confused with independent.
17	Conditional probability	A screening table gives flagged and true-error counts.	P(flagged\|error) substituted for P(error\|flagged), or the grand total used as denominator.	The keyed option must demonstrate this standard: The restricted denominator and direction of conditioning are correct.
18	Independence	Two event probabilities and their intersection are supplied.	Disjoint, independent, and complementary events treated as interchangeable.	The keyed option must demonstrate this standard: The response distinguishes a model assumption from empirical evidence of association.
19	Random-variable distribution	A net-gain game has three possible monetary outcomes.	Gross payout used instead of net gain, or expected value described as a guaranteed single result.	The keyed option must demonstrate this standard: Expected value and variability are properties of repeated chance outcomes.
20	Expected value	A warranty plan has cost, payout, and no-claim outcomes.	Probabilities that do not sum to one or an omitted cost in every outcome.	The keyed option must demonstrate this standard: Costs are included and the noninteger expectation is not presented as a guaranteed single outcome.
21	Binomial conditions	Twenty components each pass or fail inspection.	A model justified only by yes/no outcomes while constant p or independence fails.	The keyed option must demonstrate this standard: Fixed n, two outcomes, independent trials, and constant p all hold.
22	Binomial probability	A batch question asks for at least two failures.	P(X>2) instead of P(X>=2), a complement missing P(X=1), or wrong binomial bounds.	The keyed option must demonstrate this standard: The event notation matches the calculator bounds and includes the correct endpoint.
23	Binomial center/spread	A binomial model gives n and p for successful uploads.	Variance reported as SD, q omitted, or the mean interpreted as a guaranteed integer count.	The keyed option must demonstrate this standard: Mean and SD are interpreted for the count variable with the model assumptions visible.
24	Normal probability	Package weights follow a stated normal model.	The wrong tail, an unstandardized area, or use of the empirical rule for an exact calculator question.	The keyed option must demonstrate this standard: A sketch confirms the tail and the final probability is between zero and one.
25	Normal percentile	A percentile threshold must be returned in original units.	A z percentile left on the standard scale or transformed with subtraction instead of addition.	The keyed option must demonstrate this standard: The percentile has the original units and the correct below/above interpretation.
26	Sampling distribution of p-hat	Repeated samples estimate a population support proportion.	Center p-hat instead of p, standard error using n rather than sqrt(n), or no large-counts check.	The keyed option must demonstrate this standard: p, p-hat, and p0 are not interchanged, and large-counts/10% conditions are checked.
27	Proportion interval	A survey reports successes, failures, and sample size.	Null counts used in an interval, probability assigned to a fixed p, or an interval for observations.	The keyed option must demonstrate this standard: Observed successes/failures and long-run confidence language are used.
28	Proportion sample size	A poll specifies confidence level and maximum margin.	Rounding n downward, using p*=0 without evidence, or confusing confidence with power.	The keyed option must demonstrate this standard: The conservative 0.5 choice is used when needed and n is rounded upward.
29	Proportion test	A sample proportion is compared with a policy claim.	p-hat used in the null SE, hypotheses written about p-hat, or a p-value called P(H0).	The keyed option must demonstrate this standard: The null standard error uses p0 and the conclusion states evidence rather than probability of H0.
30	Two proportions	Two interfaces are evaluated in independent random samples.	Pooled SE used for an interval, subtraction order switched, or dependent groups treated as independent.	The keyed option must demonstrate this standard: The interval is unpooled; the equality test pools under H0; subtraction order is defined.
31	Chi-square expected counts	Marginal totals are supplied for a two-way table.	Observed count substituted for expected, marginal totals multiplied without dividing by N, or wrong cell margins.	The keyed option must demonstrate this standard: Expected rather than observed cell counts control the chi-square reference model.
32	Chi-square statistic/conclusion	A chi-square output includes statistic, df, and p-value.	Correlation language, a causal conclusion, or the claim that every cell is individually significant.	The keyed option must demonstrate this standard: Direction is read from residuals, and causation is not inferred from the test alone.
33	Sampling distribution of x-bar	A skewed population is sampled with a large n.	Raw-population SD reported for x-bar, no square root n, or an unjustified exact-normal claim.	The keyed option must demonstrate this standard: Population spread is separated from variability of sample means.
34	t model	A sample mean and SD are provided with a small sample size.	z used because n is numeric, df set equal to n, or sample SD treated as known population sigma.	The keyed option must demonstrate this standard: s replaces sigma and heavier tails reflect estimated standard error.
35	One-mean interval	Battery-life data include n, mean, SD, and a t critical value.	A normal interval for individual batteries, no shape check, or confidence described as sample coverage.	The keyed option must demonstrate this standard: Randomness, independence, shape/outlier checks, and contextual confidence all appear.
36	One-mean test	A fill-weight sample is tested against a target mean.	An alternative chosen after data, p-value accepted as H0 probability, or practical importance inferred automatically.	The keyed option must demonstrate this standard: Hypotheses name mu and practical importance is not inferred from significance alone.
37	Paired t	The same patients are measured before and after treatment.	Two independent samples formed from paired columns or n counted as twice the number of patients.	The keyed option must demonstrate this standard: One difference per pair is used; n counts pairs; conditions are checked on differences.
38	Two-mean t	Independent classes report means, SDs, and sample sizes.	A pooled equal-variance method assumed without need or a paired method used for unrelated classes.	The keyed option must demonstrate this standard: Separate sample summaries and design-based scope are preserved without forcing equal variances.
39	Scatterplot/correlation	A scatterplot includes one extreme explanatory value.	Correlation called causal, the extreme x-value ignored, or influence inferred solely from residual size.	The keyed option must demonstrate this standard: Direction, form, strength, and limitations are stated without causal language.
40	Regression line	A fitted score equation is used inside the observed range.	Extrapolation far beyond the data, slope interpreted without units, or intercept forced into context.	The keyed option must demonstrate this standard: Units, observed x-range, and conditional prediction wording are correct.
41	Residual/r-squared	A residual plot and r-squared are shown together.	Residual sign reversed, r-squared called percent of points on line, or pattern ignored because r is high.	The keyed option must demonstrate this standard: Residual is observed minus predicted; r-squared concerns response variability, not points on a line.
42	Influence/mixed prompt	Removing one high-leverage point changes the fitted slope.	Automatic deletion, leverage confused with residual, or no comparison of fits with and without the point.	The keyed option must demonstrate this standard: The fit is compared with and without a point, and valid data are not deleted automatically.

The four revised free-response questions

Four questions, each worth 10 points
Question	Official role	What a complete response may need
1	Primarily Practices 1 and 2	Formulate an investigative question; define population and variables; propose or critique sampling, experiment, measurement, and scope
2	Primarily Practices 3 and 4	Select and carry out an analysis; read displays or output; interpret the result and limitations
3	Inference through Practices 2, 3, and 4	Parameter, method, conditions, calculation, p-value or interval, conclusion, and design-based scope
4	Multiple content areas through Practices 2, 3, and 4	A connected investigation with several methods, intermediate results, claim evaluation, and limitations

The role descriptions are more useful than treating every FRQ as an identical 22.5-minute calculation. Question 1 can award substantial value for design language, Question 2 for analytical interpretation, Question 3 for formal inference, and Question 4 for integration. Students should practice the specific response verbs: identify, describe, calculate, compare, justify, explain, interpret, and evaluate.

Anatomy of the four FRQ roles

Thirteen recurring components across four revised questions
Question	Component	Required work	Strong evidence	Common failure
Q1	Formulate Questions	Define the population, cases, variables, relationship or comparison, and a question that expects variation.	A measurable question aligned with the proposed data	Vague words, undefined population, or a question that asks for a fixed fact
Q1	Collect Data	Design or critique sampling, random assignment, controls, blocks, measurement, and bias.	A reproducible random process and correct scope of inference	Calling haphazard selection random or confusing sampling with assignment
Q1	Design conclusion	Explain what the design can establish and which limitations remain.	Causation and generalization linked to the correct random mechanisms	Claiming a large sample alone supports both
Q2	Analyze Data	Choose displays and summaries; calculate probabilities, statistics, residuals, or model quantities.	A method that matches variable type and the investigative question	Pasting calculator output without statistical meaning
Q2	Interpret Results	Connect graphical and numerical evidence to a claim with units and context.	A direct answer that distinguishes sample evidence from population truth	Describing separate groups without comparing them
Q2	Evaluate model	Use residuals, simulation, distribution shape, or other evidence to judge model adequacy.	A named pattern or condition tied to the method	Saying the model is good because a single summary is large
Q3	Parameter and hypotheses	Define the population quantity and, for a test, write H0 and Ha in the question's direction.	Notation and prose identify the same parameter	Hypotheses about a sample statistic or equality placed in Ha
Q3	Procedure and conditions	Name the interval/test and verify randomness, independence, counts, shape, or expected counts.	Method-specific checks with values from the prompt	Generic statements such as n is large without the relevant criterion
Q3	Calculation	Show estimate, standard error, critical value or statistic, probability, and result.	Correct pooled/unpooled choice, tail, df, and sensible rounding	Opaque calculator output with wrong model or denominator
Q3	Conclusion	Use confidence or p-value evidence to answer in the population context.	Strength of evidence, direction, parameter, population, and limitations	Probability-of-H0 language or proof from failure to reject
Q4	Multi-focus setup	Scan all parts, define variables and supplied results, and identify independent entry points.	Organized work that preserves access to later parts	Stopping the entire question after one blocked calculation
Q4	Connected analysis	Combine methods from several units while using one coherent investigation.	Each calculation answers its part and feeds later reasoning transparently	Using unrelated formulas because the prompt contains many numbers
Q4	Argument and limitations	Judge a claim, practical magnitude, design, uncertainty, and possible alternative explanations.	A conclusion proportional to the evidence	Turning statistical significance into practical or causal certainty

Four original mock FRQ roles

These are instructional format examples, not released or secure College Board questions.

Mock role 1: Formulate and collect

An original prompt asks whether students using a structured note-review schedule retain more concepts after four weeks than students using their usual approach. The first part should ask for a precise investigative question naming eligible students, treatment, comparison, response measure, and follow-up time. A response such as “Does the new schedule work?” is too vague because work has no operational measure.

The design portion can block students by prior diagnostic score and randomly assign within blocks to structured review or usual review. Total assigned study time, content, assessment, and instructor contact should be kept comparable. The response must distinguish random assignment, which supports causation for participants, from random sampling, which would support generalization to the population represented by the frame. If volunteers are used, that limitation remains even after excellent assignment.

A 10-point question can distribute value across the investigative question, design details, randomization, control, bias or confounding, and scope. A student should answer every part in short labeled paragraphs. Decorative prose is less useful than an executable random process and a precise statement of what the experiment can establish.

Mock role 2: Analyze and interpret

An original prompt displays two distributions of form-completion time under different interface layouts. Students may compare shape, center, spread, and unusual values; select resistant summaries if one distribution is skewed; calculate a standardized value; or interpret a residual from a fitted relationship. The analysis must use common units and direct comparison, not separate descriptions pasted beside one another.

A later part can provide calculator output and ask what it shows. The response should choose only relevant values, connect them to the graph, and describe practical magnitude. If a high outlier raises the mean but leaves the median nearly unchanged, that is evidence about resistance, not a reason to delete the observation. If a residual plot curves, a linear model misses structure even if correlation is large.

The format rewards a connected explanation: what feature appears, how the calculation quantifies it, and what conclusion follows in context. A screenshot of output or a list of values is not a statistical interpretation. The typed answer should separate calculations from the sentence that tells the reader what they mean.

Mock role 3: Inference

An original inference prompt gives a random sample of users and asks whether a success proportion exceeds a policy target. A complete response defines p, writes H0:p=p0 and the correctly directed Ha, names a one-proportion z test, checks random sampling, the 10 percent condition, and null expected successes and failures, then calculates the test statistic and p-value.

The conclusion compares the p-value with the prespecified alpha and states the strength of evidence that the population proportion exceeds the target. It does not say the p-value is the probability the null is true. If the question instead asks for an interval, sample success and failure counts replace null counts in the condition check, and the standard error uses p-hat rather than p0.

Ten points allow scoring of reasoning as well as arithmetic. Students should show parameter, method, conditions, calculation, and conclusion even when a calculator supplies the numerical test. This structure also makes partial knowledge visible when one value is entered incorrectly.

Mock role 4: Multi-focus investigation

An original multi-focus prompt follows a school study from question formulation through data collection, descriptive analysis, probability modeling, and inference. One part might ask whether assignment supports causation, another might compare distributions, another might calculate a probability under a model, and a final part might evaluate a confidence interval alongside a practical benchmark.

The key format skill is continuity. Variable definitions and subtraction order must remain stable; an intermediate result should be carried forward transparently; a design limitation should still constrain the final claim. If one calculation is blocked, independent later parts should still receive a response. Scanning all parts before writing protects these entry points.

A strong ending separates three judgments: the statistical evidence, the estimated magnitude and uncertainty, and the scope allowed by sampling and assignment. This is what makes the fourth role multi-focus rather than merely longer. It tests whether separate course tools can support one coherent argument.

Twenty command verbs that control response format

Read the verb before deciding how much calculation or prose is needed
Command	What it asks	Complete response	Failure to avoid
Identify	Name the requested object, method, feature, or value.	Give the precise statistical term and enough context to distinguish it.	A long essay that never names the requested item.
Define	State what a symbol, variable, parameter, event, or residual means.	Include population/cases, variable, units, and subtraction order when relevant.	A formula with symbols that remain undefined.
Describe	Report the important statistical features shown by data or a design.	For distributions use shape, center, spread, unusual features, and context; for design name the mechanism.	A list of calculator values with no pattern.
Compare	State similarities and differences directly.	Use comparative words and a common scale; connect magnitude and direction to groups.	Two separate descriptions that leave the reader to compare.
Calculate	Produce a numerical result using an appropriate method.	Show formula or setup, substitution, calculator result, rounding, and units as needed.	An unexplained number or a method incompatible with the parameter.
Determine	Use supplied information to reach the requested value or decision.	Show the chain of statistical reasoning that makes the determination reproducible.	A guess based on a visual impression when a calculation is required.
Estimate	Use sample data or simulation to approximate a population or chance quantity.	Name the estimator and distinguish its observed value from the target parameter.	Calling the estimate the exact population truth.
Construct	Build an interval, display, sampling plan, experiment, model, or probability distribution.	Include every defining component and verify validity conditions.	A partial object that cannot be implemented or interpreted.
Interpret	Translate a statistic, interval, p-value, slope, residual, or r-squared into context.	Name population/variables, direction, units, uncertainty, and design limits.	Restating digits without meaning or assigning probability to H0.
Explain	Give the statistical mechanism or reason behind a result.	Connect evidence to the relevant principle in complete sentences.	Repeating the claim with because but adding no reason.
Justify	Supply evidence that proves a method, condition, design choice, or conclusion is warranted.	Use numbers from the prompt or a named design/statistical principle.	Saying valid, random, or normal without the controlling evidence.
Evaluate	Judge the adequacy of a claim, method, model, design, or conclusion.	Identify strengths, failed conditions, bias, practical importance, and scope.	Calling a result good because it is statistically significant.
State hypotheses	Write H0 and Ha about a population parameter.	Put equality in H0 and match the alternative direction to the question before data.	Hypotheses about a sample statistic or an alternative chosen after seeing results.
Check conditions	Verify the assumptions that support a probability, interval, test, or model.	Use method-specific numerical and design evidence.	A universal checklist pasted without relevance.
Conclude	Answer the investigative claim at the strength supported by evidence.	Use confidence or p-value language, direction, population, context, and design scope.	Accepting H0 as true or claiming causation from observation.
Predict	Use a fitted model for a specified explanatory value.	Substitute correctly, preserve response units, stay within the data range, and acknowledge residual variation.	Treating the prediction as guaranteed or extrapolating without warning.
Simulate	Represent a chance process and repeat it to estimate a probability or distribution.	Define labels, trial, repetition, statistic, and interpretation.	Many repetitions of a label assignment that does not match the target probability.
Design	Create a data-production process suited to the question.	Specify population/units, selection or assignment, controls, measurement, replication, and scope.	Using the word random without an executable mechanism.
Distinguish	Separate two ideas that appear similar but have different meanings.	Use definitions and a contrasting example, such as sampling versus assignment or parameter versus statistic.	Giving two synonyms rather than a real difference.
Support a claim	Use data, design, and uncertainty as an argument.	Cite the relevant graph, estimate, interval or p-value, effect magnitude, and limitations.	Selecting only favorable evidence or exceeding the design's scope.

A forty-point instructional FRQ rubric

This model assigns ten instructional checkpoints to each revised role so students can diagnose missing evidence. It is not an official scoring guideline and should not be used as a future AP score conversion.

Forty diagnostic checkpoints across a four-question practice set
Practice question	Point	Evidence	Use
Q1	1	Defines the population and cases	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	2	States a measurable investigative question	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	3	Classifies or defines variables	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	4	Names the study type	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	5	Gives an executable random sampling mechanism	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	6	Gives an executable random assignment when experimental	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	7	Uses control/comparison appropriately	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	8	Addresses bias or confounding	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	9	States generalization scope	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q1	10	States causal scope	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	1	Selects an appropriate display or summary	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	2	Calculates the requested statistic	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	3	Uses units and denominators correctly	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	4	Describes distribution shape	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	5	Compares center directly	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	6	Compares spread directly	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	7	Identifies unusual features	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	8	Evaluates a model or pattern	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	9	Connects evidence to the question	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q2	10	States a contextual interpretation and limitation	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	1	Defines the population parameter	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	2	Writes H0 and Ha or interval target	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	3	Names the correct inference procedure	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	4	Checks randomness	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	5	Checks independence or 10 percent	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	6	Checks counts, shape, or expected counts	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	7	Calculates estimate and standard error	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	8	Calculates statistic/p-value or interval	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	9	Makes the statistical decision	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q3	10	Concludes in context with correct scope	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	1	Organizes variables and supplied information	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	2	Uses design to limit later claims	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	3	Completes the first content calculation	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	4	Interprets the first result	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	5	Selects a second method	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	6	Checks the second method's conditions	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	7	Carries intermediate results forward	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	8	Evaluates a claim with evidence	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	9	Discusses practical magnitude or alternative explanations	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.
Q4	10	Synthesizes a final multi-focus conclusion	Award in an original practice rubric only when the response contains the named statistical evidence; operational College Board scoring follows the actual prompt.

Section weighting and score interpretation

MCQ and FRQ each contribute 50% of the composite result. The revised four FRQs are 10 points each, but a website should not invent how a particular future raw composite maps to scores 1 through 5. Operational score setting and form difficulty are handled by College Board.

A practice score should be broken into content and practice categories. A student can earn many calculation points yet lose design and interpretation points, or recognize MCQ answers but struggle to build a complete FRQ argument. Equal section weights mean neither pattern can be ignored.

Use format evidence rather than an unofficial AP cutoff
Diagnostic	What to record	How to respond
MCQ content	Correct by five revised units	Rebuild the lowest unit using contrast problems, then return to mixed sets
MCQ practice	Errors in question formulation, collection, analysis, or interpretation	Practice the statistical action across several content topics
FRQ components	Points or elements lost for parameter, design, condition, calculation, or conclusion	Rewrite only the failed component, then recombine a full response
Digital execution	Navigation, notation, calculator, scratch-to-screen, and review errors	Rehearse the mechanical step separately before another full timed set

Five-unit coverage in the revised exam

Official percentage ranges, not fixed item quotas
Unit	Exam weighting	Central content
1. Exploring One-Variable Data and Collecting Data	20%-30%	Investigative questions, categorical/quantitative displays and summaries, random sampling, and experimental design
2. Probability, Random Variables, and Probability Distributions	15%-25%	Simulation, probability, random variables, binomial and normal distributions
3. Inference for Categorical Data: Proportions	15%-25%	Sampling distributions, intervals/tests for proportions, and chi-square homogeneity/independence
4. Inference for Quantitative Data: Means	10%-20%	Sampling distributions, t intervals/tests, paired means, and independent means
5. Regression Analysis	10%-20%	Scatterplots, correlation, least-squares models, residuals, prediction, leverage, and influence

What fully digital means for the response format

Both multiple-choice selections and free-response answers are submitted through Bluebook. There is no paper FRQ booklet for the standard May 2027 administration. Students continue to receive scratch paper for planning and calculations and a printed reference booklet, while reference information is also available in Bluebook.

The digital response should be readable rather than ornamental. Use conventional keyboard notation, define symbols, place formulas on separate lines when needed, and follow calculations with complete contextual sentences. The symbols menu includes common statistical symbols, and College Board says FRQs will minimize unnecessary symbolic-entry demands.

Fully digital does not mean calculator-free. Built-in Desmos is available, and an approved handheld graphing calculator may also be used. The detailed Bluebook testing guide owns device setup and troubleshooting so this page can remain focused on question architecture.

Format and interface should be evaluated separately. The format defines counts, timing, score weight, question roles, and required statistical practices; the interface defines how a student navigates, enters notation, and submits. A practice PDF can imitate the first category but cannot fully rehearse the second. Conversely, clicking through Bluebook without solving aligned statistical questions rehearses mechanics but not the assessment. A complete preparation cycle therefore uses current-format questions inside a realistic digital workflow, then scores method selection, conditions, calculations, and interpretation as well as completion.

This distinction also keeps resource reviews honest. An older released FRQ can remain excellent for inference reasoning while being unsuitable as a full 2027 section simulation. Label the content value and the structural mismatch separately. Students then retain authentic statistical thinking without learning the wrong question count, response medium, or removed topic priority. The final rehearsal should still use all 42 current MCQs, four revised FRQ roles, and both 90-minute limits under realistic calculator, reference, scratch-paper, and typed-response conditions from beginning through final submission, review, and a recorded error audit.

Twenty examples of clear typed statistical notation

These examples show conventional forms a student can enter and define. They are not a requirement to copy one exact style.

Mathematical meaning must remain unambiguous on screen
Purpose	Readable entry	Meaning	Avoid
Sample proportion	$p-hat = 138/240 = 0.575$	Use p-hat for the observed sample statistic and p for the population proportion.	Do not type p=0.575 unless the population value is known.
Sample mean	$x-bar = 41.2 hours$	Define x-bar as the sample mean and preserve units in interpretation.	Do not use mu for a number calculated from the sample.
Null and alternative	$H0: p = 0.60; Ha: p > 0.60$	Equality belongs in H0 and the alternative direction comes from the question.	Do not write hypotheses about p-hat.
One-proportion z	$z = (p-hat – p0) / sqrt[p0(1-p0)/n]$	The null value controls the test standard error.	Do not substitute p-hat into the null SE.
t statistic	$t = (x-bar – mu0) / (s/sqrt(n)), df=n-1$	The sample SD and degrees of freedom identify the t model.	Do not mix s with a small-sample z reference without justification.
Confidence interval	$0.575 +/- 1.96(0.0319) = (0.512, 0.638)$	Show estimate, critical value, standard error, and endpoints.	Do not describe the endpoints as containing 95% of observations.
p-value	$p-value = P(Z >= 1.79 \| H0) = 0.0368$	The conditioning phrase under H0 preserves the meaning.	Do not type P(H0 true)=0.0368.
Paired difference	$d = after – before; mu_d is the population mean difference$	Define subtraction order before interpreting signs.	Do not analyze paired columns as independent samples.
Two proportions	$parameter: p_A – p_B$	Name which population is first so interval signs have meaning.	Do not switch order between calculation and conclusion.
Expected count	$E = (row total)(column total)/(grand total)$	Use marginal totals under independence.	Do not replace expected counts with observed counts in the condition check.
Chi-square contribution	$(O-E)^2/E$	The contribution is nonnegative; use O-E separately for direction.	Do not infer over- or underrepresentation from the squared contribution alone.
Regression line	$y-hat = b0 + b1x$	The hat marks a predicted response, not an observed value.	Do not interpret b1 as causal without randomized evidence.
Residual	$residual = y – y-hat$	Positive means observed response lies above the fitted prediction.	Do not reverse the subtraction.
r-squared	$r^2 = 0.64$	Interpret as 64% of sample response variation explained by the fitted linear relationship.	Do not say 64% of points lie on the line.
Conclusion after rejection	$Because p-value < 0.05, reject H0; the data provide evidence that …$	State evidence for Ha in context.	Do not write accept Ha as proven.
Conclusion after failure to reject	$Because p-value > 0.05, fail to reject H0; the data do not provide convincing evidence that …$	Use insufficient-evidence language.	Do not claim the null is true or groups are identical.
Confidence statement	$We are 95% confident that the population mean difference lies between …$	Name population, parameter, units, and subtraction order.	Do not assign 95% probability to the already fixed parameter.
Simulation estimate	$estimated probability = successes / repetitions$	Define what counts as a simulated success and the statistic recorded.	Do not claim repeated simulation proves an invalid random assignment.
Standard error	$SE(x-bar)=s/sqrt(n) for estimated mean uncertainty$	Distinguish estimator variability from individual-value spread.	Do not interpret SE as the SD of raw observations.
Scope of inference	$Random assignment supports causation; random sampling supports generalization.$	State exactly which design features are present.	Do not use the words random study without naming selection or assignment.

Twenty decisions for evaluating a practice resource

Separate durable content from outdated structure
Resource feature	Format judgment	Correct use
A five-choice practice test is labeled 2027.	Its answer-choice format is outdated. Revised MCQs have four choices.	Use it only for content practice after remapping; do not use it to rehearse guessing or pacing.
A practice test contains 40 MCQs.	It reflects the older count, not the revised 42-question section.	Add two original aligned items or use a revised full set for timing.
A practice test contains six FRQs.	It reflects the prior free-response structure.	Select four by revised role or use a current four-question set.
A student prepares to handwrite all FRQs in 2027.	The May 2027 exam is fully digital; paper-only rehearsal misses response-entry mechanics.	Continue scratch work by hand but practice final typed responses in Bluebook style.
A student expects no paper at all.	College Board says scratch paper and a printed reference booklet remain available, with reference also in Bluebook.	Rehearse a deliberate screen-scratch-type workflow.
A shared prompt has three probability questions.	This matches one of the revised MCQ set structures.	Build one event definition or probability model and reuse it consistently.
A shared prompt has three regression questions.	This matches the other identified revised MCQ set structure.	Read graph/output once and preserve variable roles, units, and fitted equation across items.
An MCQ asks for a p-value interpretation.	The format can test conceptual reasoning without a long calculation.	Choose the option describing a tail probability under H0, not the probability H0 is true.
An FRQ gives calculator output.	Students can be assessed on choosing and interpreting a method rather than reproducing arithmetic.	Name the parameter, conditions, relevant output, and contextual conclusion.
An FRQ part says justify.	A bare numerical answer does not satisfy the command.	State the statistical principle, design feature, condition, or evidence that proves the response.
An FRQ part says calculate.	Show setup or enough work to make the source of the result clear.	Use calculator output responsibly and preserve meaningful digits until the final value.
An FRQ part says interpret.	Translate the quantity into population, variables, units, direction, and uncertainty.	Do not restate the number using different punctuation.
A question uses a topic removed from the revised framework.	It may be historical enrichment but should not be presented as core May 2027 coverage.	Label it clearly and prioritize current five-unit material.
A resource lists nine units.	It maps the older course organization.	Translate durable topics into the five revised units and remove no-longer-required content.
A resource says geometric distribution is required.	College Board removed it from the revised required framework.	Do not devote core 2027 practice time to it unless a teacher assigns enrichment.
A resource teaches regression-slope inference as an exam target.	That inference unit was removed from the revised required course.	Keep interpretation, residuals, r-squared, prediction, leverage, and influence.
A student memorizes every printed formula.	The reference information reduces recall demands but does not choose a method or write conclusions.	Practice identifying parameters, conditions, and interpretations without being told the procedure.
A response uses p-hat and p interchangeably.	Digital notation is not the problem; the parameter/statistic distinction is.	Define conventional keyboard notation once and use it consistently.
A student uses only Desmos and has not rehearsed it.	Built-in availability does not create fluency.	Practice lists, distributions, intervals/tests where supported, and graph interpretation before timed work.
A student brings a handheld calculator.	An approved handheld remains allowed in addition to built-in Desmos.	Verify the current policy and know one dependable workflow; a second device does not replace preparation.

AP Statistics exam-format glossary

Forty terms for reading questions and resources accurately
Term	Meaning	Why it matters in the revised format
Section	A timed major part of the exam with its own question type and score weight.	The revised exam has a 90-minute MCQ section and a 90-minute FRQ section.
Item	One scored multiple-choice question or a discrete task within an assessment.	The revised MCQ section contains 42 items.
Stem	The question text that defines the context, data, and task before the answer choices.	Read the requested statistical action before being pulled into story details.
Option	One of the four answer choices in a revised MCQ.	A keyed option must be best supported, not merely partly true.
Distractor	A plausible incorrect option built from a common conceptual or computational error.	Review why each distractor fails to expose the precise misconception.
Keyed answer	The option designated as correct for the item.	A sound practice resource can explain the key from definitions, conditions, or calculations.
Shared prompt	One context, graph, table, or output used by several MCQs.	The revised section includes two three-question shared-prompt sets.
Standalone item	An MCQ with its own independent prompt.	Most practice sets should mix standalone items with the two set structures.
Free response	A question requiring the student to generate rather than select an answer.	The revised section contains four 10-point questions typed in Bluebook.
Subpart	A labeled task inside an FRQ, often dependent on or connected to other parts.	Scan all subparts so an early block does not hide later accessible work.
Command verb	The word that specifies the response action, such as calculate, justify, or interpret.	Match the amount of computation and prose to the verb.
Statistical practice	A recurring action: formulate questions, collect data, analyze data, or interpret results.	FRQ roles are described through these practices rather than content alone.
Content unit	One of five revised groups of required statistical knowledge.	Percentage ranges describe coverage across the exam, not a fixed item order.
Question role	The principal practice function assigned to a revised FRQ position.	Q1 design, Q2 analysis, Q3 inference, and Q4 multi-focus require different rehearsals.
Point	A unit in an instructional or operational scoring guideline.	Each revised FRQ has 10 points, but specific operational guidelines depend on the actual question.
Raw score	The unconverted performance accumulated across scored components.	A future raw-to-AP-score conversion should not be invented by a practice website.
Composite	The weighted combination of section results before reporting the 1-5 AP score.	MCQ and FRQ each contribute 50% in AP Statistics.
AP score	The reported score from 1 through 5 after College Board's scoring process.	It is not a simple permanent percentage grade.
Bluebook	College Board's application for digital AP testing and practice previews.	Beginning May 2027, both AP Statistics sections are completed in it.
Response field	The on-screen area where a student types an FRQ answer.	Separate formulas and contextual conclusions so scoring evidence remains visible.
Symbols menu	A Bluebook aid for entering commonly used mathematical and statistical symbols.	Use it when helpful, while conventional keyboard notation remains acceptable when defined.
Scratch paper	Physical paper supplied for planning and calculations during the digital exam.	Label work by question and transfer the final reasoning to Bluebook.
Reference information	Printed and digital formulas/tables supplied for AP Statistics.	It supports arithmetic but does not choose a method, check conditions, or interpret results.
Built-in Desmos	The calculator available inside Bluebook.	It can supplement or replace a handheld workflow when the student has practiced it.
Approved handheld	A calculator permitted by the current AP calculator policy.	It remains allowed in addition to built-in Desmos; verify the live policy.
Parameter	The population quantity targeted by an interval, test, or investigation.	Define it before hypotheses or a conclusion so the response has a stable target.
Statistic	A sample quantity used as evidence about a parameter.	Digital notation must keep p-hat distinct from p and x-bar distinct from mu.
Condition	A design or distribution requirement supporting a method's reference model.	A complete FRQ cites method-specific numerical and contextual evidence.
Standard error	The estimated or known sampling-distribution spread of an estimator.	It appears in intervals and tests but changes formula by parameter and null model.
Critical value	A distribution cutoff used with standard error to set confidence margin.	Its size depends on confidence level and, for t, degrees of freedom.
Test statistic	The standardized distance between observed evidence and a null value.	Show numerator direction and correct null standard error before quoting calculator output.
p-value	A tail probability under H0 for the observed statistic or one more extreme.	Its interpretation is a high-value FRQ component and common MCQ distractor source.
Pooled estimate	A common proportion estimate used under H0:p1=p2.	It belongs in the two-proportion test SE, not the ordinary interval SE.
Unpooled estimate	Separate group proportions used to estimate unrestricted interval variability.	It belongs in the two-proportion confidence interval.
Degrees of freedom	A reference-distribution index after estimated constraints are considered.	Use n-1 for one-sample t and (r-1)(c-1) for chi-square association.
Residual	Observed response minus model-predicted response.	It can be calculated in an MCQ or interpreted and diagnosed in an FRQ.
Leverage	Potential influence from an explanatory value far from the center of x.	A high-leverage point can have a small residual and still affect the line.
Influence	The actual change in fitted results after removing an observation.	Compare fits rather than deleting a point automatically.
Scope of inference	The population and causal reach justified by data collection and assignment.	It should appear in design FRQs and in conclusions across the exam.
Practical importance	Whether an estimated effect is large enough to matter in context.	It is evaluated separately from statistical significance.

Calculator and reference information in the format

College Board expects a graphing calculator with statistical capabilities or a permitted nongraphing calculator that has the required statistical capabilities. Built-in Desmos appears in Bluebook, and the current policy allows approved handhelds. Check the live policy before the exam because a static article should not freeze a model list.

Reference information supports probability distributions, descriptive statistics, sampling distributions, confidence intervals, tests, and related work, but it does not decide which formula applies. Students must still distinguish parameters from statistics, intervals from tests, pooled from unpooled standard errors, paired from independent data, and association from causation.

The best format rehearsal uses the same calculator plan and supplied reference information while answering 42 four-choice MCQs and four typed FRQs. Practice content from older years can be remapped, but full-section timing should use the revised counts.

Frequently asked questions about the AP Statistics format

What is the AP Statistics exam format in 2027?

A fully digital Bluebook exam with 42 MCQs in 90 minutes and four 10-point FRQs in 90 minutes; each section is worth 50%.

How many answer choices are on each MCQ?

Four in the revised format, down from five in the previous design.

Are there question sets?

Yes. Two sets contain three questions sharing a prompt: one probability/random-variables/distributions set and one regression set.

How many FRQs are there?

Four, each worth 10 points in the revised description.

What does FRQ 1 assess?

Primarily Formulate Questions and Collect Data, the first two statistical practices.

What does FRQ 2 assess?

Primarily Analyze Data and Interpret Results.

What does FRQ 3 assess?

Inference through the collection, analysis, and interpretation practices.

What does FRQ 4 assess?

Multiple content areas through collection, analysis, and interpretation.

Is the exam paper or digital?

Beginning in May 2027, both sections and all responses are completed in Bluebook.

Is scratch paper allowed?

College Board’s revision FAQ says students continue to receive scratch paper for planning and calculations.

Is a reference sheet provided?

Yes. Printed reference information remains available, and reference information is also accessible in Bluebook.

Can students use a handheld calculator?

Yes, an approved handheld with statistical capabilities may be used in addition to built-in Desmos.

Does digital notation need perfect typesetting?

No. It must be unambiguous. Define conventional keyboard notation and use the symbols menu when helpful.

Do old exams match the new format?

They contain durable content but may use 40 MCQs, five choices, six FRQs, old units, and removed topics. Remap before timing.

Is the exact 2027 date part of the format?

No. Date and structure are separate facts. See the date-status page for the unresolved subject day.

Official sources

College Board sources checked July 18, 2026
Official source	What it verifies
AP Statistics assessment	42 MCQs, four FRQs, 90 minutes per section, 50/50 weighting, and the four FRQ roles
AP Statistics revisions	Fully digital May 2027 delivery, four-choice MCQs, shared-prompt sets, removed topics, and course changes
AP Statistics course page	Five units and official unit weighting ranges
AP calculator policy	Calculator expectations, built-in Desmos, and handheld rules
Reference information for AP exams	Printed/digital reference materials and subject-specific details
Bluebook	Official digital practice and testing application