How closely do TOEFL mock test scores match real scores?

How is the ETS practice test different from the real test?

The difference ETS states is one of format. Practice Test 1 is a practice test aligned with tests given from January 21, 2026, and the document is a version adapted to be taken on paper (Practice Test 1 cover, checked 2026-10-30). The real test is taken on a computer at a test center or on a home computer (Home Edition) (ETS Test Content, checked 2026-10-30).

ETS describes TOEFL TestReady as a personalized prep portal that brings practice tests, courses, guides, books, performance insights and progress tracking together in one place (ETS Test Preparation, checked 2026-10-30). Neither the information pages we checked nor the ETS "TOEFL iBT Technical Manual" (the Technical Manual below) say how TestReady practice test scores correspond to real scores.

Why can your score be different on the real test?

Among the score calculation and scoring methods ETS describes, we see the five below as places where mock test scores and real scores can diverge. Treating them as sources of difference is our view; the ETS documents we checked (including the Technical Manual) do not separately explain differences between practice test scores and real scores. In each paragraph, the first part is ETS fact, and our view is marked separately with "Our view:". All page numbers below follow the PDF page order.

Adaptive modules and score calculation: ETS says it designed Reading and Listening as two-stage adaptive (the difficulty of the second module depends on the answers in the first module) (ETS "TOEFL iBT Test: 2026 Update Test Blueprint and Specifications Document", the Specifications Document below, page 3, note 5, checked 2026-10-30). The ETS "Updated TOEFL iBT Test Overview" (the Test Overview below) explains that performance on the first part is used to select the content of the second part, matching task difficulty to the test taker's level (Test Overview page 3). The Technical Manual says Reading and Listening scores reflect both the number of questions answered correctly in the two modules and the difficulty of the modules on the test taker's path, and that raw scores are converted to reported scores through equating (a statistical procedure that adjusts for difficulty differences between test forms) for each test form (Technical Manual pages 28 and 34, checked 2026-10-30). The same pages state that reported scores are not the same as the number or percentage of questions answered correctly. The formula and conversion tables are not published. Our view: without evidence that a mock test selects the second module the same way and goes through the same equating, the same number of correct answers can produce a different score.

Scoring methods: The table on page 2 of the Specifications Document lists the Speaking tasks and the Writing tasks Write an Email and Write for an Academic Discussion as AI Scored, and the 10 Build a Sentence items in Writing and the Reading and Listening items as Machine Scored. The Technical Manual says Email, Discussion and the Speaking tasks are scored with ETS's AI scoring engine together with human scoring, and that human raters assign the score for responses where the AI's confidence is low or that are difficult (Technical Manual pages 35 and 39). Each test taker's responses are scored by two or more raters (page 40), and Build a Sentence is scored 1 or 0 per item by matching against the correct answer (pages 34 and 39). Our view: a mock test's scoring is separate from the ETS scoring system, so there is no basis for assuming the same response will get the same score.

Rounding to the nearest 0.5: The total score is the average of the four section bands (1 to 6) rounded to the nearest 0.5 (ETS Understanding TOEFL Scores and Test Overview page 21, checked 2026-10-30). Our view: near a rounding boundary, even a small difference shows up as 0.5 in the total score (example 1 below).

Unscored questions: The Technical Manual says Reading modules can contain 15 unscored questions and Listening modules may include 12 (Technical Manual page 28). It does not say which questions they are. The Specifications Document says their purpose is usually to collect item performance data for later use in scored questions (Specifications Document page 3, note 1). Our view: unscored questions do not count toward the score, so they are unlikely to explain a score difference. However, the number of questions you answer on the real test and the number counted in your score can differ.

Measurement error: The Technical Manual gives the standard error of measurement (SEM) for the total score as 0.32 (1 to 6 band), measured with field test data from one test form, and explains that the chance the true score falls within 1 SEM above or below the observed score is about 68% (Technical Manual page 48). Our view: real test scores also carry measurement error, so we do not read a 0.5 difference in total score between two tests directly as a difference in ability.

How should you read the gap between mock test and real scores?

When you look at a gap, put the four section bands side by side before looking at the total score, and do not read a mock test score as a forecast of your real score. The following are examples we wrote for illustration.

Example 1

• Wrong reading: "Mock test total 5.0, real total 5.5 (both on the 1 to 6 band), so all four sections went up by 0.5."
• Right reading: "A rise of 0.5 in just one of the four sections can change the total score from 5.0 to 5.5."
• Why: The total score is the average of the four sections rounded to the nearest 0.5. In the examples ETS gives, an average of 5.125 becomes a total of 5, and an average of 5.25 becomes a total of 5.5 (1 to 6 band; Understanding TOEFL Scores; Test Content, checked 2026-10-30). Section bands of 5, 5, 5 and 5.5 average 5.125 (total 5); if one more section rises, giving 5, 5, 5.5 and 5.5, the average is 5.25 (total 5.5) (our calculation). The rounding rule and the 0 to 120 comparable total score ranges are in a table in How do TOEFL 1 to 6 scores compare with 0 to 120 scores?.

Example 2

• Wrong reading: "My mock test total was 4.5 (1 to 6 band), so I will get 4.5 on the real test too."
• Right reading: "The mock test 4.5 is a reference score from that mock test; the real total is set separately by the modules and scoring of the real test."
• Why: The ETS documents we checked contain no table matching practice test scores to real scores (see "What we did not find in ETS documents" below). Real Reading and Listening scores are equated for each test form, reflecting both the number of correct answers in the two modules and module difficulty (Technical Manual pages 28 and 34).

Reference estimates are not ETS scores; they are scores Unlock Prep calculates using criteria that reflect ETS scoring criteria. Unlock Prep's AI judges Writing responses on each task's criteria. Unlock Prep does not yet publish figures comparing these estimates with real test scores (unlockprep.com mock test page, checked 2026-10-30).

What has ETS disclosed, and what has it not?

ETS has disclosed how the practice test and the real test differ in structure and how real test scores are calculated, but the documents we checked (including the Technical Manual) do not say how practice test scores correspond to real scores.

What ETS has disclosed (checked 2026-10-30; pages = PDF pages)

• Practice Test 1 is aligned with tests given from January 21, 2026, is not an exact copy, and was adapted for a paper format (cover).
• Reading and Listening are two-stage adaptive, and performance on the first part selects the content of the second part. Writing and Speaking are linear (all test takers who receive the same test form complete the same tasks) (Specifications Document page 3, note 5; Test Overview page 3).
• Reading and Listening scores are equated for each test form, reflecting the number of correct answers in the two modules and module difficulty, and the summed Writing and Speaking raw scores are converted using weighted equipercentile linking (Technical Manual pages 28, 34 and 35).
• Email, Discussion and Speaking use an AI scoring engine together with human scoring, and a human scores the response when the AI's confidence is low. Build a Sentence is scored 1 or 0 by matching against the correct answer (Specifications Document page 2 table; Technical Manual pages 34, 35 and 39).
• Reading modules can contain 15 unscored questions, and Listening modules may include 12 (Technical Manual page 28).
• The total score is the average of the four sections rounded to the nearest 0.5, and the standard error of measurement for the total score is 0.32 (1 to 6 band) (Understanding TOEFL Scores; Test Overview page 21; Technical Manual page 48).

What we did not find in ETS documents

• A table matching Practice Test 1 or TestReady practice test scores to real scores, or the degree of agreement between them
• A table converting raw scores to the 1 to 6 bands: the method is disclosed (a conversion table is built for each test form; Technical Manual page 34), but the table is not published.
• Which questions are unscored
• Cutoff scores, such as how well you must do in the first module to be routed to the upper module
• What has changed since the note still in the Specifications Document saying small changes may be made before the test formally launches (page 3)

What differs between sources

• Reading and Listening raw score range: Specifications Document page 2 table and Technical Manual page 41 Table 8 = 0 to 35 / Test Overview page 21 Table 1 = 0 to 30. This article does not calculate with raw scores.

This article did not compare agreement figures published by companies or blogs. Because the main pages of Practice Test 1 are images, we compared only the cover sentences as text.

What should you use mock test scores for?

Our view is that you should not read a mock test score as a forecast of your real score, but use it as a record for finding which section and which questions cost you points.

• For each attempt, write down the four section bands (1 to 6) separately. If you record only the total, you end up reading a change in one section as a change across all four, as in example 1.
• Get used to the new format with ETS practice materials first. Criteria for choosing a mock test are in How to choose a TOEFL mock test for the 2026 TOEFL iBT.
• Check your target score in the notices of the institutions you are applying to. ETS says it does not set a passing score and that each institution sets its own requirements (Understanding TOEFL Scores, checked 2026-10-30). How to read institution requirements is covered in What TOEFL score do I need?.