Calibration study

How close are Bandwise bands to real results? We are measuring it, in the open.

Every band on this site is an estimate. Listening and Reading bands come from the mark-to-band tables IELTS publishes for its own papers, applied to tests those tables were never measured on; Writing and Speaking bands come from an AI examiner. This page is the study that turns those estimates into measurements: what it collects, how far it has got, and exactly what we are allowed to say at each step.

timed answers shared by consenting learners: 0
untimed or practice-mode answers, stored but excluded: 0
learners who entered an official result: 0
of them paired with a timed sitting within 14 days: 0

Why the bands are estimates

IELTS measures every version of its test. Nobody has measured ours yet.

A real IELTS paper gets its own mark-to-band table because versions differ in difficulty. Bandwise tests are original, so the published tables are the closest anchor we have, and a result on the day can differ from our estimate by up to a band.

The study closes that gap in phases. First, enough timed answers to every question to see which questions behave like real test items and which need replacing. Then, learners who share a recent official result, so each form’s table can be corrected against the real scale rather than borrowed from it. Later, examiners marking our Writing and Speaking material blind, to measure how closely the AI examiner agrees with them.

Each phase has a sample size fixed in advance and a pass mark it must reach on that full sample. A phase that fails does not stop the product; it stops the claim. Until a step is earned, the wording on results pages stays on the row below.

Where the study stands: Today

0 timed answers shared by consenting learners and 0 official results entered so far. No form has yet reached the Phase 1 sample of 200 timed answers per item and 250 persons, so every band on the site is still an estimate from the published anchors.

The claim ladder

What may be said about a band at each stage, and nothing more. The marked row is the one the data so far support.

What Bandwise may say about its band estimates at each stage of the study; the current stage is marked
Stage reached	Listening and Reading bands	Writing and Speaking bands	Difficulty labels
TodayCurrent stage	“Estimated band from IELTS’ published anchors; may differ by up to a band”	“Estimated band (AI), practice only; beta; typical range ±1 band”	Hidden or marked “target, not measured”
After Phase 1	add “items validated on N attempts, alpha = x”	unchanged	“Measured: easier/harder than our reference form”
After Phase 2	“Calibrated against N official results; typically within half a band”	unchanged	Relative to the reference form’s official anchor
After Phase 3	unchanged	“Agreement with examiners: QWK x, y% within one band, measured on N scripts (date)”	unchanged
Phase 4	per-form “based on N attempts”	refreshed figures	refreshed

No stage permits “official”, “real exam questions”, “certified by IELTS examiners” or a guaranteed band. Stages are earned per form, and the wording in the product moves up a row only after a monthly report has recorded the decision; the row marked here is the one the data collected so far would support.

Progress

How far each form has got

Counted from the study store as this page loaded. Nothing here identifies a learner: the store holds a random code per browser, and this page reads only totals.

Phase 1: timed answers per form

A form needs 200 timed answers on its thinnest item, from 250 different persons, before its item statistics are reported. Each bar shows the count against that threshold.

Academic Practice Test 000100 · Listening
ListeningAcademicCollecting
ac-000100 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000100 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000100 · Reading
ReadingAcademicCollecting
ac-000100 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000100 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000101 · Listening
ListeningAcademicCollecting
ac-000101 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000101 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000101 · Reading
ReadingAcademicCollecting
ac-000101 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000101 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000102 · Listening
ListeningAcademicCollecting
ac-000102 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000102 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000102 · Reading
ReadingAcademicCollecting
ac-000102 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000102 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000103 · Listening
ListeningAcademicCollecting
ac-000103 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000103 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000103 · Reading
ReadingAcademicCollecting
ac-000103 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000103 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000104 · Listening
ListeningAcademicCollecting
ac-000104 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000104 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000104 · Reading
ReadingAcademicCollecting
ac-000104 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000104 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000105 · Listening
ListeningAcademicCollecting
ac-000105 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000105 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000105 · Reading
ReadingAcademicCollecting
ac-000105 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000105 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000106 · Listening
ListeningAcademicCollecting
ac-000106 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000106 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000106 · Reading
ReadingAcademicCollecting
ac-000106 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000106 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000107 · Listening
ListeningAcademicCollecting
ac-000107 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000107 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000107 · Reading
ReadingAcademicCollecting
ac-000107 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000107 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000108 · Listening
ListeningAcademicCollecting
ac-000108 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000108 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000108 · Reading
ReadingAcademicCollecting
ac-000108 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000108 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000109 · Listening
ListeningAcademicCollecting
ac-000109 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000109 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000109 · Reading
ReadingAcademicCollecting
ac-000109 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000109 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000110 · Listening
ListeningAcademicCollecting
ac-000110 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000110 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000110 · Reading
ReadingAcademicCollecting
ac-000110 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000110 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000111 · Listening
ListeningAcademicCollecting
ac-000111 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000111 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 000111 · Reading
ReadingAcademicCollecting
ac-000111 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
ac-000111 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 1 · Listening
ListeningAcademicCollecting
academic-01 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
academic-01 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 1 · Reading
ReadingAcademicCollecting
academic-01 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
academic-01 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 2 · Listening
ListeningAcademicCollecting
academic-02 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
academic-02 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
Academic Practice Test 2 · Reading
ReadingAcademicCollecting
academic-02 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
academic-02 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 1 · Listening
ListeningGeneral TrainingCollecting
general-01 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
general-01 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 1 · Reading
ReadingGeneral TrainingCollecting
general-01 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
general-01 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000201 · Listening
ListeningGeneral TrainingCollecting
gt-000201 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000201 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000201 · Reading
ReadingGeneral TrainingCollecting
gt-000201 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000201 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000202 · Listening
ListeningGeneral TrainingCollecting
gt-000202 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000202 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000202 · Reading
ReadingGeneral TrainingCollecting
gt-000202 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000202 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000203 · Listening
ListeningGeneral TrainingCollecting
gt-000203 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000203 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000203 · Reading
ReadingGeneral TrainingCollecting
gt-000203 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000203 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000204 · Listening
ListeningGeneral TrainingCollecting
gt-000204 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000204 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000204 · Reading
ReadingGeneral TrainingCollecting
gt-000204 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000204 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000205 · Listening
ListeningGeneral TrainingCollecting
gt-000205 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000205 Listening: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.
General Training Practice Test 000205 · Reading
ReadingGeneral TrainingCollecting
gt-000205 · 40 items
Timed answers on the thinnest item
0 of 200
Persons with a timed sitting
0 of 250
gt-000205 Reading: 0 of 200 timed responses per item (min), 0 persons.
Official results paired with this form: 0 (bands 5 / 6 / 7 / 8: 0 / 0 / 0 / 0); 30 in every band are needed to fit its own table.

Phase 2: official results to anchor the scale

A module needs 200 (Academic) or 150 (General Training) learners whose official result pairs with a timed sitting, with at least 30 of them in each whole band from 5 to 8, before a form’s table can be corrected.

Academic
Recruiting
Learners with an official result paired to a timed sitting
0 of 200
0 official results entered without a timed sitting inside the 14-day window, so not usable yet.
Listening: paired learners per whole band (30 needed in each)
- Band 5
  0 of 30
  Not yet
- Band 6
  0 of 30
  Not yet
- Band 7
  0 of 30
  Not yet
- Band 8
  0 of 30
  Not yet
Reading: paired learners per whole band (30 needed in each)
- Band 5
  0 of 30
  Not yet
- Band 6
  0 of 30
  Not yet
- Band 7
  0 of 30
  Not yet
- Band 8
  0 of 30
  Not yet
General Training
Recruiting
Learners with an official result paired to a timed sitting
0 of 150
0 official results entered without a timed sitting inside the 14-day window, so not usable yet.
Listening: paired learners per whole band (30 needed in each)
- Band 5
  0 of 30
  Not yet
- Band 6
  0 of 30
  Not yet
- Band 7
  0 of 30
  Not yet
- Band 8
  0 of 30
  Not yet
Reading: paired learners per whole band (30 needed in each)
- Band 5
  0 of 30
  Not yet
- Band 6
  0 of 30
  Not yet
- Band 7
  0 of 30
  Not yet
- Band 8
  0 of 30
  Not yet

A learner counts once per module, the latest result standing. Listening is the same paper in both modules, so a Listening form pairs with learners from either; a Reading form pairs only with learners of its own module.

Item statistics

What the answers say about each question

Published per form as soon as the Phase 1 sample is reached: how easy each item is, how well it separates stronger from weaker candidates, and which items are retired or under review.

No form has enough data yet

Item statistics are published for a form once every one of its questions has at least 200 timed answers, so that each facility figure is known to within about ±0.035 and a question can be retired on evidence rather than on a hunch. No timed sitting has been shared yet.

Only real data appears on this page

Every figure here is read from the live study store, which holds nothing but answers shared under research consent. Rehearsal data generated by scripts/simulate-responses.ts is written to a separate directory and is never shown here, however complete it looks.

Take part

How to help, and how to leave

Taking part is optional, costs nothing beyond the practice you were doing anyway, and can be undone completely at any time.

Three steps, all optional

Switch sharing on
Open the Research section of your progress page and turn on sharing. Your browser creates a random participant code; nothing about you is attached to it, and nothing is collected until you have switched it on.
Sit Listening and Reading sections timed
Only whole, timed, exam-mode sittings count. Practice-mode and untimed sittings are stored but excluded from every statistic, and a second sitting of the same form is treated as practice on seen questions.
Add an official result if you have one
If you took IELTS in the last 60 days, enter the bands and test date from your Test Report Form. A timed sitting within 14 days either side of that date anchors the form to the real scale. Never upload the form itself, your candidate number or your name.

You must be 16 or older to take part. Saying no changes nothing about how Bandwise works for you.

Open the research section Browse the test library

Withdrawing is as easy as joining

Switching sharing off stops collection from that moment. Pressing “Delete everything I shared” on the same page erases every answer and any official result held under your code from the study store at once, and the code is discarded; a later opt-in creates a new one.

There is no backup index or derived table to purge, so deletion is complete. Totals already published cannot be recalculated, but they contain nothing about you.

Your participant code is shown on the research section. It is the only key to your records: keep it if you ever want to ask for deletion from another device.

Methods

How the numbers are computed

The short version. Every threshold on this page is read from one place in the code, and the full protocol is a document in the repository, not a promise on a page.

Which answers count

Timed, whole-section, exam-mode sittings only; a participant’s first sitting of each form. Untimed rows are stored and counted as excluded, never analysed.

Phase 1: does each question work?

For every item, the facility (share correct, with its standard error), the corrected point-biserial against the rest-of-section score and, for choice items, how each option was chosen by high and low scorers. Items outside the facility range or with no discrimination are retired; weak ones are reviewed. Per form, KR-20 and the standard error of measurement in marks. With enough persons, a one-parameter Rasch model gives each item a difficulty with fit statistics, and learners who sat more than one form link the forms on a common scale.

Phase 2: is our band the official band?

Learners who enter a recent official result and sat a form timed inside the window give (raw score, official band) pairs. Once every whole band 5–8 has enough pairs, the mark-to-band table that best reproduces the official bands is fitted, with bootstrap intervals. The form passes only if bias is within 0.25 band at every band with a 95% interval excluding half a band, at least 70% of pairs are within half a band, and the correlation is at least .80.

Phase 3: how good is the AI examiner?

Writing scripts and Speaking recordings, collected under a separate consent, marked blind by two former examiners; agreement measured on a held-out third that was never used to tune the scorer. Not covered by this page yet.

What is never collected

No name, email address, IP address, account or device identifier; no essays, recordings or free text beyond an objective answer; no image of a Test Report Form. The participant code is random and stored only in your browser. This page shows counts and estimates only, and published reports suppress any cell under ten learners.

The full protocol, consent text and monthly runbook are in docs/design/calibration-study.md in the Bandwise repository; the study design and the derivation of every sample size are in docs/research/03-fidelity-audit.md §4; the data contract is src/lib/calibration/types.ts. Monthly reports, when there are any, go to docs/reports/.

Thresholds the page reads

One source in the code; nothing on this page hard-codes a number.

Timed answers per item before item statistics are reported: 200
Persons per form before Rasch difficulties are trusted: 250
Anchored learners per module: 200 Academic, 150 General Training
Anchored learners in each whole band 5–8: 30
Official result to timed sitting, either direction: 14 days
Facility range outside which an item is retired: 0.15–0.95
Point-biserial below which an item is reviewed: 0.15

Take part How the tests are made How scoring works

IELTS measures every version of its test. Nobody has measured ours yet.

The claim ladder

How far each form has got

Phase 1: timed answers per form

Academic Practice Test 000100 · Listening

Academic Practice Test 000100 · Reading

Academic Practice Test 000101 · Listening

Academic Practice Test 000101 · Reading

Academic Practice Test 000102 · Listening

Academic Practice Test 000102 · Reading

Academic Practice Test 000103 · Listening

Academic Practice Test 000103 · Reading

Academic Practice Test 000104 · Listening

Academic Practice Test 000104 · Reading

Academic Practice Test 000105 · Listening

Academic Practice Test 000105 · Reading

Academic Practice Test 000106 · Listening

Academic Practice Test 000106 · Reading

Academic Practice Test 000107 · Listening

Academic Practice Test 000107 · Reading

Academic Practice Test 000108 · Listening

Academic Practice Test 000108 · Reading

Academic Practice Test 000109 · Listening

Academic Practice Test 000109 · Reading

Academic Practice Test 000110 · Listening

Academic Practice Test 000110 · Reading

Academic Practice Test 000111 · Listening

Academic Practice Test 000111 · Reading

Academic Practice Test 1 · Listening

Academic Practice Test 1 · Reading

Academic Practice Test 2 · Listening

Academic Practice Test 2 · Reading

General Training Practice Test 1 · Listening

General Training Practice Test 1 · Reading

General Training Practice Test 000201 · Listening

General Training Practice Test 000201 · Reading

General Training Practice Test 000202 · Listening

General Training Practice Test 000202 · Reading

General Training Practice Test 000203 · Listening

General Training Practice Test 000203 · Reading

General Training Practice Test 000204 · Listening

General Training Practice Test 000204 · Reading

General Training Practice Test 000205 · Listening

General Training Practice Test 000205 · Reading