Methodology
Every exam on IvyTest is described in one file, built from the university's own documents. The questions, the mock exams and MedChance are made from that file. Facts checked on official pages are logged with their link and date.
One free guide per university, with its sources and when they were checked.
How it fits together
Official documents go in at the top. Three kinds of file are built from them, and everything you use comes out of those files. Each step below links to how it works.
Read on the university's or the ministry's own site.
taxonomies/<university>.jsonSubjects and subtopics, questions per subject, time, answer options and marking.
exams/<exam>.tsDates, fees, places and admission rules, with the official sources behind them and when they were checked.
Builds
videocourse/<university>.jsonEach exam's syllabus, mapped onto video lessons that several universities share.
Builds
Kept current by
The exam file
For every university there is one file, built from its official documents: the subjects and subtopics, the questions per subject, the time, the answer options and the marking. Where a paper has separately timed sections with a pass mark in each, as at Humanitas, the file says so.
Subject and subtopic names are copied character for character from the official document, never translated, reworded or recapitalised. Every question and every statistic is filed under them.
A figure the university has not published is left empty or marked as estimated. No questions are written for an option count nobody published.
When a university's admission criteria and its sample papers disagree, the criteria win, and the file notes what the samples show.
Automated tests fail when a new version of a file breaks one of these rules.
31 of the 52 mock exams include at least one estimated figure. Each one says which, before you start.
From its exam file
HUMAT
On the paper
Tap a subject to see its subtopics, named as the file names them.
Timed sections
Scientific Thinking & Academic Literacy
20 questions, 45 min
Biology & Chemistry
20 questions, 35 min
Mathematics & Physics
20 questions, 40 min
Every figure on this card is published by the university.
Questions
Every question is written by IvyTest for one university's exam, from a brief made out of that exam's file.
A live question from Cattolica's bank, as the simulator shows it.
Five options where the paper has five, four where it has four. The subject and subtopic are copied from the file, and the import lists every name it receives, so a stray one shows.
Each question carries a worked explanation, why each wrong option is tempting and why it is wrong, a difficulty and a recommended time.
The import refuses a question with the wrong number of options for its exam, one for a format the university has not published, one with a missing figure, and wrong-answer notes on the correct option. Near-copies of questions already in the bank are reported.
The bank is checked for repeated options, duplicates and answers that are the longest option more often than chance. A question is flagged once candidates have marked it for review twice, or, after ten answers, when one wrong option is picked more than the right one or fewer get it right than guessing would. Broken questions are retired.
Sources
Exam dates, fees, places, cutoffs and formats change every year. Each one is read on the university's or the ministry's own page, and the site says when.
Every fact checked on an official page is logged with the exact figures, the link, the day it was read, whether the site uses it and what is still unclear. Research from elsewhere is checked against the official page first; what cannot be checked is recorded as open, not used.
What the page says
Humanitas publishes its full ranking on its ESSE3 admissions portal, linked from its admissions page. The last row marked Admitted or Registered is position 332; position 333 has the same score and is Not admitted.
Figures
1085 rows: 145 Registered, 5 Admitted, 270 Not admitted, 483 Rejected (under a section minimum), 182 Cancelled. Last admitted 463/960. Mean of the 150 admitted 612.8/960.
Sources
Still open: Five rows still read Admitted rather than Registered; the page shows no publication date.
Each exam guide lists its sources and when they were last checked. A script re-reads every cited page and compares it with the copy saved last time.
Unchanged
The page reads exactly as it did when a person last checked the guide against it, so its date moves forward.
Changed
The new text is saved and listed for a person. A suggested correction is applied only if the sentence behind it is on the page word for word, with its numbers, and nothing goes live until a person approves it.
Unreachable
Listed for a person. Nothing changes.
Dates are never moved forward in bulk to clear the list.
The official pages the guides cite are also read about once a day for the coming cycle's exam dates and application windows.
A page changes and a new date appears.
The date is reported with the sentence it comes from.
It is kept only if that sentence is on the page word for word, contains the date and falls in the coming cycle.
An admin approves it before it is recorded or sent to anyone who asked for an alert.
MedChance and MedMatch
MedChance estimates your chances at one university from a short diagnostic. MedMatch helps you choose where to apply. Both say what is known, and say what is not.
Subjects are weighted by their share of the exam's questions, read from the same exam file.
Each university's diagnostic questions come from its own question bank, in proportion to those weights. Everyone gets the same questions, in the same order.
A score is corrected for the exam's number of options and its penalty for wrong answers, then compared with that university's own cutoff.
Each cutoff is marked verified, from a ranking or decree the university published, or estimated. Where the written test does not decide admission, there is no percentage, and MedChance says what does decide it.
MedChance does not test language proficiency, by design. Where an exam tests it, the result says how much of the exam it could not test.
Every prediction is saved when it is given, so it can be compared with the outcomes candidates report. No accuracy figure is published yet.
Fit
How many of your preferences a school meets, from recorded facts: country, teaching language, fees, city size and the subjects its exam tests.
Odds
Only from your own MedChance result for that school. Never another school's result, and never its reputation.
With nothing to rank on, MedMatch ranks nothing and lists the schools by region.
Video course
Each university's syllabus is mapped onto a shared set of lessons. A lesson's scope is set by everything those syllabuses ask of it, not by its title.
It is made once. Every university whose exam tests that material gets the same lesson, under the headings its own exam uses, word for word.
What we do not claim
IvyTest's free practice papers are written by IvyTest in the exam's current format. Old official papers are described in the guides only after someone has checked them, and never reproduced.
A date, pass mark or fee the university has not announced is shown as not announced. A mock built on an estimate says which figure is estimated.
Predictions are saved so they can be compared with real outcomes. No accuracy figure is published yet.
Each university is sold as its own Complete Admissions System, which covers that university only.
See it for yourself
Each exam guide shows the format, the dates and the sources, and when they were checked. The demo runs the real simulator.
