← Acrolly Research

How Acrolly Analyses JEE PYQs

Source filters, difficulty buckets, question-type labels, and limitations behind Acrolly Research reports.

Snapshot 2026-08-19

Source corpus

Reports aggregate rows from acrolly.question_bank. At snapshot time:

  • JEE Advanced: 1086 questions, 22 papers, 2016–2026
  • JEE Main: 6801 questions, 82 sessions, 2023–2026

Advanced filter: reference ~ '^JEE_ADVANCED_[0-9]{4}_PAPER' AND reference NOT ILIKE '%MOCK%' AND reference NOT ILIKE '%CHAPTER%'

Main filter: reference ~ '^JEE_MAINS_[0-9]{4}_' AND reference NOT ILIKE '%MOCK%'

Grouping rules

(subject, chapter) — Thermodynamics exists in both Physics and Chemistry

Unit filters use ai_stream (e.g. Organic Chemistry, Mechanics, Calculus).

Difficulty scale

0–100 integer difficulty_level

  • Main: Easy 0–49, Medium 50–64, Hard 65–79, Very hard 80–100
  • Advanced: Medium 0–64, Hard 65–79, Very hard 80–100 (scores cluster high; almost no Easy band)

Question types

  • mcq_single_correct → Single correct
  • mcq_multiple_correct → Multiple correct
  • integer → Integer / numerical
  • matching → Matching
  • paragraph_mcq → Paragraph MCQ
  • paragraph_integer → Paragraph integer
  • reasoning → Assertion–reason

Integer / numerical is a response format, not a claim that the item is purely computational.

What we do not claim

  • Official NTA weightage or future paper predictions
  • Calculation-complexity or conceptual-vs-computational scores (not stored)
  • That every official Main shift is present in the bank

Figures reflect historical question distribution in Acrolly's tagged PYQ bank. They are not a prediction of future papers.

FAQ

Where do the numbers come from?
From Acrolly’s question_bank table, filtered to tagged PYQ references only. Advanced uses JEE_ADVANCED_<year>_PAPER-*; Main uses JEE_MAINS_<year>_<MONTH>_<day>_MORNING|EVENING. Mocks are excluded.
Why isn’t calculation complexity in these reports?
That label is not stored as a first-class column on PYQ rows today. We do not invent proxies that would look precise but be wrong.
Can I download the data?
Yes. Each report links to a JSON file under /research/data/ for LLM and developer use.

Browse all reports · Free PYQs