Thinking global, living local

Do CBSE Past Papers Actually Help? Depends Which of the 11 Subjects You Ask.

September 17, 2026

Parsing 30,734 questions from five years of CBSE Class-XII papers shows that "practise past papers" is good advice for Accountancy and Physics and close to useless for English Core and Biology — the coaching-industry folklore treats one subject's pattern as if it applied to all eleven.

Education Data · CBSE Board Exams · Exam Preparation

Do CBSE Past Papers Actually Help? Depends Which of the 11 Subjects You Ask.

Skip to article content
Text Size

Revised · v1.1.0 · what changed

1,955CBSE Class-XII papers parsed, 2022–2026
49%of Accountancy's high-mark marks sit on a recurring topic
0%of English Core's high-mark marks do
10/10sample-paper sessions, 2015-16–2024-25, testing Physics's “electric field”

"Solve the last five years' papers" is close to universal advice from CBSE coaching material. It is also, on the evidence, true for some subjects and nearly meaningless for others. A parse of 1,955 Class-XII question papers from cbse.gov.in (2022–2026, main and compartment sittings, 30,734 questions extracted) can measure this directly: for each subject, what share of the high-mark (≥4-mark) marks sit on a topic — a 2–4-word phrase — that recurs across at least three of the five exam years. Call it the repetition index.

The repetition index, by subject

The split does not track how big a subject's syllabus is. It tracks how formulaic its questions are — and that is a very different sort by subject than most study advice assumes:

SubjectPapersHigh-mark QsRepetition index, %
Accountancy5010949
Physics6412044
Economics6117338
Mathematics629022
Business Studies5411916
History6716716
Geography6022914
Political Science602007
Chemistry721335
Biology621695
English Core722590
Repetition index by subject (2022–2026 corpus) Accountancy49% Physics44% Economics38% Mathematics22% Business Studies16% History16% Geography14% Political Science7% Chemistry5% Biology5% English Core0% Share of ≥4-mark marks sitting on a phrase recurring in ≥3 of 5 exam years (2022–2026).
Repetition is a subject property, not a syllabus-size property — Physics and English Core have similar syllabus scope but sit at opposite ends of the index.
Read this as lexical repetition, not conceptual repetition. The index measures repeated wording, mined as 2–4-word phrases, not repeated ideas. Chemistry and Biology score 5% not because the board tests wildly different concepts each year, but because their questions describe the same chemistry in differently-worded ways (“calculate emf of the following cell…” vs “consider the following standard electrode potential values…” is one topic, two phrasings). English Core scoring 0% is correct by design — its long-answer questions quote unseen passages, which never repeat.

What actually recurs, and where

Accountancy is the most repetitive paper in the corpus. Two topics alone — admission/retirement of a partner (14 papers across 2023/25/26, 139 marks) and the balance sheet incl. Common Size Balance Sheet (20 papers across 2024/25/26, 139 marks) — plus profit-sharing-ratio changes, cash flow from investing activities, and dissolution of the firm make up most of the high-mark paper, year after year.

Physics has exactly two topics that recur at high marks in 4 of the 5 years, across the entire 11-subject corpus: electric field (14 papers, 84 marks) and magnetic field (11 papers, 73 marks). Parallel-plate capacitors and refractive index follow at 3-of-5 years. No other subject has a single content topic that reliable.

Economics leans on a recurring numerical format — the “for a hypothetical economy, it is known that…” problem appears in 14 papers worth 133 marks — plus aggregate demand, commercial banks/RBI, and gross value added. History circles a small set of sourced narratives (Vijayanagara Empire, Quit India Movement, Salt Satyagraha, Bombay Deccan riots) even though its index reads a modest 16%, because those set-pieces sit specifically in the 8-mark long-answer slot. Business Studies reuses “functions of management” (planning, directing) almost verbatim across three straight years.

A separate finding: the same skill-subject block, six times over

Away from topic recurrence, the corpus's structural analysis turns up a more actionable pattern for students taking any of CBSE's Skill Education electives (Artificial Intelligence, Information Technology, Web Applications, Taxation, Financial Markets Management, Typography & Computer Applications). Every one of these papers devotes roughly 10 of its 50–60 marks — 17–20% — to an identical five-unit Employability Skills block: communication, self-management, ICT basics, entrepreneurship, and green skills. The wording repeats near-verbatim across subjects — the same active-listening question appears in both the Taxation and Web Applications papers; the same “Ctrl+M adds a slide” item appears in both AI and Typography.

The single highest-leverage move for a Skill Education student is this block, not the subject content. It is graded, worth a fixed ~10 marks, and identical across whichever of the six papers a student sits — unlike any subject-specific topic, which recurs at best 3 years out of 5.

Does this hold over more years? Ten years of official sample papers say: mostly, yes

The repetition index above is built on five years of administered exam papers (2022–2026) — cbse.gov.in itself does not publish real Class-XII papers from earlier sessions in a form this analysis could fetch. But CBSE separately publishes an official Sample Question Paper (SQP) before each session, illustrating that year's format and a slice of its content, going back to 2015-16. Searching that decade-long SQP archive for the specific topics already found recurring in the real-paper analysis gives an independent, longer-run check on the same claim:

Subject · topicSQP sessions with the subjectSessions where the topic appears
Physics · “electric field”10 (2015-16 to 2024-25)10 of 10
Physics · “magnetic field”10 (2015-16 to 2024-25)10 of 10
Economics · “aggregate demand”10 (2015-16 to 2024-25)10 of 10
Accountancy · “balance sheet”7 (2018-19 to 2024-25)7 of 7
Accountancy · partner admission/retirement7 (2018-19 to 2024-25)2 of 7

Physics's two field topics and Economics's “aggregate demand” check out in every single sample paper CBSE has released in the last decade — a far stronger confirmation than the original 4-of-5-years finding, and evidence that this isn't a quirk of the 2022–2026 window. Accountancy's broad “balance sheet” topic is equally universal. But the specific partnership-admission/retirement cycle that dominates Accountancy's high-mark real-exam questions (14 of 20 papers across three exam years) shows up in only 2 of 7 sample papers.

Why the Accountancy result diverges, and why that's not a contradiction. A Sample Question Paper is one illustrative paper per session; the real-exam repetition index is built from an entire exam year's worth of papers — multiple sets and sittings, including compartment exams. A topic can recur reliably across a whole year's real papers without appearing in the single SQP that year, simply because the SQP is a smaller sample of the same syllabus, not a shrunk copy of the full exam. This divergence is a fact about SQPs' sampling method, not a reason to doubt the original real-paper finding — and it is exactly the kind of check that could have gone the other way; it didn't for Physics or Economics, which is why those two results are reported as a genuine strengthening. CBSE did not publish an Accountancy SQP for 2015-16 through 2017-18 in this corpus, so that subject's window is 7 years, not 10.

What would break this analysis

Method caveats, stated plainly. (1) Topics were mined automatically as recurring n-grams, not hand-picked, so the conclusion isn't smuggled in by a curated list — but recurrence is counted in exam years, not papers, because papers within one sitting overlap roughly 75% and would otherwise manufacture false recurrence. (2) Hindi-medium text extracts as mojibake in this corpus (embedded-font code-point mapping), so only the English-medium ~54% of parsed questions could be analysed; the index does not cover Hindi-medium-specific patterns. (3) Marks for most of the 2022–2026 archive were parsed heuristically from paper text rather than reconciled against each paper's own instructions (only a 32-paper 2024-25 subset was exactly reconciled) — treat every index value as accurate to a few percentage points, not exact.

The takeaway

"Practise past papers" is not one piece of advice; it's eleven, and they don't agree. For Accountancy and Physics students, five years of past papers genuinely rehearse a large share of what will be asked again — and for Physics and Economics specifically, a decade of official sample papers backs that up further still. For English Core, Biology, and Chemistry students, the same advice mostly teaches format and timing, not content, because those subjects are built not to repeat. And for anyone taking a Skill Education elective, the highest-certainty ten marks in the entire paper are a shared syllabus block that has nothing to do with the subject's name.

Sources: 1,955 CBSE Class-XII question papers, main and compartment sittings, downloaded directly from cbse.gov.in/cbsenew/question-paper.html, academic years 2022–2026; 30,734 questions parsed (16,633 usable English-medium text), extracted with pdftotext -layout and mined for 2–4-word recurring phrases (n-grams appearing in ≥3 of 5 exam years, boilerplate stripped). Paper-architecture detail (Employability Skills block, section structure) from a separately reconciled 32-paper 2024-25 subset whose marks were checked against each paper's own General Instructions. The ten-year extension uses CBSE's own Sample Question Paper archive (cbseacademic.nic.in, sessions 2015-16 to 2024-25, sourced from this account's corpus), searched directly for the phrases already found recurring in the real-paper analysis — a targeted presence check, not an independent re-mining of the SQP text. cbse.gov.in and cbseacademic.nic.in could not be fetched directly from this piece's research environment for anything beyond what the underlying repository already held; no post-2024-25 SQP or pre-2022 real exam paper is claimed here that was not already in that corpus. Full analysis, methodology and the underlying DuckDB tables are documented in this account's own CBSE_paper_patterns research repository.

Revision history.
  • v1.1.0 — 17 September 2026 — added a ten-year (2015-16–2024-25) extension using CBSE's official Sample Question Paper archive, confirming the electric-field/magnetic-field and aggregate-demand findings hold in every sample paper released in the last decade, and finding a genuine divergence on Accountancy's partnership-cycle topic.
  • v1.0.0 — 17 September 2026 — first published.
Related on this blog: Thirty-One Years of Asking Parliament the Same Question About Fertiliser — another piece built the same way — read years of one specific document type end to end to find what's actually constant versus what changed.

About this article: Researched, written and edited by Umashankar Triplicane Dwarakanathan, with AI research assistance; every figure is meant to trace to the primary source cited. See the Editorial Policy for how sourcing, AI use and corrections work.

Topics
Umashankar Triplicane Dwarakanathan
Contact Us
Umashankar Triplicane Dwarakanathan
Investment Promotion & Energy-Sector Leader · Chennai, Tamil Nadu, India
LinkedIn → GitHub → Email +91 78273 81696
How this site works

Data-led analysis of India's trade, currency and industrial policy. Every article is built from primary official sources, and every figure links back to the release, table or filing it came from.

Sources. DGCI&S TradeStat (imports/exports, HSN-wise) · PIB (government press releases, January 2017 to today, refreshed daily) · RBI (circulars, balance of payments) · MoSPI (CPI/WPI, IIP) · PARIVESH (environmental clearances) · CCIL (bond yields) · BIS (policy rates) · SEBI, NSE/BSE and SEC filings for company data.

Interpretation. Figures carry their vintage and retrieval date; estimates and press-reported numbers are labelled as such; where sources disagree, both are shown. Corrections are made visibly, never silently. Articles are written with AI assistance from the cited sources — AI-generated text can misstate figures even when working from real material, so verify any number that matters to a decision against the linked primary source.

footer

Browse all articles by topic

Every piece on this blog, grouped. Or read the full index.

Agriculture & FertilisersAI ToolsChemicalsClimate & CarbonEnergy & FuelsGas & LNGImport SubstitutionIndustrial PolicyMarkets & FinanceMobility & EVPrices & InflationTextilesTrade & Tariffs

Each topic is a live archive page that updates itself as pieces are labelled. It replaces a hand-kept list that had fallen 18 articles behind.