CSAT · UPSC Paper II

The whole of CSAT, taken apart — and made calm.

A math-and-statistics house took the qualifying paper apart — both halves. You practise the real exam: timed quant with a specialist’s read on every trap, and comprehension that’s reasoning from a passage, not a vocabulary lottery. One pass, both halves, no fear.

Both halves, live. Comprehension on CSAT is a reasoning test — in ten years, not one question tested vocabulary or grammar. It’s trainable. We built the practice for it.

Sign in & take your free attempt

Free to start, no card. Your first graded set — both halves, with its diagnostics — free, so you feel the engine work once before you decide a thing.

Free · no sign-in

Read the worked solutions first — free.

Every UPSC CSAT question, solved in full: the reasoning, the exact answer, and the trap in each wrong option named. Free to read, forever, with or without a pass — see the work before you decide a thing.

Read the free CSAT solutions →

Try one — free, no sign-in

Quant · try it free, no sign-in

UPSC CSAT 2024, Paper II, Q18 — a real past-year question.

421 and 427, when divided by the same number, leave the same remainder 1. How many numbers can be used as the divisor in order to get the same remainder 1?

Choose an answer

Pick an option above, then tap to see the read.

If a divisor dd leaves remainder 11 on both numbers, then dd divides both 421−1=420421-1=420 and 427−1=426427-1=426 exactly. So dd must be a common divisor of 420420 and 426426 — i.e. d∣gcd⁡(420,426)d \mid \gcd(420, 426).

Find the gcd the fast way, by subtraction, not by factorising two 3-digit numbers under a clock: 426−420=6,gcd⁡(420,6)=6  (since 420=70×6 exactly).426 - 420 = 6, \qquad \gcd(420, 6) = 6 \ \ (\text{since } 420 = 70 \times 6 \text{ exactly}).

So dd must be a divisor of 66: the candidates are {1,2,3,6}\{1, 2, 3, 6\} — four numbers.

The one guard that decides the whole question: a divisor that leaves a remainder of 1 must itself be greater than 1 — you cannot leave “remainder 1” when dividing by 1 (dividing by 1 always leaves remainder 0). Drop d=1d=1.

Valid divisors: {2,3,6}\{2, 3, 6\} — three. Quick check: 421=6×70+1421 = 6\times70+1; 421=3×140+1421=3\times140+1; 421=2×210+1421=2\times210+1 — all leave remainder 1, confirmed. Answer: (c) 3.

The skill this rewards: “same remainder on two numbers” is a gcd-of-the-differences problem, every time, with no exceptions — and the boundary guard (divisor must exceed the remainder) is the one line a volume-drilled student skips because it never bites on “nicer” remainders. This is not a hard calculation; it is a discipline question.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(a) 1 — reveals “solved a different, easier question.” A solver who reaches gcd⁡(420,426)=6\gcd(420,426)=6 and stops, answering “the divisor is 6, so there is 1 valid divisor,” has correctly done the hard part (the gcd) and then quietly substituted “which number is the answer” for “how many numbers work.” This is not a calculation error — the arithmetic is perfect — it is stopping one conceptual step early. The tell: if your answer to a “how many” question is the same number as an intermediate quantity you computed along the way, ask whether you actually answered “how many” or just reported that quantity.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(b) 2 — reveals incomplete enumeration under time pressure. The solver correctly reaches the valid set {2,3,6}\{2, 3, 6\} and then drops one — commonly 66 itself, on the false instinct that “the divisor can’t equal the gcd” (it can; 66 divides both 420420 and 426426 exactly, leaving remainder 11 on the originals, same as 22 and 33 do). The tell: when a problem says “how many,” count your final set out loud, item by item, before you write the digit down — don’t infer the count from memory of the derivation.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(c) 3 — correct, and here is what “correct” required you to actually do. Not just find a gcd (a mechanical step most AR/NT questions share) — apply the boundary guard as part of the same breath as finding the divisor set, not as a separate “oh wait” check afterward. That is the skill CSAT is actually testing in this family: discipline at the boundary, not raw number-theory knowledge.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(d) 4 — the deliberately engineered near-miss, and the one to actually study. This is the entire set of divisors of 6 — {1,2,3,6}\{1, 2, 3, 6\} — counted without applying the remainder guard at all. Every step of the real method up to the gcd is done correctly; the error is a single missing sentence: “but the divisor must be greater than the remainder it leaves.” This is the signature trap of the Number Properties family on CSAT (the classification taxonomy names it divisor-vs-remainder-slip) — it is precisely the kind of near-miss where a student who did 95% of the correct work still gets the mark taken away, because the exam is testing whether you know the boundary condition exists, not whether you can find a gcd. The tell, generalized: whenever a question fixes a remainder rr and asks for valid divisors, the answer set is “divisors of the gcd that are >r> r” — never “divisors of the gcd,” full stop. Build that filter into the method itself, not as an afterthought check.

See the read for every option

(a) 1

(a) 1 — reveals “solved a different, easier question.” A solver who reaches gcd⁡(420,426)=6\gcd(420,426)=6 and stops, answering “the divisor is 6, so there is 1 valid divisor,” has correctly done the hard part (the gcd) and then quietly substituted “which number is the answer” for “how many numbers work.” This is not a calculation error — the arithmetic is perfect — it is stopping one conceptual step early. The tell: if your answer to a “how many” question is the same number as an intermediate quantity you computed along the way, ask whether you actually answered “how many” or just reported that quantity.

(b) 2

(b) 2 — reveals incomplete enumeration under time pressure. The solver correctly reaches the valid set {2,3,6}\{2, 3, 6\} and then drops one — commonly 66 itself, on the false instinct that “the divisor can’t equal the gcd” (it can; 66 divides both 420420 and 426426 exactly, leaving remainder 11 on the originals, same as 22 and 33 do). The tell: when a problem says “how many,” count your final set out loud, item by item, before you write the digit down — don’t infer the count from memory of the derivation.

(c) 3 — the key

(c) 3 — correct, and here is what “correct” required you to actually do. Not just find a gcd (a mechanical step most AR/NT questions share) — apply the boundary guard as part of the same breath as finding the divisor set, not as a separate “oh wait” check afterward. That is the skill CSAT is actually testing in this family: discipline at the boundary, not raw number-theory knowledge.

(d) 4

(d) 4 — the deliberately engineered near-miss, and the one to actually study. This is the entire set of divisors of 6 — {1,2,3,6}\{1, 2, 3, 6\} — counted without applying the remainder guard at all. Every step of the real method up to the gcd is done correctly; the error is a single missing sentence: “but the divisor must be greater than the remainder it leaves.” This is the signature trap of the Number Properties family on CSAT (the classification taxonomy names it divisor-vs-remainder-slip) — it is precisely the kind of near-miss where a student who did 95% of the correct work still gets the mark taken away, because the exam is testing whether you know the boundary condition exists, not whether you can find a gcd. The tell, generalized: whenever a question fixes a remainder rr and asks for valid divisors, the answer set is “divisors of the gcd that are >r> r” — never “divisors of the gcd,” full stop. Build that filter into the method itself, not as an afterthought check.

And the reading half, to the same standard —

Verbal · try it free, no sign-in

UPSC CSAT 2023, Paper II, Q21 — a real past-year question.

UPSC CSAT (Paper II), 2023 — Q21

Passage

To tackle the problem of pollution in cities, policy makers think that drastic actions like temporary use of odd-even number scheme for vehicles, closing schools, factories, construction activities, and banning the use of certain type of vehicles are a way forward. Even then the air is not clean. Vehicles more than 15 years old comprise one percent of the total; and taking them off the road will not make any difference. Banning certain fuels and car types arbitrarily is not proper. Diesel engines produce more PM2.5 and less CO₂ than petrol or CNG engines. On the other hand, both diesel and CNG engines produce more NOₓ than petrol engines. No one has measured the amount of NOₓ that CNG engines are emitting. Arbitrary bans on vehicles that have passed mandated fitness tests and periodic pollution tests are unfair. What is needed is the scientific and reliable information about the source of pollutants on a continuing basis and the technologies that will work to reduce pollution from them.

Question: Which one of the following statements best reflects the most logical and rational implication conveyed by the passage?

Choose an answer

Pick an option above, then tap to see the read.

The one-line takeaway (this is the whole pitch, proved, not asserted): every wrong answer here was beaten by rereading the passage’s own sentences, never by knowing a word the passage didn’t already explain. That’s what “the answer is in the text, not your word-stock” means in practice.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(a) — borrows the mood, not the reason. The passage is skeptical of arbitrary curbs, and so is (a) — but read closely, the passage never says curbs are hard to implement. It says the opposite: they were implemented (odd-even ran, factories and schools closed, vehicles were banned) and still failed — “even then the air is not clean” — because they were ineffective and unfair, not because they were hard to carry out. (a) borrows the passage’s real skepticism about blunt bans but relabels the reason as an implementation problem the passage never raises — and even read as charitably as possible, it still only keeps the failure half and drops the passage’s actual point, the constructive half (“what is needed is…”).

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

Why (b) — proof from the text, not opinion. Read the passage’s own arc: it lists blunt actions (odd-even, closures, arbitrary bans) and then says “even then the air is not clean” — they don’t work. It shows exactly why: 15-year-old vehicles are “one percent of the total” (banning them can’t matter), and diesel-vs-CNG trades one pollutant for another (nobody’s even measured the NOₓ). Then it names what would work: “scientific and reliable information about the source of pollutants… and the technologies that will work.” That’s a two-part argument — blunt measures fail, and evidence-based measures are what’s needed — and (b) is the only option that states both halves. Nothing here required knowing a hard word. It required tracking the passage’s own logic from its first sentence to its last.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(c) — the passage says the opposite. (c) calls for penalising drivers “without periodic pollution tests” — but the passage explicitly defends vehicles that have passed those tests, calling bans on them “unfair.” (c) doesn’t misread a detail; it inverts the passage’s own position. This is the trap built for a reader who remembers the topic (penalties, pollution tests) but not the direction the passage actually argued.

You dodged it — here's why it works anyway, since the exam varies it.

The trap caught you — and it's built to catch almost everyone the first time. Here's how it works.

(d) — sounds reasonable, isn’t there. “Absence of laws → arbitrary decisions” is a plausible real-world claim about how governments behave — but the passage never says this. It critiques arbitrariness itself, never once blaming a legal vacuum for it. This is the purest form of the trap: an inference that sounds smart and even sounds like something a policy expert might say, built from outside knowledge the passage supplies no evidence for.

See the read for every option

(a) Arbitrary curbs on vehicles to reduce pollution are difficult to implement.

(a) — borrows the mood, not the reason. The passage is skeptical of arbitrary curbs, and so is (a) — but read closely, the passage never says curbs are hard to implement. It says the opposite: they were implemented (odd-even ran, factories and schools closed, vehicles were banned) and still failed — “even then the air is not clean” — because they were ineffective and unfair, not because they were hard to carry out. (a) borrows the passage’s real skepticism about blunt bans but relabels the reason as an implementation problem the passage never raises — and even read as charitably as possible, it still only keeps the failure half and drops the passage’s actual point, the constructive half (“what is needed is…”).

(b) Knee-jerk reactions cannot solve the problem of pollution but an evidence-based approach will be more effective. — the key

Why (b) — proof from the text, not opinion. Read the passage’s own arc: it lists blunt actions (odd-even, closures, arbitrary bans) and then says “even then the air is not clean” — they don’t work. It shows exactly why: 15-year-old vehicles are “one percent of the total” (banning them can’t matter), and diesel-vs-CNG trades one pollutant for another (nobody’s even measured the NOₓ). Then it names what would work: “scientific and reliable information about the source of pollutants… and the technologies that will work.” That’s a two-part argument — blunt measures fail, and evidence-based measures are what’s needed — and (b) is the only option that states both halves. Nothing here required knowing a hard word. It required tracking the passage’s own logic from its first sentence to its last.

(c) A heavy penalty should be enforced on those driving without periodic pollution tests.

(c) — the passage says the opposite. (c) calls for penalising drivers “without periodic pollution tests” — but the passage explicitly defends vehicles that have passed those tests, calling bans on them “unfair.” (c) doesn’t misread a detail; it inverts the passage’s own position. This is the trap built for a reader who remembers the topic (penalties, pollution tests) but not the direction the passage actually argued.

(d) In the absence of laws to deal with the problems of pollution, the administration tends to make arbitrary decisions.

(d) — sounds reasonable, isn’t there. “Absence of laws → arbitrary decisions” is a plausible real-world claim about how governments behave — but the passage never says this. It critiques arbitrariness itself, never once blaming a legal vacuum for it. This is the purest form of the trap: an inference that sounds smart and even sounds like something a policy expert might say, built from outside knowledge the passage supplies no evidence for.

This is the read on your every miss, both halves — across a real ~20–30 question set. Sign in for your first graded set, free →
Every miss like this gets the same read, a few times a week, on our Telegram channel. Free. Join

CSAT is qualifying — clear it, and your GS score counts.

CSAT is a qualifying paper. Clear it, and your General Studies score is what decides your rank. Miss the qualifying mark, and that year's GS score doesn't count — however strong it is, it's the same attempt, gone. Because CSAT doesn't carry marks toward the final rank the way the optional does, many capable, serious aspirants give it less prep time than it needs, and some lose an attempt to it as a result.

The gate is clearable, and it's more predictable than it looks. It rewards preparation more than talent. That's the whole reason we took it apart.

We didn't guess at this exam. We measured it.

Weightage

Three question-families carry 73% of every quant mark.

73%

Data interpretation: 5 of 444 — about 1%

Ten years of CSAT. Five DI questions. The pie-and-bar-chart drilling most aspirants over-invest in is the rarest thing on the paper.

Three question-families carry 73% of every CSAT quant mark.

Arithmetic, number properties, and logical-quant reasoning — 324 of 444 questions across ten years. And data interpretation, the charts everyone over-drills? Five questions in ten years.

10 years of CSAT (2016–2025), 444 quant questions. Uniform marking (2.5 marks each, no partial credit) means share of questions = share of marks, exactly.

Arithmetic, number properties, and logical-quant reasoning — 324 of 444 questions across ten years. Over half the paper is just arithmetic and number properties (228 of 444). And data interpretation — the pie and bar charts everyone over-drills — is five questions in ten years. About 1%. Most aspirants spend their prep in the wrong place. We'll show you where the marks actually are.

Traps

The exam is trap-first, not maths-first.

By our distractor analysis of ten years of papers, the wrong answers cluster into a closed set of trap families — and a deliberately "almost-right" answer, the kind engineered to catch the solver who's nearly there, shows up again and again. You don't lose marks because the maths is hard. You lose them because the wrong answer was made to look right. We teach you to see the mechanism.

Time

CSAT is won on the clock — about 90 seconds a question.

On our solve-time map, only about one in six quant questions is a genuine quick win. The "consider the following" and data-sufficiency items are the biggest time-sinks — high cost, few easy marks — and they're exactly where an unprepared aspirant burns the clock. Winning is triage, not finishing everything. We built the practice around that.

What the engine actually does — and what it isn't.

What it is

  • You sit the real exam, not a self-report. Exam-faithful timed papers — both halves — you select an option, with real CSAT mechanics and real −1/3 negative marking. Because the engine sees exactly what you picked, the diagnosis is precise, not a vibe.
  • A specialist reads every miss — on both halves. On quant: why the option you chose was built to trap you — the rate you inverted, the ratio you read as a fraction, the off-by-one — the cleaner method, the one skill it reveals. On comprehension: why the keyed answer is the one the passage actually supports, and why the option that merely sounded right doesn't survive the text. That's the read a volume-player structurally cannot write.
  • Your worksheet, built from your misses. After a paper, a guided sheet that targets your weak patterns — across both halves, winnable problems first, with the method shown — not a generic next test.
  • A schedule that owns your prep, not a streak that guilts it. It teaches you spaced repetition, sequences your next sittings, and pushes the next one to your calendar in a tap.
  • An honest readiness read. A calm, conservative signal of where you stand — necessary, not sufficient — never a pass/fail verdict you'd bet a year on.
  • Built on free knowledge, not walled from it. Every worked solution, both halves, stays free to read — with or without the pass.

And what it isn't. We'd rather tell you up front.

  • Not a coaching class or a lecture factory. No videos, no live teaching, no first-time instruction in a topic you've never seen.
  • Not where you learn CSAT from scratch. The free worked solutions and technique notes are for that. The engine assumes you're practising to score.
  • Not a fear machine. No countdown, no "you'll fail," no dramatised negative marking, no leaderboard, no streak guilt.
  • Not a guarantee. No promised clearance, no rank, no score. We never sell relief from cutoff fear.
  • Not a vocabulary drill or a grammar app. CSAT comprehension is reasoning from a passage, and that's how we practise it — we don't hand you word-lists for a test that doesn't reward them.

Keep your coaching. Keep your test series. The engine is the diagnostic-and-practice layer on top that turns your attempts into a specialist's read on you. It works with how you prepare. It doesn't replace it.

Why a maths house? Because the half that fails most people is ours.

The quantitative half is the one that trips serious, capable candidates — and it's the half a math specialist can take apart down to the mechanism of every trap. That's the thing a generic factory can't author: not just the right answer, but why the wrong one was built to look right. That's the edge.

And the method — take the paper apart, curate a gold bank, adapt it to you — is exactly what a rigour-and-statistics house does for a living. The maths is why you can trust that the rigour is real, not a slogan.

The honest both-halves frame

We don't solve reading with equations — comprehension isn't a formula, and we won't pretend it is. What a rigour-and-statistics house does bring to it is discipline: every keyed answer is defensible from the passage, authored by a verbal specialist, with the trap in each wrong option named the same way we name a quant trap. The answer is always there in the text — something you find, not a feeling you gamble on. Both halves are live, held to one standard. The maths is why you can trust that the rigour on the reading half is real, not a slogan.

If you've been told your English is the problem, here's the truth the exam itself proves: in ten years, not one CSAT comprehension question tested vocabulary or grammar. Every mark is reasoning from a short passage — roughly six in ten turn on a single skill: choosing the inference the passage actually supports over the one that merely sounds right. The word-stock you fear you're missing is not the thing the exam punishes. Comprehension is rule-governed reading. It can be trained — and that's exactly the half we built the practice for.

Seven question-types. Zero vocabulary tests. Ten years.

Every one of 284 CSAT verbal-reasoning questions since 2016 sorts into just seven recurring types — six built on a short passage, one a free-standing logic puzzle — and not one turns on a rare word. It’s a reasoning test, not a memory game.

One loop. It gets sharper every time you use it.

  1. Pick a paper, sit it for real. Real CSAT mechanics, real −1/3 negative marking, your pace first — nothing is graded against you until you choose it.
  2. Get the specialist's read. For every miss: why that distractor was built to catch you, the cleaner method, the skill it reveals.
  3. Get your worksheet. A guided sheet auto-built from your weak patterns — winnable first, method shown.
  4. It schedules your next sitting. Spaced to make it stick, pushed to your calendar in a tap.

A schedule that works for you, not over you.

Most apps hand you a streak and let it guilt you. We don't. The engine teaches you spaced repetition — the settled science that returning to a topic across days, instead of cramming it once, is how it lasts to exam day — and it sequences your next sittings for you. One tap puts the next one in your calendar. Miss a day? Nothing breaks. There's no streak to lose, and a missed sitting is never a setback. The schedule works for you, not over you.

The knowledge is free. Forever.

Reading a worked solution never costs anything, with or without a pass — that's the line that never moves. Our free public CSAT solutions are rolling out now, both halves, year by year. What you pay for is the engine: it measures you, diagnoses you, builds your worksheet, and schedules you. Knowledge is free. The personalisation is the product.

Start free. Pay once, when you're ready.

Start free. Read our worked solutions free, always — and take your first graded set — both halves, with its diagnostics, free, so you feel the engine work once before you decide anything. No card to start.

When you're ready for the full loop, here's the price — one number, the real one.

Our Mathematics Daily Practice was piloted by 18 aspirants before we opened it to everyone.

CSAT Pass

₹699 · one-time

The real price. No inflated “MRP,” no countdown.

One-time pass — valid through the 2027 Prelims. No subscription, nothing auto-renews. A 7-day refund, no questions, on the order page.

Get the CSAT Pass — ₹699

Why is a specialist's full-cycle CSAT practice ₹699, when a focused test series runs ₹500–₹3,000 and CSAT bundled into a prelims package sits inside ₹5,000–₹40,000? Because of how this is built. We take the paper apart once and that work serves every aspirant after — so serving one more costs us a few rupees, not the faculty, brand, and ad spend a coaching package carries. ₹699 isn't a sale. It's the actual shape of the thing — and it's something a person-heavy model genuinely can't match.

And because you've been shown a fake "discount" before — here's exactly how we'll show you a price, every time:

How we price

The price you see is the real price.

You know the trick. A price slashed in half. A banner that's been “ending tonight” for months. A “70% off” that somehow never runs out. The high number was never real — so the saving isn't either.

We don't do that. Our price is the price — the same whether you open this today or next month. And if we ever discount, it's a real campaign with a real end date you can hold us to.

Honest answers to what you're probably thinking.

Do I need to be good at maths for this?

No. That's what it's for. It diagnoses where you're losing marks and rebuilds your practice around exactly that — and when you're stuck, it routes you to the free worked solution. You don't start good. You get there.

Is this instead of my coaching?

No. Keep your coaching and your test series. This is the layer on top that turns your attempts into a specialist's read on you. It complements how you prepare; it doesn't replace it.

Is this the whole paper, or just the maths half?

The whole paper. Both halves are live — quant and comprehension, one pass. We started with quant because it's the half a math house can take apart down to the mechanism of every trap, and the comprehension half is held to the same standard: every keyed answer defensible from the passage, every wrong option's trap named. And here's the relief most CSAT prep won't tell you — comprehension is reasoning from a passage, not a vocabulary test. In ten years, not one question turned on word-stock or grammar. It's far more trainable than the fear suggests.

Will it tell me whether I'll clear?

It gives you an honest, conservative read on where you stand — necessary, not sufficient. It will never hand you a pass/fail verdict you'd bet a year on. We think the truth serves you better than false comfort.

Is it really paid? What's free?

Reading a worked solution is free, always — our public CSAT solutions are rolling out now, both halves — plus your first graded set, both halves, with its diagnostics, free. The engine — repeat graded attempts, the specialist read at depth, your worksheets, the schedule — is the one-time ₹699 pass. No subscription.

Will the negative marking punish me here?

The engine uses the real −1/3 marking because the exam does — but here it's data we teach from, not a stick to beat you with. It shows you which misses were the engineered trap, so the marking becomes something you learn to beat, not something to dread.

Start with one free attempt. See the engine work before you decide a thing.

Sign in & take your free attempt

Free to start, no card. Every solution stays free to read whether you ever buy or not. One-time price, no subscription, 7-day refund.

Every worked solution and every guide chapter is free to read, forever, with or without the pass.

Also taking Maths? Its Daily Practice is free for everyone through the 2026 Mains.