CAT Good Attempts: How Many Questions to Attempt in Each Section
There is no universal "good attempt" number in CAT, and any page that gives you one without attaching an accuracy figure is telling you nothing useful. Attempts and accuracy are a single decision, not two — and the marking scheme decides where the balance sits.
With +3 for a correct answer and −1 for a wrong MCQ, CAT is far more forgiving than SSC or banking exams. A blind four-option guess breaks exactly even. That changes the right answer to "how many should I attempt" more than most candidates realise.
The number that reframes everything
(0.25 × 3) − (0.75 × 1) = 0. A blind guess costs you nothing in marks, only seconds. Eliminate two options and the same guess is worth +1.00. Candidates carrying caution over from exams with harsher penalties routinely under-attempt by five or six questions.
Quick Answer (30-Second Read)
Working ranges
VARC 16–20 · DILR 10–14 · QA 14–18
At 80%+ accuracy
The real rule
Accuracy first, then volume
Below 70%, attempt fewer
- Attempt count means nothing without accuracy. 18 at 85% beats 24 at 60% in every section.
- Never leave a TITA blank. Wrong and blank both score zero, so a reasoned guess is free upside.
- Guess an MCQ after eliminating two options — expected value +1.00. Below that, the time cost outweighs it.
- DILR is measured in sets, not questions. Two sets finished beats four touched.
- Find your own number by plotting attempts against final marks across five mocks. The turning point is yours.
Marking scheme per the official CAT information bulletin at iimcat.ac.in. Attempt ranges below are working benchmarks, not official figures.
CAT Mock Test — VARC, DILR & QA on the real interface
Your optimal attempt count cannot be read off a table. It comes from your own accuracy curve, and five timed mocks is what it takes to plot it.
The arithmetic nobody shows you
Work it through with the real marking scheme and the "how many attempts" question answers itself. Take a VARC section of 24 questions:
| Attempts | Accuracy | Correct | Wrong | Gained | Lost | Net |
|---|---|---|---|---|---|---|
| 24 | 55% | 13 | 11 | 39 | −11 | 28 |
| 22 | 65% | 14 | 8 | 42 | −8 | 34 |
| 20 | 75% | 15 | 5 | 45 | −5 | 40 |
| 18 | 85% | 15 | 3 | 45 | −3 | 42 |
| 14 | 93% | 13 | 1 | 39 | −1 | 38 |
The peak sits at 18 attempts, not 24 — and the candidate attempting every question finishes 14 marks behind despite touching six more. But look at the last row too: at 14 attempts and 93% accuracy the score falls again. Over-caution costs as much as recklessness, just more quietly.
The curve, not the number
Marks rise with attempts, peak, then fall. Where the peak sits depends entirely on your accuracy — a more accurate candidate peaks further right. Your job is not to hit some published figure but to find where your curve turns over, which the last section of this page shows you how to do.
Section by section
Working benchmarks to test against your own data, not rules to follow blindly.
| Section | Questions | Working range | Target accuracy | Why this range |
|---|---|---|---|---|
| VARC | 24 | 16–20 | 80%+ | RC sets reward re-reading. Rushing all 24 collapses accuracy across the whole section |
| DILR | 22 | 10–14 | 85%+ | Set-based. Two or three sets finished cleanly, not questions counted individually |
| QA | 22 | 14–18 | 80%+ | Some questions are deliberately time sinks. Leaving four is correct, not a failure |
DILR is the section where the attempt count misleads most. Ten questions attempted from two fully solved sets is a strong performance. Fourteen questions attempted across four half-cracked sets is usually a poor one, because DILR questions depend on the full arrangement — a set you did not crack yields close to nothing regardless of how many of its questions you marked.
In VARC, the temptation is the opposite. Twenty-four questions look attemptable because each RC question is short. But the reading is the expensive part, and skimming a passage to reach every question attached to it drops accuracy across all of them at once.
Where does your accuracy curve actually turn over?
The table above is a starting point. Five timed mocks give you the real answer, plotted from your own attempts and marks.
- Accuracy vs attempts
- Marks lost to negatives
- Blank TITA count
When a guess is worth taking
| Situation | Chance correct | Expected marks | Verdict |
|---|---|---|---|
| Blind guess, 4 options | 25% | (0.25×3) − (0.75×1) = 0.00 | Break-even. Costs seconds, not marks |
| Eliminated 1 | 33% | (0.33×3) − (0.67×1) = +0.32 | Weakly positive |
| Eliminated 2 | 50% | (0.50×3) − (0.50×1) = +1.00 | Attempt it |
| Fairly confident | 70% | (0.70×3) − (0.30×1) = +1.80 | Always attempt |
| Any TITA | anything | No penalty exists | Always attempt |
The break-even row is the one worth internalising. In SSC CGL a blind guess loses you marks outright. In CAT it does not. That single difference means the correct instruction is "eliminate first, then guess freely" rather than "when in doubt, leave it" — and candidates who move between exam types without adjusting leave marks on the table every attempt.
TITA questions change the count
TITA questions — type in the answer — carry no negative marking at all. A wrong TITA and a blank TITA both score zero.
This means your attempt target should be read as an MCQ target. TITA questions sit outside it entirely, because there is no version of the arithmetic in which leaving one blank is correct.
Two minutes left in QA. Three TITA questions untouched. You cannot fully solve any of them.
The instinct. Leave them — you are "not sure", a habit from exams with penalties.
The arithmetic. Wrong TITA = 0. Blank TITA = 0. Identical downside, +3 upside.
The move. Put your best reasoned estimate on all three. If you narrowed one to "around 40", type 40.
Expected value of guessing: positive. Of leaving blank: exactly zero.
Check the report after every mock for how many TITA questions you left blank. It is the single easiest habit to fix and the one candidates are least aware of, because a blank question does not feel like a mistake in the way a wrong answer does.
Finding your own number
Five mocks, three columns. That is the whole method.
| Mock | QA attempts | Accuracy | QA marks | Read this as |
|---|---|---|---|---|
| 1 | 12 | 92% | 32 | Room to push |
| 2 | 15 | 87% | 37 | Still rising |
| 3 | 18 | 83% | 42 | Still rising |
| 4 | 20 | 70% | 36 | Turned over — the edge is around 18 |
| 5 | 22 | 64% | 30 | Past the peak, losing marks |
These figures illustrate the method rather than describing any real candidate — yours will differ, and that is the point. Run the same three columns per section, because your optimum in VARC and your optimum in QA are different numbers driven by different constraints.
Recalculate every few weeks. As accuracy improves the peak moves right, so an attempt count that was optimal in September will be too cautious by November. Candidates who fix a number early and never revisit it end up under-attempting by the time it matters.
If both accuracy and attempts look bad
Fix accuracy first, always. Speed built on shaky concepts produces wrong answers faster, and under negative marking that actively loses marks. Accuracy is a stable property you can raise through study; attempt count is a dial you can turn in a week once the foundation holds.
Key Takeaways
- Attempts and accuracy are one decision. 18 at 85% beats 24 at 55% by 14 marks in a 24-question section.
- A blind CAT guess breaks exactly even — it costs seconds, not marks. Eliminate two options and it is worth +1.00.
- TITA questions sit outside your attempt target. Wrong and blank both score zero, so never leave one.
- DILR is counted in sets, not questions. Two sets finished beats four half-attempted.
- Over-caution costs as much as recklessness, just more quietly — the curve falls on both sides of the peak.
- Plot attempts against marks across five mocks. The turning point is your number, and it moves as accuracy improves.
People also search for
How many questions should I attempt in CAT?
There is no single correct number, because attempts and accuracy are one decision rather than two. As working benchmarks at 80 percent accuracy or better, VARC sits around 16 to 20 of 24, DILR around 10 to 14 of 22, and QA around 14 to 18 of 22. But these are starting points to test against your own mock data. Plot your attempts against your final marks across five mocks and the point where marks stop rising is your genuine optimum.
Is it better to attempt more questions or be more accurate in CAT?
Accuracy first, in almost every case. Accuracy is a stable property reflecting how well you know the material, and raising it improves your score at every attempt level at once. Attempt count is a tactical dial you can turn within a week once accuracy is solid. Pushing attempts while accuracy sits below about 70 percent actively loses marks under the minus-one penalty. That said, over-caution has a cost too — the marks curve falls on both sides of its peak.
What is a good number of attempts in CAT VARC?
Around 16 to 20 of the 24 questions, at 80 percent accuracy or better. The temptation in VARC is to attempt all 24, because each individual question looks short. The expensive part is the reading, though, and skimming a passage to reach every question attached to it drops your accuracy across all of them simultaneously. Reading two passages properly and leaving a third usually outscores rushing all of them.
How many DILR questions should I attempt?
Think in sets rather than questions. Two to three sets solved completely, which typically works out to 10 to 14 questions, is a strong DILR performance. DILR questions within a set depend on one shared arrangement, so a set you failed to crack yields close to nothing no matter how many of its questions you marked. Fourteen questions spread across four half-solved sets almost always scores worse than ten questions from two finished ones.
Should I guess in CAT if I do not know the answer?
On MCQs, guess once you have eliminated at least two of the four options — that gives an expected value of +1.00 marks, which is clearly worth taking. A blind four-option guess has an expected value of exactly zero, so it costs you nothing in marks but does cost seconds, making it rarely worthwhile. On TITA questions always attempt, because there is no penalty at all and a wrong answer scores identically to a blank one.
Does CAT have negative marking for TITA questions?
No. Type In The Answer questions carry no negative marking, so a wrong TITA and an unattempted TITA both score zero. This means your attempt target should really be read as an MCQ target, with TITA questions sitting outside it entirely — there is no version of the arithmetic where leaving one blank is the right call. Check your mock report for how many you left blank, since a blank question does not register as a mistake the way a wrong answer does.
How many attempts are needed for 99 percentile in CAT?
The honest answer is that percentile depends on rank rather than attempts, and rank depends on how everyone else performed on that year's paper. Two candidates with the same attempt count and very different accuracy end up in completely different percentile bands. Rather than targeting an attempt figure, target accuracy above 80 percent in each section and let your attempt count settle wherever your own accuracy curve peaks — then track sectional percentiles in mocks to calibrate.
Why do I score less when I attempt more questions?
Because you have pushed past the peak of your accuracy curve. Every additional attempt beyond a certain point comes from progressively less certain ground, so the marginal question is increasingly likely to be wrong and cost you one mark rather than gain three. In a 24-question section, attempting 24 at 55 percent accuracy scores 28 while attempting 18 at 85 percent scores 42. More work, fewer marks. The fix is to attempt fewer and read more carefully, not to work faster.
How is CAT attempt strategy different from SSC or banking exams?
The penalty ratio changes everything. CAT awards +3 and deducts −1, so a blind four-option guess breaks exactly even. SSC CGL deducts 0.50 against 2 marks, and IBPS deducts 0.25 against 1, so blind guessing there is closer to break-even but the practical guidance has hardened into caution. Candidates moving from those exams to CAT routinely under-attempt by five or six questions because they carry the caution across without recalculating. Eliminate two options in CAT and guessing is clearly profitable.
Should my attempt strategy change between mocks and the real exam?
No, and treating them differently is a common way to lose marks. If you attempt cautiously in mocks and then push hard in the real exam, you are running an untested strategy on the only attempt that counts. The point of mocks is to find the attempt count where your marks peak and then rehearse it until it is automatic. Recalculate every few weeks as your accuracy improves, since the optimum moves right over time — but change it in a mock first, never on exam day.
Like what you read? Share with a friend
Senior Content Writer
Co-Founder at PrepGrind, working on education content, product development, and student-focused learning resources. I also write articles and guides for MBA/MBS entrance exams, covering preparation strategies, exam updates, and useful resources for students.
