CAT Attempt Strategy by Percentile Target: Work Backwards From the Number
Most attempt advice is a single figure handed to everyone. Attempt twenty in quant. Read three passages. Do four DILR sets. It is confidently stated and it is useless, because the right attempt count is a function of two things that differ enormously between candidates — the percentile you actually need, and the accuracy you can hold while reaching for it.
So work backwards instead. Start from the target, convert it into a rough scaled score, and only then ask how many attempts at what accuracy gets you there. Two candidates chasing 95 percentile — one holding 85% accuracy, one holding 65% — need materially different counts. Telling both "attempt twenty" guarantees that one under-attempts and the other bleeds marks.
Targets, and what they cost in attempts
Figures below are working bands rather than target lines. The marks required for a given percentile shift by several points between cycles, and every table on this subject — including this one — is reconstructed from shared scorecards rather than published by the IIMs. Plan against the range, and build in the margin the underlying data already carries.
| Target | Rough scaled score | At 85% accuracy | At 70% accuracy | At 60% accuracy |
|---|---|---|---|---|
| 99+ | 95–105 | ~42 attempts | ~56 attempts | Not reachable |
| 95 | 66–74 | ~29 attempts | ~39 attempts | ~50 attempts |
| 90 | 53–61 | ~24 attempts | ~32 attempts | ~41 attempts |
| 85 | 45–52 | ~20 attempts | ~27 attempts | ~34 attempts |
Three marks for a correct MCQ, one deducted for a wrong one. Rearrange it and you get the attempt count a target demands at whatever accuracy you can actually sustain — which is a far more honest planning tool than any universal number.
Look at the bottom-right corner and the top-right corner together. At 60% accuracy, 99 percentile is not reachable inside 68 questions — you would need more attempts than the paper contains. That is not a motivational statement, it is arithmetic, and it makes the point better than any amount of advice about working harder.
Above a certain target, accuracy stops being one lever among several and becomes the only one. A candidate at 60% accuracy chasing 99 does not have an attempts problem to solve. They have an accuracy problem wearing an attempts problem's clothes.
Splitting the target across three sections
The overall number is not the binding constraint for most candidates. Sectional minimums are — almost every institute screens them separately, so a lopsided profile fails filters a balanced one clears.
Which means the split matters more than the total. Divide your target score across the three sections in proportion to your current sectional percentiles rather than evenly, then check that the weakest one still clears the minimums at your target institutes.
Buy attempts with passage selection
Do not attempt more passages. Attempt the same number, chosen better — ninety seconds of scanning converts a hard passage into an easier one at no cost in reading time.
Attempts are the wrong metric entirely
Count sets completed, not questions attempted. Three sets finished beats five started, because a part-built grid answers close to nothing.
The first pass is where attempts come from
Fifteen minutes harvesting everything you can see a route to will add more attempts than any amount of pushing on hard questions.
Which lever to pull, given where you are
Take your last three mocks. Average the attempts and the accuracy separately, by section. Then read the row that matches.
CAT Mock Test — get the two numbers this page runs on
None of the above works without your real attempt count and accuracy, by section. One full paper produces both.
Where the arithmetic stops helping
There is a second limit worth naming honestly. Expected value describes what happens on average across many decisions, and you are sitting one paper. If you are close to a sectional cut-off, the spread of outcomes matters alongside the average, and a positive-value gamble that could drop you below a threshold is not automatically correct. The arithmetic is a planning tool rather than a rule for every individual question.
Which lever should you pull?
Attempts and accuracy need opposite fixes and are routinely confused. The section-wise split tells you which one you actually have.
- Attempts vs accuracy, per section
- Sectional percentiles
- Time per question
Questions candidates actually ask
How many questions should I attempt in CAT?
There is no single number, and any page that gives you one is guessing on your behalf. The right count depends on the percentile you need and the accuracy you can hold while reaching for it — score equals attempts multiplied by four times your accuracy, minus one. Two candidates targeting 95 percentile, one at 85% accuracy and one at 65%, need materially different attempt counts, so work backwards from your own two numbers.
How many attempts do I need for 99 percentile in CAT?
Around 42 at 85% accuracy, or roughly 56 at 70% — and at 60% accuracy it is not reachable inside 68 questions at all, because you would need more attempts than the paper contains. That last figure is worth sitting with. Above a certain target, accuracy stops being one lever among several and becomes the only one that matters.
Is it better to attempt more questions or be more accurate?
It depends on where your accuracy currently sits. Below about 65%, additional attempts add very little and cost time you could spend getting the ones you take right — so attempt fewer. Between 65% and 80%, both levers work and attempts move faster, which is where most candidates and most available gains sit. Above 85% on a low attempt count you are being too careful, and adding attempts will raise your total.
Should I attempt the same number of questions in every section?
No — and in DILR, attempts are the wrong metric altogether. Count sets completed rather than questions attempted, because the questions hang off a shared structure and a part-built grid answers close to nothing. Split your target across sections in proportion to your current sectional percentiles rather than evenly, then check the weakest still clears the sectional minimums at your target institutes.
Does the attempt formula apply to TITA questions?
No, and the difference works in your favour. Typed-answer questions carry no negative marking, so their expected value is positive at any accuracy above zero and a blank scores exactly what a wrong answer scores. Every equation on this page governs multiple-choice questions only. Attempt every typed-answer question you meet, whatever your target and whatever your confidence.
How do I find my real accuracy?
Take your last three mocks and average attempts and accuracy separately, by section, rather than looking at overall figures. Accuracy means correct answers as a proportion of questions you chose to attempt — not of all 68. Candidates routinely compute it the second way, which produces a flattering number that makes every subsequent planning decision wrong in the same direction.
Is the expected-value approach always right?
As a planning tool, yes. As a rule for every individual question, not quite — expected value describes what happens on average across many decisions, and you are sitting one paper. If you are close to a sectional cut-off, the spread of outcomes matters alongside the average, so a positive-value gamble that could drop you under a threshold is not automatically correct. It is a narrow exception, and worth naming rather than pretending one equation covers everything.
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.
