How Many Questions Should You Attempt in CAT for a 99 Percentile? (The Data)

Most people preparing for CAT quietly believe the exam is a race to solve everything. It isn't. The data says something almost the opposite: you can leave roughly a third of the paper untouched and still walk out with a 99 percentile.

There were 68 questions in CAT 2024. You do not need to solve all of them to reach a high percentile. But “45 correct answers” and “45 attempts” are very different: 45 correct answers earn 135 marks before any wrong-MCQ penalty, while 45 attempts can produce many different scores. Choosing which questions to pursue, and which to leave, is central to the strategy.

Let's look at the actual numbers.

First, the math of the paper

Before attempts make sense, you need the structure they sit inside. CAT 2024 ran 68 questions across three sections, 40 minutes each:

Every question is worth +3 marks. Every wrong MCQ costs you −1. The non-MCQ (TITA) questions, where you type the answer instead of picking an option, carry no negative marking — which matters more than most people realise, and we'll come back to it.

That's 68 questions and 204 marks on the table. Now, how many of them do you actually need?

The section-wise attempt data

Here is a way to think about section-wise attempt budgets. The ranges below are illustrative starting points, not validated thresholds for 90, 95 or 99 percentile. The totals are the actual sums of the section ranges. Adjust your own budget using timed mocks and accuracy, rather than adopting someone else’s target.

Read the bottom row again: a higher attempt budget is not automatically a higher score. You can leave questions untouched and still perform well. Your correct answers, wrong MCQs, slot scaling and the year’s score distribution determine the result — not attempts alone.

Notice something else: the higher budget puts more attempts into VARC than DILR or QA. That can be a sensible personal strategy, but it is not a rule for every top scorer. Some DILR sets and QA questions consume much more time than others; part of the skill is recognising when to move on.

Attempts alone are a lie — accuracy is the other half

If attempts were the whole story, you could just fill in 68 answers and win. Negative marking makes sure you can't. Every wrong MCQ doesn't just fail to score, it actively pulls your total down, which means a paper attempted carelessly can score lower than a smaller, cleaner one.

For context, the following are approximate CAT 2025 score–percentile figures reported by Cracku, not CAT 2024 or guaranteed CAT 2026 targets. Raw marks, scaled marks and accuracy are different measures, so there is no single accuracy percentage attached to each percentile.

The single most useful line in all of this data: 40 questions attempted at 85% accuracy beats 48 questions at 65% accuracy. Fewer, cleaner attempts win. Every time you talk yourself into a shaky guess on an MCQ to push your attempt count up, you're usually trading a −1 for the feeling of having "done more."

There is a nuance by section, too. VARC, DILR and QA have different time demands, but a wrong MCQ costs −1 in all three. A genuine two-option elimination can make a guess worthwhile; it does not remove the penalty. Track whether your uncertain answers actually help your score instead of using a blanket accuracy rule.

This is also where no-negative-marking TITA questions earn their keep. A wrong answer carries no marking penalty, though time spent still has an opportunity cost. Leaving a solvable TITA blank can mean missing available marks.

What the data actually tells you to do

Strip it all down and the strategy writes itself: attempt selectively, protect your accuracy, and stop chasing volume.

In practice that means going into every section planning to leave questions behind. It means having a rule for when to abandon a question — you might use a two-and-a-half-minute checkpoint for an individual QA question without traction, then adjust it using your own results; a DILR set needs a different checkpoint — instead of stubbornly feeding time into one problem while easier marks sit unread further down. And it means treating your attempt count not as a score to maximise but as a budget to spend wisely.

The problem is that almost none of this is visible from your raw mock score. For an all-MCQ illustration, 50 attempts with 30 correct and 20 wrong score 70. So do 30 attempts with 25 correct and 5 wrong. Those students look identical on the scoreboard, but one needs to protect accuracy while the other may need to expand coverage. Break the mock down attempt by attempt to distinguish them. Our companion article explains why scores plateau and why diagnosis matters.

The trap: mistaking activity for a plan

The reason this data matters is that the instinct it corrects is so strong. Under exam pressure, leaving questions blank feels like failure, so people attempt more, accuracy drops, negative marking bites, and the score falls — the exact opposite of what they intended. Then they walk out saying the paper was tough, when really the strategy was.

Knowing the numbers ahead of time helps you override that instinct. Once you understand that a smaller, accurate attempt count can outscore a larger, careless one, skipping stops feeling like giving up and starts feeling like a deliberate choice.

Where Marg fits in

This is the whole reason Marg exists, and yes, it's our tool. Marg is a free AI mentor for CAT 2026 that reads your mocks the way this article reads the exam: it separates your score from your execution, so it can tell you whether you're losing marks to weak accuracy, wasted attempts, or genuine gaps — and then it builds daily practice around whichever one is actually holding you back. Instead of a number you can't act on, you get the attempt-and-accuracy breakdown that tells you what to change before the next mock. You can reconstruct all of this yourself with a careful spreadsheet after every mock. Marg can help you examine that evidence and decide what to test next.

The short version

You don’t need to solve the whole CAT paper. You need an attempt budget that your accuracy can support. Use historical score–percentile figures as context, not a promise. Be selective in DILR and QA, remember that TITA has no wrong-answer penalty but still uses time, and measure each mock by correct answers, wrong MCQs and time lost — not just how much of the paper you touched.

Preparing for CAT 2026? Marg is a free AI mentor that breaks down your mocks by attempts and accuracy, tracks the mistakes you keep repeating, and builds daily practice around your real weak spots.