CAT Test Strategy: What to Attempt and What to Leave (Section-wise Guide for CAT 2026)

Sep 27

8 min read

CAT test strategy comes down to one decision, repeated about seventy times in two hours: attempt this question, park it, or leave it. Most aspirants prepare for months and then make that decision by instinct on exam day. The ones who score 99+ percentile make it by rule. This guide lays out those rules section by section, with the numbers behind them, so you can practise them in your mocks between now and CAT 2026.

If you would rather hear this live, the same method is taught in FIIB's free Smarter CAT Preparation masterclass series by Abhishek Anand (Founder, Mentors Capital; MBA, IIM Bangalore). Episode details and registration are on the CAT Test Strategy masterclass page.

The short answer

CAT has three sections of 40 minutes each: VARC, DILR and Quant. Time does not carry over between sections.
A correct MCQ earns +3. A wrong MCQ costs −1. TITA (type-in-the-answer) questions carry no negative marking.
99 percentile in recent CATs has needed roughly 100 to 105 out of 198 marks, depending on the slot and year. That is about half the paper, answered at 85%+ accuracy.
So the job is not to finish the paper. The job is to find the easiest half, answer it accurately, and refuse to be dragged into the other half.

Everything below is about how to do that under a clock.

Why attempt strategy matters more than one more formula

Percentile is relative. You are ranked against roughly three lakh people sitting the same paper, and the questions that separate the 90th percentile from the 99th are rarely the hardest ones. They are the medium questions that most people leave because they ran out of time on a hard question earlier.

Think about the arithmetic. Three wrong MCQs wipe out one correct one. Six minutes sunk into a DILR set you eventually abandon is six minutes you did not spend on a set you could have solved. A 40-minute section does not reward courage; it rewards selection.

Abhishek Anand puts it this way in the masterclass: CAT is a series of small decisions made under a clock, and the students who practise the decisions score higher than the students who only practise the maths.

The 30-second scan

Before you solve anything, read the question or the set once and classify it. You are allowed about 30 seconds for this, and you should train yourself to stick to it.

Attempt now. You recognise the type, you can see the method, and you are confident of finishing in the time this question deserves (roughly two minutes for a Quant MCQ, eight to ten minutes for a DILR set).
Park. You know how to do it, but it looks long or has a trap you want to check. Mark it for review and move on. You will come back if time allows.
Leave. You do not see the method within 30 seconds, or the set has too many variables to hold in your head. Do not touch it. Do not guess unless it is TITA.

The rule most students break is the second one. Parking feels like giving up. It is not. Parking is how you protect the next ten questions from the one in front of you.

Section-wise attempt order

VARC (Verbal Ability and Reading Comprehension)

Recent CAT VARC sections have had 24 questions: four RC passages (16 questions) and eight Verbal Ability questions (para jumbles, para summary, odd sentence out, para completion).

Scan all four passages first. Spend the first two minutes reading the opening paragraph of each and rank them by comfort. Abstract philosophy and dense social science usually cost more time per mark than narrative or business passages.
Attempt three passages you like. Leave the fourth unless you have eight minutes left with everything else done.
Do the VA questions between passages, not at the end. Para summary and odd-one-out are high accuracy for most students and take under two minutes each. Para jumbles are TITA, so a guess costs nothing, but do not spend more than two minutes forming a sequence.
Target for 99 percentile: 15 to 17 attempts at 80%+ accuracy. Attempting all 24 at 60% accuracy scores lower.

DILR (Data Interpretation and Logical Reasoning)

DILR is where percentiles are made and lost, because the variance between sets is so large. A typical section has four or five sets of four to six questions each; two sets are usually doable by most serious aspirants, one or two are hard, and one is often a trap dressed as an easy set.

Spend the first five minutes reading every set. Do not solve. Rank them.
Start with the set whose structure you recognise: an arrangement, a distribution, a table with straightforward totals. Finish it completely; partial sets waste the setup time you invested.
Move to the second easiest set. If you have solved two sets cleanly by minute 28, you are on track for 95+ percentile in DILR.
Use the remaining time on the third set only if you can see the way in. Otherwise pick off any standalone or TITA questions in it and stop.
Never start a fourth set with less than eight minutes left. The set-up time alone will eat what remains.
Target for 99 percentile: two complete sets plus two or three questions from a third, at near-perfect accuracy. That is 12 to 14 attempts.

QA (Quantitative Aptitude)

Quant rewards breadth of easy questions rather than depth on hard ones. A 22-question section usually has eight to ten questions that a well-prepared student can solve in under two minutes, another six to eight that take three to four minutes, and four to six that are designed to consume time.

First pass, 15 to 18 minutes: go through all 22 questions in order, applying the 30-second scan. Solve only the ones you can see the method for immediately. Mark the rest as park or leave.
Second pass, 15 minutes: return to the parked questions in the order you marked them, easiest first.
Last five to seven minutes: check any TITA questions you have left blank and put an educated guess in; there is no penalty. Do not guess on MCQs unless you have eliminated two options.
Arithmetic (percentages, ratios, time-speed-distance, averages) and basic algebra have been the largest share of the section for several years and are the highest-return topics for a first pass. Geometry and number theory questions are more often the ones designed to slow you down.
Target for 99 percentile: 14 to 16 attempts at 85%+ accuracy.

Attempts versus accuracy: the numbers

The table below is a rough guide built from recent years' patterns. Exact cut-offs move with the difficulty of each slot, so treat these as targets for your mocks, not guarantees.

90 percentile: around 60 to 65 marks. Roughly 26 to 30 attempts across three sections at 75% to 80% accuracy.
95 percentile: around 75 to 80 marks. Roughly 32 to 35 attempts at 80% accuracy.
99 percentile: around 100 to 105 marks. Roughly 40 to 44 attempts at 85%+ accuracy.
99.5 percentile: around 110 to 115 marks. Roughly 44 to 48 attempts at 88%+ accuracy.

Notice what is not on that list: attempting 60 questions. Nobody who scores 99.5 does it by attempting everything. They do it by attempting the right two-thirds and getting almost all of it right.

Now compare this with your own last mock. If you attempted 45 questions at 65% accuracy, you scored roughly 45 × 0.65 × 3 − 45 × 0.35 × 1 = about 72 marks. Drop to 36 attempts at 85% accuracy and the same knowledge gives you 36 × 0.85 × 3 − 36 × 0.15 × 1 = about 86 marks. Fourteen more marks, from attempting fewer questions. That is the whole argument for attempt strategy in one line.

A mock-test routine for the last ten weeks

Taking more mocks does not raise your score. Changing something after each mock does. Here is the routine the masterclass recommends.

Take the mock at CAT timing. Same slot, same three-section order, no pauses. Attempt strategy only works if you practise it under the real clock.
Before checking answers, write down your attempts per section and your predicted accuracy. The gap between predicted and actual accuracy is your biggest diagnostic. If you thought you were 90% accurate and were 70%, you are attempting questions you should be parking.
Sort every question into four buckets: attempted and correct, attempted and wrong, left and could have solved, left and correctly left. The second and third buckets are where your marks are.
For every wrong attempt, ask one question: should I have known at the 30-second mark that this was a leave? If yes, that is a selection error, and selection errors are fixed faster than knowledge gaps.
Pick one change for the next mock. Not five. One. "Read all DILR sets before starting" or "stop at two RC passages if minute 25 has passed." Then take the next mock and see if the change held.
Track attempts, accuracy and score in a single sheet across all mocks. After four or five mocks, the pattern is obvious: most students find their score rises when attempts fall and accuracy climbs, and then rises again when attempts creep back up at the higher accuracy.

Common mistakes that cost percentile

Starting DILR without reading all the sets. The first set on the page is not chosen to be the easiest.
Finishing an RC passage you should have abandoned. Sunk cost is not a reason to answer four more questions at 50% accuracy.
Guessing on MCQs to "not leave marks on the table." A blind guess has a 25% chance of +3 and a 75% chance of −1. The expected value is zero at best.
Skipping TITA questions. These have no negative marking; an educated guess is free.
Changing your attempt plan on exam day. If your last five mocks worked with 14 Quant attempts, exam day is not the day to try 20.
Treating the mock score as the goal. The mock is a diagnostic. The decisions you change afterwards are the goal.

Where to go from here

Practise the 30-second scan in your next mock and nothing else. Then come to the live sessions and bring your numbers.

FIIB's Smarter CAT Preparation is a free, seven-episode live series on Zoom, every Sunday from 20 September to 1 November 2026, taught by Abhishek Anand. The episodes cover test strategy, the 99+ playbook, DILR, VARC, Quant, mock analysis and a final strategy lab with an RC practice session. Each episode is one hour with 15 minutes of Q&A, and every registered aspirant gets the Zoom link by email immediately.

Frequently asked questions

How many questions should I attempt in CAT to get 99 percentile?

Roughly 40 to 44 attempts across all three sections at 85% or better accuracy, based on recent years' score-to-percentile patterns. The exact number moves with each slot's difficulty, so use it as a target for mocks rather than a fixed rule.

Is it better to attempt more questions or be more accurate in CAT?

Accuracy, almost always. With +3 for a correct MCQ and −1 for a wrong one, three wrong answers cancel one right one. Most students gain more marks by cutting low-confidence attempts than by adding attempts.

Should I attempt all RC passages in CAT?

Usually not. Three passages attempted well typically score more than four attempted in a rush. Leave the fourth unless you have finished everything else with time to spare.

How much time should I spend on a DILR set before leaving it?

If you have not found a way into the set within three to four minutes of starting it, stop. If you are solving but slowly, finish it; a half-solved set wastes the setup time you already spent.

Is the FIIB CAT masterclass free?

Yes. The Smarter CAT Preparation series is free and runs live on Zoom. Registration and episode details are on the masterclass page.