Skip to content
GATE Prep

GATE CS preparation — the complete strategy

A realistic plan for GATE Computer Science, built around weightage, timed practice and error journals rather than finishing more lectures.

AI-Shala Team 12 min read
32 WEEKS WHERE THE MARKS ARE Algorithms DBMS OS Networks Digital Compilers MOCK SCORES every Sunday

Most GATE aspirants do not fail on syllabus coverage. They finish the syllabus. They fail in the gap between recognising a problem and solving it under time pressure with a rank on the line.

This guide is about closing that gap. It assumes you are aiming at a top-1000 rank and have somewhere between six and twelve months.

Start from weightage, not from the syllabus order

The syllabus is a list. It is not a plan, and it is not ordered by what actually earns marks. Over recent years the distribution in CS has been broadly stable:

AreaApproximate weight
Engineering Mathematics + Discrete Maths13–15%
Algorithms + Data Structures12–14%
Operating Systems8–10%
DBMS7–9%
Computer Networks7–9%
Theory of Computation7–9%
Compiler Design5–7%
Computer Organisation & Architecture7–9%
Digital Logic5–6%
Programming (C)5–7%
General Aptitude15%

Two observations that change how people plan:

General Aptitude is 15% and most aspirants under-prepare it. It is the highest marks-per-hour section in the paper. Thirty minutes a week, consistently, is enough to secure most of it.

Maths plus Algorithms is roughly a quarter of the paper. If these two are shaky, no amount of Compiler Design will rescue the rank.

The three-phase plan

Phase 1 — Foundations (months 1–4)

Goal: cover every subject once, properly, with problems attempted as you go.

  • Do not watch a lecture without solving problems the same day. The illusion of understanding is strongest immediately after a good explanation.
  • Maintain one notebook per subject with only the results you keep re-deriving. If you look something up twice, it goes in the notebook.
  • Finish a subject with a 25-question timed set before you move on. Not to score — to find out what did not stick.

Order matters less than people think, with one exception: do Discrete Mathematics and Engineering Mathematics early. They underpin Algorithms, TOC and DBMS, and doing them late means relearning those subjects twice.

Phase 2 — Problem density (months 4–8)

This is where ranks are made, and where most preparation quietly stalls into re-watching lectures.

  • Previous year questions, topic-wise. Not chronologically. Take one topic and do every PYQ on it across fifteen years in one sitting. The pattern in how a topic is examined becomes obvious in a way it never does when questions are scattered.
  • Start an error journal. Every wrong answer gets one line: what the question tested, and the specific reason you got it wrong — concept gap, misread, arithmetic, or time panic. Review the journal weekly.

That last category matters. Most aspirants assume their mistakes are conceptual. When they actually classify them, a large fraction turn out to be misreads and time pressure — which require completely different fixes.

Phase 3 — Exam craft (final 8–10 weeks)

  • One full mock a week, at the exam’s actual time of day. Cognitive endurance at 9am is not the same as at 9pm, and you are training for a specific slot.
  • Spend longer on the debrief than on the mock. Three hours of paper, four hours of analysis. Anyone can take mocks; the rank comes from what you extract from them.
  • Build a skip strategy. Decide in advance how long a question gets before you leave it. Most aspirants lose more marks to one question they refused to abandon than to any topic they never studied.

Time, realistically

Full-time students: 6–8 focused hours a day is achievable and sufficient. More than that, most people are re-reading rather than learning.

Working professionals: 12–15 hours a week over 11 months works. Two hours on weekday mornings and a longer weekend block is the pattern that survives contact with a job. Three hours a week does not work, and it is kinder to know that at the start than in January.

The number that actually predicts outcomes is not hours studied. It is problems attempted under timing.

The negative marking calculation

In CS, wrong MCQ answers carry a penalty; NAT questions do not. This produces a simple rule that many aspirants get wrong:

  • NAT questions: always attempt. There is no downside. An informed guess is free.
  • MCQs: attempt if you have eliminated at least two options. At that point the expected value turns positive.
  • MCQs with no elimination: leave them. Pure guessing across a paper is statistically close to neutral and practically corrosive, because it encourages guessing on questions you could have solved with another ninety seconds.

What to actually cut

If you are behind, cut depth, not subjects. Every subject appears in the paper; a subject you have never opened is guaranteed lost marks. A subject you have covered to 60% depth still earns the straightforward questions, which is where most of its marks live anyway.

The exception is very late in the cycle. In the final month, do not start anything new. Consolidate what is already 70% there — moving a subject from 70% to 90% is faster and worth more than moving one from 0% to 40%.

The honest part

No plan guarantees a rank, and anyone selling you one is selling you something else. What a good plan does is remove the failure modes that have nothing to do with ability: uneven coverage, no timed practice, no feedback loop, and the slow drift into passive study that feels productive and is not.

The single highest-leverage change most aspirants can make is to have someone competent look at their error journal every week and tell them the truth about it.


Our GATE programme is built around exactly that loop — small cohorts, weekly one-to-ones, and 40+ mocks with personal debriefs. Our last cycle produced AIR 11 in CS.

AI-Shala Team

Research & Engineering

Written collectively by the people who build and teach here — engineers, researchers and mentors who spend their week with the problems these posts describe.

GATEGATE CSexam strategycomputer science

Keep reading

400 APPLICATIONS certificates, mostly THE ONE THAT GETS READ /sim-to-real-arm it runs, and there is a video eleven months of evenings
Careers
9 min read

AI and robotics internships in India — how to actually get one

What research and deeptech teams look for in intern applications, why most get rejected in thirty seconds, and what to build instead of another certificate.

AI-Shala Team Read
IN SIMULATION true state · clean torque · fixed timestep THE GAP ON HARDWARE sensor noise · latency · loop jitter
Robotics Pillar guide
10 min read

Sim-to-real transfer — why your policy fails on hardware

A policy that is perfect in simulation and useless in the lab is not a bug in your training. It is a specific, diagnosable gap. Here is how to find and close it.

AI-Shala Team Read
fifty papers read Submitted Under review Revise Accepted
Research Pillar guide
11 min read

Publishing your first AI paper as an undergraduate

What actually stands between a capable student and a first publication — and the order of operations that gets you from reading papers to submitting one.

AI-Shala Team Read

Reading is the easy part. Doing it with someone is better.

Our programmes exist because attention, not information, is what actually changes an outcome.