A data-backed look at why CAT mock scores stop improving — percentile math, the attempts-vs-accuracy tradeoff, and the real reasons aspirants plateau.

Every CAT aspirant hits the same wall eventually: scores climb for a few mocks, then stop. You study the same number of hours, take the same number of tests, and yet the percentile refuses to move. It feels personal. It isn't — it's math, and once you see the numbers, the plateau makes a lot more sense.

1. The percentile math nobody explains properly

Most aspirants chase a "score target" without understanding how forgiving — or unforgiving — the percentile curve actually is. CAT uses a normalization process to convert raw marks into a scaled score, and then into a percentile, which is why two candidates with the same raw score in different slots can land on different percentiles (Careers360).

CAT 2024 and CAT 2025 had a maximum possible score of 204 across 68 questions (Careers360), but very few candidates need anywhere close to that. In Cracku’s reported CAT 2025 data, a scaled score of about 84.8 corresponded to 99 percentile (Cracku), and in CAT 2025 a 99.90 percentile corresponded to an overall score of just 111.4 (iQuanta).

What trips people up is how sharply the curve bends near the top. A gap of just 5–10 marks can swing your percentile significantly at the higher end (Toprankers) — which means the difference between a call from an old IIM and no call at all is often smaller than aspirants think. And percentile alone doesn't tell the full admissions story either: most B-schools build a composite score using academics, work experience and other profile factors on top of your CAT percentile (Career Launcher), while individual IIMs also set their own sectional and category-wise cutoffs that a strong overall percentile alone won't automatically clear (Toprankers).

The takeaway: stop chasing a fixed score. Reverse-engineer your target from the percentile your shortlisted colleges actually need, with a safety margin — because at the top of the curve, small mark differences carry outsized weight.

2. Why "more attempts" is quietly costing you marks

This is the part most aspirants get backwards. CAT rewards precision, not volume — and the negative marking scheme is the reason why.

Because a correct MCQ earns +3 and a wrong one costs −1, every incorrect attempt isn't just a lost mark — it's a lost opportunity too, a net swing that can work out to roughly 4 points per mistake once you account for both (CATMock). At high percentile bands, the exam is genuinely unforgiving: 5 to 7 wrong MCQs can materially lower a score; the resulting percentile depends on the whole paper and the year (Quantifiers).

This plays out predictably across sections. There is no universal attempt count for a 99+ percentile. For illustration, 40 all-MCQ attempts with 36 correct and 4 wrong score 104 raw marks — without answering every question on the paper (CollegeDekho). Two candidates with near-identical preparation can land at different percentiles because one prioritized selective, confident attempts while the other went for volume (Quantifiers).

The takeaway: your mock score isn't just a knowledge problem. It's a decision-making problem — which questions you choose to attempt, and which you have the discipline to leave alone.

3. Why scores plateau even when effort doesn't drop

Here's the part that frustrates aspirants the most: the plateau usually isn't a ceiling. It's a broken feedback loop.

The most common reason cited across mock-analysis research is depressingly simple — students take a mock, glance at the score, and move to the next one without structured review, which can leave scores flat even after many mocks (Career Launcher). Toppers do the opposite: they treat mock review as more valuable than the mock itself, often spending more time analysing a paper than they spent attempting it (Quantifiers).

The deeper issue is that most "stuck" scores share a signature — a mistake pattern that resurfaces across mocks because it was noted once and never actually fixed. A wrong-answer type that shows up in three consecutive mocks isn't bad luck; it's a gap sitting in exactly the same place each time (Quantifiers). Broader research on stagnating scores backs this up, pointing to shallow analysis, repeating the same error types across mocks, and undiagnosed concept gaps as the primary drivers of a plateau — not a lack of effort or hours studied (Optima Learn). The takeaway: if your score has been flat for 3+ mocks, more mocks alone may not fix it. Finding and correcting the specific, repeating pattern will.

4. So what actually moves the needle?

Put the three pieces together and a clear picture emerges:

  • Percentile math rewards precision at the margins — a few marks matter more than aspirants assume.
  • Accuracy beats attempts — selective, confident answers consistently outscore aggressive guessing.
  • Plateaus are pattern problems — the fix isn't more practice, it's diagnosing why the same mistake keeps happening.

The common thread across all three is diagnosis. Generic advice like "solve more RCs" or "take more mocks" doesn't tell you which specific decision — a set you shouldn't have picked, an option-elimination habit, a rushed final calculation — is quietly capping your percentile.

This is the exact gap Marg was built to close. Instead of another generic study plan, Marg looks at how you specifically get questions wrong, tracks the pattern across sessions, and helps you choose focused RC, DILR and QA practice to test that working diagnosis (trymarg.com). If your mock scores have been flat for a few tests in a row, the fastest way to find out why is a proper diagnosis, not another mock. You can try Marg for free and start exploring your specific pattern; a reliable diagnosis needs evidence.