How Many CAT Mocks Should I Take? I Took 20 in 90 Days

Twenty-five to thirty-five across a full preparation cycle, at roughly one a week once your syllabus is covered - and the number matters far less than the gaps between them. I took 20 mocks in 90 days between 10 May and 8 August 2026, going from 41 to 91 overall. The median gap was 6 days, but seven of the nineteen gaps were two days or less, and those seven produced no usable signal, because there was no room to revise anything in between. On the research side, the reason a mock is worth so much is that it is three well-evidenced techniques at once - and the strongest of them, retrieval practice, beat elaborative study by about 50% on a delayed test in a 2011 Science experiment.

So the useful question is not "how many mocks". It is "how many mocks can I properly analyse", and that number is smaller than you want it to be.

The actual log, with the gaps

Every attempt, in order, with the interval from the previous one. Scores are out of 198. The last three are real CAT 2020 papers sat as practice rather than coaching mocks.

Nineteen gaps: eight of exactly 7 days, four of 5-6 days, and seven of two days or less. Median 6, mean 4.7.

Twenty mocks, about thirteen signals

Here is the arithmetic nobody does. A mock earns its place by what you learn from analysing it. If you sit the next one before you have analysed the last one, the second paper is not a second data point - it is a re-measurement of the same untreated state, plus fatigue.

Strip out the seven attempts that came within 48 hours of the previous one and I ran roughly 13 properly separated mocks in three months, not 20. The clustered ones were not wasted exactly - the last three, the real CAT 2020 papers, were deliberately bunched at the end - but they did not produce the thing a mock is for.

The clearest illustration is 29 and 30 July. A 96 and then a 40, with 2 marks in DILR. Two papers, 24 hours apart, and the only honest conclusion available from the pair is that I had not analysed the first before sitting the second. What that 40 actually cost me was the analysis of the 96.

Why a mock is worth more than its questions

A full mock happens to combine the three best-evidenced techniques in the learning literature simultaneously. That is worth understanding, because it explains why the analysis matters more than the count.

None of those studies used a mock exam, and I explain below why that matters. But the mechanisms are the right ones: a mock forces you to retrieve rather than recognise, to choose a method rather than execute a signposted one, and to have every topic live at once rather than one chapter at a time.

Every one of those benefits is realised when you engage with what happened, not when you record a score. Which is why the count is the wrong unit.

The budget, done honestly

Work backwards from hours rather than forwards from a target number.

• Sitting a full mock: 2 hours.
• Analysing it properly, split across two passes days apart: 2 to 3 hours. The first pass is still reacting to the score; the second consistently finds what the first missed. That is not a preference, it is what the spacing evidence and the retrieval-practice evidence jointly imply.
• Total cost per mock: 4 to 5 hours.

If you can give mocks and mock analysis 6 to 8 hours a week for the last 20 weeks before CAT, that is 120 to 160 hours, which buys 24 to 32 mocks. That is where my 25-to-35 range comes from - transparent arithmetic on a time budget, not a finding. If your weekly budget is 4 hours, the honest answer is closer to 16 mocks properly analysed, and 16 analysed beats 35 skimmed.

A cadence that follows from all this

• Before your syllabus is broadly covered: one mock a fortnight, purely for calibration. More is wasted because most of the paper is testing what you have not studied.
• Once it is covered, through to a month before the exam: one a week. Analyse it across two passes before the next one. Eight of my nineteen gaps were exactly this, and they are the ones that moved anything.
• The final month: up to two a week if you can still analyse both. If you cannot, take one.
• Never two inside 48 hours, with one exception: deliberately simulating back-to-back pressure, or working through past papers as a block at the end, which is what my final cluster was.
• Stop about 3-4 days before the exam. Nothing you learn from a mock in that window can be converted into a change.

Does more mocks actually mean a higher score?

In my log, the improvement is real and the attribution is not clean. First five mocks against last five:

Two caveats I have to attach to my own numbers. The last five include three real CAT 2020 papers sat at home rather than proctored coaching mocks, which flatters them. And DILR moved 2.2 marks over the whole quarter while the overall number climbed 19. Twenty mocks did not fix my weakest section. If a number of mocks alone were the mechanism, it should have.

That is the strongest argument I have against chasing a count: the section that stalled is the one I never changed my analysis of.

Where this evidence does not reach

• There is no dose-response study of mock exams. Not for CAT, not for any competitive entrance exam that I could find and read. Anyone telling you "take 40 mocks" or "take 15" is giving you an opinion, and so am I. The difference is that mine comes with the time arithmetic attached.
• The three studies in the table above did not study mocks. They studied American schoolchildren and undergraduates learning word lists, science texts and school algebra, in laboratories and classrooms. The mechanisms transfer as descriptions of what a mock does. The effect sizes do not transfer as predictions about your score.
• My sample is one candidate, 20 papers, 13 weeks. There is no control group, no comparison cadence, and no way to separate mock count from three months of ordinary study running alongside it.
• I have not sat CAT. Every percentile above is a mock percentile against a coaching cohort, and mock percentiles do not convert - in the same log a 54 and a 96 both returned the 90th percentile. See why that number does not travel.
• The 4-to-5 hours per mock is my estimate, not a measurement. If you analyse faster than that and still get something out of it, the arithmetic changes and so does the number.
• Nothing here accounts for burnout, which is a real constraint on cadence and one no study I read even measured.

Sources

• Mock log, 20 attempts, 10 May to 8 August 2026 - first-party, published in full above and section by section in the 20-mock post.
• Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science, 331(6018), 772-775.
• Rohrer, D., Dedrick, R. F., Hartwig, M. K., & Cheung, C.-N. (2019). A randomized controlled trial of interleaved mathematics practice. Journal of Educational Psychology. Read via ERIC.
• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks. Psychological Bulletin, 132(3). Read via eScholarship.

The gap column in that first table is the only reason any of this was visible to me, and it exists because Karma Yogi stores a mock with its date and its analysis sessions attached rather than as a score in a spreadsheet. Count your analysis passes, not your mocks.