Do mock percentiles inflate your actual CAT percentile?

Nobody has published paired data that would settle it, and the widely-quoted "3 to 8 points" range has no primary source I could find. What this log shows is that no stable offset exists even between two mocks: a 54 and a 96 both returned the 90th percentile, and six pairs invert outright.

Why did I score higher but get a lower percentile?

Because the two scores came from different papers with different cohorts. A percentile describes where you sat among the people who attempted that specific paper, so an easier paper or a stronger cohort pushes the same raw score down the distribution. In this log it happened six times in eighteen attempts.

Is a 90 percentile in a SIMCAT the same as 90 percentile in CAT?

No, and they are not comparable in either direction. A coaching mock cohort is self-selected and much smaller than the roughly 2.5 lakh who sit CAT, and is scored on a different paper. Treat a mock percentile as a position check within that mock series, not as a forecast.

Should I track mock percentile or raw score?

Raw score, per section. It is the figure you control and the only one comparable across papers from the same provider. Percentile moves with paper difficulty and cohort, so a flat percentile trend can hide a genuinely improving score - and a rising one can hide a stalling one.

Are IMS and TIME percentiles comparable?

No. They are computed against completely different groups of test-takers on completely different papers, so averaging an IMS percentile with a TIME one produces a number that describes nobody. Keep them apart and read each series on its own.

How many mocks do I need before the trend means anything?

On this evidence, more than you would expect from percentile alone, because percentile is noisy across papers. Raw sectional scores settle down faster. Fifteen coaching mocks over three months was enough to see a clear raw-score trend and not enough to see any percentile pattern at all.

Why do coaching institutes publish percentiles if they are unreliable?

Because a percentile is the shape of the answer candidates want, and it does serve a real purpose within one mock series - it tells you roughly where you sit among people preparing for the same exam. The problem is the conversion, not the number: it is being read as a CAT prediction rather than as a position within that cohort.

Can I use past CAT papers to predict my percentile?

Better than a coaching mock, with a caveat. A past paper scored against that year real cohort is at least measuring you against people who actually sat CAT. But attempting it at home, untimed and unproctored, is a different event from the exam, so the score is flattered even where the cohort is not.

What is a good percentile in CAT mocks?

A more useful question is whether your raw score is rising. In this log the overall raw score ran from 41 to 96 while the percentile wandered between 49 and 97 with no reliable relationship to it. If the score is climbing and accuracy is holding, the preparation is working whatever the percentile says that week.

Where does the data in this post come from?

Eighteen attempts logged personally in Karma Yogi between 10 May and 8 August 2026 - fifteen IMS SIMCAT, PreSimCAT and TIME AIMCAT mocks, plus three CAT 2020 papers sat as practice. Every row is published above with the paper it came from.

The Journal

CAT Mocks

10 September 2026
8 min read

Anish Guruvelli

Karma Yogi

Mock Percentile vs Actual CAT Percentile: My 18 Papers

A 54 and a 96 both returned the 90th percentile in my own mock log. Six pairs invert outright. Which is why "mocks inflate your percentile by 3-8 points" cannot be true in either direction - and what to track instead.

Across 18 logged attempts between 10 May and 8 August 2026, a raw score of 54 returned the 90th percentile and a raw score of 96 also returned the 90th percentile. Same candidate, same three-month window, a 42-mark gap, and the identical percentile. Six other pairs in the same log invert outright - a higher score coming back with a lower percentile than a lower one.

That is the whole answer to "how much do mocks inflate my percentile." The question assumes there is a stable offset to correct for. In this data there isn't one, in either direction, and the number people quote for it turns out to have no source at all.

The claim everyone repeats, and where it comes from

Search this and you will be told, confidently and repeatedly, that mock percentiles run 3 to 8 points above your eventual CAT percentile. I could not find a primary source for that range. It appears in aggregated summaries without a sample size, a cohort description, or a year.

The most credible version of the argument is 2IIM's "The Truth About Mock CATs and Percentiles", from July 2020. Rajesh Balasubramanian - who has sat CAT repeatedly and scored 100 percentile more than once - argues that coaching percentiles are guesstimates from look-up tables, built on a couple of thousand serious proctored candidates and presented as though the cohort were ten times that.

I think he is right. The post is about 700 words long and contains no table and no data. It has been the best available answer for six years, and it is an assertion. What follows is the same argument with the receipts.

The log

Eighteen attempts with a recorded overall percentile. Fifteen are coaching mocks - IMS SIMCAT and PreSimCAT, and one TIME AIMCAT. Three are real CAT 2020 papers, sat as practice, scored against that year's actual cohort.

Date	Paper	Overall score	Percentile
10 May	PreSimCAT 03	41	49
16 May	SIMCAT 2	61	67
23 May	SIMCAT 1	74	88
25 May	AIMCAT SA2701	68	83
30 May	SIMCAT 101	65	80
6 Jun	SIMCAT 3	44	70
13 Jun	SIMCAT 102	63	78
20 Jun	SIMCAT 4	91	96
27 Jun	SIMCAT 103	54	90
28 Jun	SIMCAT 104	47	67
5 Jul	SIMCAT 5	49	72
11 Jul	SIMCAT 105	80	85
18 Jul	SIMCAT 6	65	82
27 Jul	SIMCAT 7	60	83
29 Jul	PreSimCAT 02	96	90
1 Aug	CAT 2020 Slot 1	94	97
3 Aug	CAT 2020 Slot 2	84	94
8 Aug	CAT 2020 Slot 3	91	96

Two mocks in the same period returned no percentile at all, so they are not in this table. The full section-by-section version, including those, is in the 20-mock log.

The six inversions

An inversion is a pair where the higher score came back with the lower percentile. If a mock percentile were a fixed transformation of a mock score, there could not be any.

Lower score	Its percentile	Higher score	Its percentile
74 (SIMCAT 1)	88	80 (SIMCAT 105)	85
44 (SIMCAT 3)	70	47 (SIMCAT 104)	67
91 (SIMCAT 4)	96	96 (PreSimCAT 02)	90
54 (SIMCAT 103)	90	60 (SIMCAT 7)	83
54 (SIMCAT 103)	90	65 (SIMCAT 6)	82
54 (SIMCAT 103)	90	80 (SIMCAT 105)	85

SIMCAT 103 appears three times because a 54 on it was worth more than a 60, a 65 and an 80 on other papers. Nothing about the candidate changed. The paper's difficulty and the group that sat it did.

Why a percentile cannot travel between papers

A percentile is not a property of your performance. It is a statement about the people you were measured against, on that day, on that paper. Change the group and the same performance produces a different number - which is exactly what the table above shows happening within a single candidate's log over twelve weeks.

Three things vary between any two mocks:

Paper difficulty. A hard paper compresses scores, so a modest raw score climbs the distribution. SIMCAT 103's 54-for-90 is that.
Cohort composition. A weekend national mock and a Tuesday-evening practice paper draw different people, and a small cohort is noisier at the top than a large one - which is precisely where you care about the number.
When it was scored. Percentiles computed early, before most of the cohort has attempted, move as more results land.

What about the real CAT papers?

The three attempts at the end are CAT 2020 Slots 1, 2 and 3, sat as practice and scored against that year's real cohort. The contrast with the coaching mocks is stark: those three average 89.7 raw for the 95.7th percentile, while the fifteen coaching mocks average 63.9 raw for the 78.7th.

Do not read that as evidence coaching mocks deflate your percentile. I want to be explicit, because it is the obvious misreading and it would be a better story than the truth. Three problems with it:

They came last. August attempts against May and June ones. Twelve weeks of preparation sits inside that gap, and the raw scores climbed accordingly.
They were not sat under mock conditions. A past paper attempted at home, untimed against a live cohort, is not the same event as a proctored national mock.
n is three. That is an anecdote with a percentile attached.

What those three rows genuinely support is narrower and more useful: a percentile scored against CAT's real 2020 cohort behaved differently from one scored against a coaching cohort. Which is the same finding as the rest of the post - the number is about the group, not about you.

The honest limits of this post

This is one candidate, eighteen papers, twelve weeks, and two coaching providers. It is enough to demonstrate that a stable mock-to-CAT offset does not exist in this data. It is nowhere near enough to produce the correct conversion factor, and no such factor is offered here.

I also have no actual CAT percentile to compare against - I have not sat CAT yet. Anyone claiming a precise inflation figure needs paired data: the same candidates' mock percentiles and their real CAT percentiles, at scale. The coaching institutes hold that data. None of them publish it.

What to track instead

If percentile cannot travel between papers, the practical consequence is that your percentile trend across mocks is close to meaningless as a progress signal, and most people are using it as their primary one.

Track raw score, per section. It is the thing you control and the only figure comparable across papers from the same provider.
Track accuracy separately from attempts. A rising attempt count with falling accuracy lowers your score while feeling like progress.
Keep providers apart. An IMS percentile and a TIME percentile are two different cohorts. Averaging them produces a number describing nobody.
Use percentile within a single mock series only, as a rough position check, never as a prediction.

That is what Karma Yogi records a mock as: sections with their own scores, attempts, accuracy and percentile, kept apart by provider, so the trend you read is the one you can act on. The percentile predictor built on this same data reports a band rather than a number, for exactly the reasons above.

Sources

Score and percentile log, 10 May to 8 August 2026 - first-party, published in full above and in the 20-mock log.
2IIM, The Truth About Mock CATs and Percentiles, July 2020.
CAT paper structure - 68 questions, 204 marks, 120 minutes - unchanged since CAT 2024.

End of essay

- Anish Guruvelli

Common questions

Do mock percentiles inflate your actual CAT percentile?: Nobody has published paired data that would settle it, and the widely-quoted "3 to 8 points" range has no primary source I could find. What this log shows is that no stable offset exists even between two mocks: a 54 and a 96 both returned the 90th percentile, and six pairs invert outright.
Why did I score higher but get a lower percentile?: Because the two scores came from different papers with different cohorts. A percentile describes where you sat among the people who attempted that specific paper, so an easier paper or a stronger cohort pushes the same raw score down the distribution. In this log it happened six times in eighteen attempts.
Is a 90 percentile in a SIMCAT the same as 90 percentile in CAT?: No, and they are not comparable in either direction. A coaching mock cohort is self-selected and much smaller than the roughly 2.5 lakh who sit CAT, and is scored on a different paper. Treat a mock percentile as a position check within that mock series, not as a forecast.
Should I track mock percentile or raw score?: Raw score, per section. It is the figure you control and the only one comparable across papers from the same provider. Percentile moves with paper difficulty and cohort, so a flat percentile trend can hide a genuinely improving score - and a rising one can hide a stalling one.
Are IMS and TIME percentiles comparable?: No. They are computed against completely different groups of test-takers on completely different papers, so averaging an IMS percentile with a TIME one produces a number that describes nobody. Keep them apart and read each series on its own.
How many mocks do I need before the trend means anything?: On this evidence, more than you would expect from percentile alone, because percentile is noisy across papers. Raw sectional scores settle down faster. Fifteen coaching mocks over three months was enough to see a clear raw-score trend and not enough to see any percentile pattern at all.
Why do coaching institutes publish percentiles if they are unreliable?: Because a percentile is the shape of the answer candidates want, and it does serve a real purpose within one mock series - it tells you roughly where you sit among people preparing for the same exam. The problem is the conversion, not the number: it is being read as a CAT prediction rather than as a position within that cohort.
Can I use past CAT papers to predict my percentile?: Better than a coaching mock, with a caveat. A past paper scored against that year real cohort is at least measuring you against people who actually sat CAT. But attempting it at home, untimed and unproctored, is a different event from the exam, so the score is flattered even where the cohort is not.
What is a good percentile in CAT mocks?: A more useful question is whether your raw score is rising. In this log the overall raw score ran from 41 to 96 while the percentile wandered between 49 and 97 with no reliable relationship to it. If the score is climbing and accuracy is holding, the preparation is working whatever the percentile says that week.
Where does the data in this post come from?: Eighteen attempts logged personally in Karma Yogi between 10 May and 8 August 2026 - fifteen IMS SIMCAT, PreSimCAT and TIME AIMCAT mocks, plus three CAT 2020 papers sat as practice. Every row is published above with the paper it came from.

Keep Reading

Spaced Repetition for CAT Revision: What 317 Experiments Say

The meta-analysis behind spaced repetition covers 317 experiments and 14,811 people, and it says less about revising a CAT mock than every study-tips post assumes. Here is what it does say, what it does not, and the schedule I actually run.

Read

Study Techniques· 11 min read

Active Recall vs Rereading: What 200 Students Showed

In a 2011 Science experiment, 101 of 120 students learned more from retrieval practice than from an elaborative study method - and only 30 of them expected to. That gap between what works and what feels like it works is the whole reason people reread.

Read

CAT Quant· 12 min read

Interleaving Quant Practice: 61% vs 38% in One RCT

A pre-registered randomised trial across 54 maths classes found mixed practice beat topic-wise practice 61% to 38% on a delayed test. My own log shows why that matters for CAT: the section I drilled topic-wise improved nine marks, and the section that needs selection improved two.

Read

Exam Anxiety· 11 min read

Exam Anxiety and Mock Scores: I Scored 96, Then 40

Two days apart, my best mock and my worst. The anxiety research explains less of that gap than you would expect - about 12% of variance at best - and the arithmetic of two outliers explains a lot more. What to check before you diagnose yourself.

Read