CAT Past Papers vs SIMCAT: The Same Score, Two Percentiles

A raw score in the band 59 to 65 returned a percentile anywhere from 66.91 to 98.46 across my CAT mocks. That is eleven attempts, six raw marks wide, and a 31.55-point spread in the number every aspirant reads first. Nothing was wrong with the scoring. The papers were different, the cohorts that sat them were different, and I sat them at different points in the season. This post lays out the whole log, 39 scored full mocks from 10 May to 22 October 2026, and the single habit it taught me: never compare two mock percentiles unless they came from the same provider, the same kind of paper and roughly the same month.

It extends two earlier posts on this site: the 18-paper version, written in August, and the log of my first 20 mocks. What is new is the second half of the season, when I started taking real past CAT papers from 2020 to 2025 as practice, which gave the data a comparison it did not have before.

What the log contains

• 40 full mocks logged, 39 with scores. One is a XAT official mock with no CAT-style section scores, so it is excluded.
• 12 real CAT papers (2020 slots 1 to 3, 2021 slots 1 to 3, 2022 slots 1 to 3, 2023 slot 1, 2024 slot 1, 2025 slot 1), each taken as a timed 120-minute practice attempt on the IMS platform, which reports a percentile for it.
• 27 coaching mocks: IMS SIMCAT and PreSimCAT papers and one TIME AIMCAT.

Every number below comes from my own Karma Yogi log. I did not take any figure from a coaching site or a forum.

The headline table: raw 59 to 65

Read it by type. The five past papers returned 91.04 to 96.62. The six coaching mocks returned 66.91 to 98.46. A 61 on SIMCAT 2 in May was the 66.91st percentile. A 61 on CAT 2022 Slot 3 in September was the 95.23rd. The raw score is identical and the percentile is 28.32 points apart.

The averages, and why they mislead on their own

The past papers look far steadier, and that is partly an artefact of timing. I took all twelve between 1 August and 18 October, when my own scores were higher than in May. That is why I added the third row. Restricted to the same window, the coaching mocks average 89.8 instead of 80.8, so more than half of the gap against the second row is simply the calendar. The remaining gap, 96.3 against 89.8, and the much narrower range on past papers, is the part I would call a real pattern, though ten coaching mocks is a small sample. I cannot say why it exists, because I do not know how IMS computes the percentile it reports for a past paper, and I have not found that stated anywhere. What I can say is that the past-paper percentiles moved far less than the coaching-mock ones.

Two more things the table shows

• Percentiles are not monotonic in score. SIMCAT 108 (6 October) scored 73 and returned 76.04. SIMCAT 15 (17 October) also scored 73 and returned 96.07. SIMCAT 114 scored 63 and returned 98.46, higher than SIMCAT 14's 90, which returned 97.17.
• A bad day reads differently in each number. SIMCAT 107 (30 July) scored 40 and returned 49.4. SIMCAT 112 (22 October) scored 39, my lowest raw score of the season, and returned 59.24. Ten points of percentile separate two nearly identical raw scores.

The rule I follow now

1. Trend raw score and sectional accuracy, not percentile. Raw score and accuracy are mine. Percentile belongs to the cohort.
2. Compare percentiles only within one provider and one paper type. A past-paper percentile and a SIMCAT percentile are different measurements, even when they carry the same label.
3. Treat a mock percentile as a weak ranking signal, not a prediction. The cohort for an IMS mock is a self-selected group of one provider's students. The cohort for CAT is roughly 2.5 lakh people. See how CAT normalizes across slots for why even the real exam does not map raw score to percentile one for one.
4. Look at the last five, not the last one. A 98.46 and a 59.24 sit nine days apart in my log (13 and 22 October). Seven of October's ten mocks returned above 95, and one returned 59.24.

If you want a rough range for a target, the CAT percentile predictor on this site shows the observed spread for each section rather than a single number, for exactly the reason in the table above.

Where this is weak

• One candidate, one season. Thirty-nine attempts by me are a log, not a study. Your numbers will differ.
• Past-paper percentiles are not official CAT percentiles. They are what IMS reported for a practice attempt. Treat none of them as a prediction of your real percentile.
• Timing is confounded with paper type. I took the past papers late. The August-onward row removes most of that, not all of it.
• I do not know IMS's method for percentiles on past papers, so the pattern above is described, not explained.
• Total marks may differ by year. I compare raw scores across papers because that is what I logged. If a paper's maximum differs, a raw score is not strictly comparable. Check each year's official paper structure before leaning on a cross-year comparison.

Logging both the raw score and the percentile for every mock, and tagging whether it was a past paper or a coaching mock, is what made this table a two-minute query instead of an afternoon in a spreadsheet. Karma Yogi stores per-section score, attempts, accuracy and percentile for each mock, so the comparison is already sitting in your log.