My 20 CAT Mocks: 41 to 91, With Every Crash Left In

Almost every "how I improved my CAT score" post shows a clean upward line. Mine is not clean. Over three months and 20 mocks my overall score went from 41 to 91, but it did that by way of a 44, a 47, and a 40 - and the 40 came two days after a 96.

This is the actual log, pulled from my own tracker, with nothing smoothed. I am publishing the crashes because they are the part that taught me something, and because a monotonic graph would tell you nothing you could use.

The Full Log

Twenty full mocks, May 10 to August 8. Scores are out of 198. The last three are actual CAT 2020 papers rather than coaching mocks, which is why they read differently.

The Headline Numbers

Comparing my first five mocks against my last five:

• Overall: 61.8 average, rising to 81.0
• Percentile: 73.4 average, rising to 94.2
• QUANT: 26.8, rising to 35.6
• VARC: 16.6, rising to 24.8
• DILR: 18.4, rising to 20.6

That last line is the one worth staring at. Over three months of deliberate work, my DILR barely moved. Two marks. If you only ever looked at the overall score climbing from the 40s to the 90s, you would conclude everything was working. Section-level numbers say something much less flattering: two sections improved substantially and one did not budge.

Crash One: 40 Marks, Two Days After 96

On 29 July I scored 96, my best mock to that point, driven by a 50 in QUANT. On 30 July I scored 40, with 2 marks in DILR.

Nothing about my knowledge changed in 48 hours. What changed was which sets I chose. My own analysis note from that mock, written the same week, reads: "Could have definitely gotten 30 marks." The sets were doable. I picked wrong, committed to the wrong one early, and had no time left when I realised.

This is the single most important thing the log taught me, and it only became visible because the scores sat next to each other in a table. A 96 followed by a 40 is not variance in ability. It is variance in triage - and triage is a trainable skill that most people, including me, spend far less time on than content revision.

Crash Two: 5 Marks in VARC

On 28 June my VARC score was 5, at the 15th percentile. My note on the question that broke it: the passage discussed only Descartes, the answer option said "Descartes and other philosophers", and I took it. A distortion I would spot instantly on a calm day.

VARC was my most volatile section all quarter: 3, then 36, then 5, then 39, then 9. Look at that sequence. There is no learning curve in it at all - it is close to noise. Averaging it into an overall score made it invisible; my overall scores across those same five mocks were 41, 91, 47, 80 and 60, which reads as "inconsistent" rather than "your reading accuracy is the thing that is inconsistent".

What the Aggregate Score Hid

Three months of data, and the honest summary is not "I improved 50 marks". It is:

• QUANT was the engine. It rose steadily and reliably - 21, 25, 27, 32, 35 - because it responds to exactly the thing I was doing most of, which is topic revision and problem sets. This is the section where effort converts to marks most predictably.
• VARC was noise, not a trend. Its swings had more to do with whether I read carefully on the day than with anything I studied that week.
• DILR was a plateau, and I did not notice for two months. It looked fine because the overall number was rising. It was not fine.

If I had been tracking only the overall score, I would have spent the whole quarter feeling like everything was working. I would have kept doing more QUANT, because QUANT kept rewarding me, and my weakest section would have stayed weakest right into the actual exam.

The Three Things That Actually Moved the Number

Stripping out everything that did not survive contact with the data:

1. Analysing every mock more than once. I logged multiple analysis passes per mock, days apart. The second pass consistently found things the first missed, because the first pass is still emotionally attached to the score. The error-type classification framework is what made a second pass worth doing rather than just re-reading solutions.

2. Writing down "could have gotten X" for every section. This became my most useful single number - not what I scored, but what the paper was worth to me with better selection. The gap between the two is pure triage loss, and it was frequently 15-20 marks in DILR. A score tells you where you are. That gap tells you what is actually available.

3. Reading section scores, not overall scores. The overall number is what coaching platforms put in front of you and it is the least actionable figure available. Every real decision I made came from a section-level or topic-level number.

What I Would Do Differently

I would treat set selection in DILR as a distinct skill with its own practice, from week one, rather than assuming it would improve as a side effect of solving more sets. Two marks of improvement in three months says clearly that it did not.

I would also stop reading a single mock as a signal. My 96 and my 40 are two days apart. Neither is my level. The five-mock rolling average - which is the only place my improvement actually looks real - is.

Why the Log Exists at All

None of the above was visible in a spreadsheet, which is where I started. It became visible when the mocks, the section scores and every analysis session sat in the same place, so that "my DILR has not moved since May" was something I could see rather than something I had to remember.

That is the entire reason I built Karma Yogi: log a mock with its section scores, log each analysis pass against it, and let the section trends surface on their own. It is free, and it works the same way for JEE, NEET, GRE and GMAT sections as it does for CAT's.

If you take one thing from this post, take the least motivational sentence in it: my overall score nearly doubled while my weakest section stood still, and the overall score is what hid it. Track sections.