How I hacked CAT in 2013
A narration of how I hacked a national aptitude test with the information I gained by observing and figuring out what it meant.
In most of the situations I encounter in my life I like to observe and speculate the reasons for the same without unnecessary prodding. Let me share an incident in which I profited massively from the way I observe and try to make sense of what I observe.
For this, I need to explain an aptitude test which is conducted in India. The name of the exam is the Common Admission Test (commonly referred to as CAT). This test is conducted every year, and attempted by roughly 200,000 graduates in India competing for a seat in the most prestigious business schools of the country, Indian Institutes of Management (IIMs), among other top business schools. The exam is very similar to GMAT in a lot of ways but has plenty of differences too. CAT has four sections 1. Quantitative Ability, 2. Data Interpretation, 3. Logical reasoning and 4. Verbal Ability. In CAT 2013, Quantitative Ability (20 questions) and Data Interpretation (10 questions) were clubbed to a single section (let’s call section 1), and Logical Reasoning (10 questions) and Verbal Ability (20 questions) were clubbed together to form the second section (let’s call section 2). While the result of GMAT is a score, the result of CAT is percentile-based rank. While one can appear for GMAT any time of the year, CAT can be attempted only once a year during a month-long test window (according to exam pattern in the years I tested. This was because of lack of sufficient centres for a computer-based test accommodating 200,000 candidates in one day).
I wanted to get into one of the top business schools in India and for that one has to get 99+ percentile in CAT to have a chance of getting in (without affirmative action). I first attempted the test in 2011, I went in casually without much preparation and scored 78.xx percentile. For the next year, I buckled down, put in a lot of time and effort, bought the books, attempted lots of mock tests and previous questions, tried different strategies for time management, and did everything experts recommended to do.
Since most of the candidates for CAT are engineers, one who scores high in section 2 (the section with 20 verbal ability questions) will score higher overall percentile. This was common knowledge and I prepared accordingly. A small overview of the exam pattern during 2011–13 can be found here (at one place in the article the total number of questions is given as 120, it’s wrong and there are slight differences in the pattern those years too). I wrote CAT in 2012 and I scored 89.94 percentile. 99+ percentile seemed far out of reach.
I put my preparations for CAT next year on the back burner. I registered for mock tests but didn’t take any. Around July it was time to register for CAT 2013. I knew that with any amount of preparation I could do till October-November, I’ll not be able to slingshot my score from 89.xx to 99.xx (to get scores above 95.xx, the effort required increases disproportionately). So I had to figure out something completely radical to score 99+ percentile.
In 2013, the exam window for CAT was from 16 October 2013 to 11 November 2013. There were two slots every day, morning and evening. Candidates can select any available slot within the exam window. Similar to GMAT, it is said that questions for CAT are taken from a question bank with difficulty level assigned to each question. But GMAT has an algorithm to serve different questions to the candidate according to the performance of the candidate. But in CAT all the candidates appearing in the same slot have the same set of questions (order of the questions are changed though). Since the questions are different for different slots of the exam, the testing authorities used a process termed as “normalization”, to rank the candidates from different slots. What exactly is normalization? No one other than the testing agencies knew (IIMs supervised testing, IIM Indore in 2013. Prometric conducted the test).
My thought process may seem obvious while reading, but I think these are obvious in hindsight. There is a lot of speculation in my thoughts and these are not presented as absolute facts.
I thought that if I could figure out what normalization was, I could have an edge over other candidates. So I collated what I knew about the exam (broadly):
1. The exam is held over 27 days (October 16 to November 11), 50 slots (I knew the actual number at the time. I’m now assuming 50, exams were not conducted on all the days. Anyway not important data).
2. All the candidates of the same slot get the same questions. The questions for different slots are taken from a question bank with matching the difficulty level across all the slots.
3. The result of the exam is a percentile-based rank where there is no distinction between the candidates appearing in different slots.
Normalization is the method used to compare and bring together the scores of candidates from different slots. Even though the knowledge of how normalization is done is not in the public domain, but it ought to have a logic and it needs to make sense to a rational mind. So I tried to speculate from the perspective of the person devising the exam.
Even though the questions of the same difficulty are chosen from the question bank, there would be differences, so taking the absolute marks and piling to a single ranking is not ideal (that’s why there is normalization). But what if the final ranking is decided with rank within one slot and the absolute difference in mark with the rest of the candidates in the same slot? The logic is, if candidate A and candidate B are the top rank holders of slot A and slot B respectively, but candidate A outperforms the second rank holder in slot A (in absolute marks) by more margin than candidate B was able to outperform the second rank holder in slot B, candidate A will have a higher rank than candidate B in the final collated rank list. It seemed reasonable and fair to create a rank list in this manner, considering the constraints. It may seem slightly complex but it is simple enough to be made into an algorithm. In all likelihood, it won’t be as simple as what I thought but I assumed it would be some variant of that with more factors taken into consideration (this was the best I could speculate). So with this assumption in place, I have to figure out a way to beat other candidates in my slot by a big margin.
At this point, I had a bit of luck. I’ll explain. What if I don’t have to figure out a way to best others in my slot, because I’ll have an advantage over other candidates in my slot? That does sound good. I stumbled across just that (with a fair bit of speculation). I got a bit lucky but I was actively looking for something similar.
I registered for CAT on the very first day of registration window, July 29 (I didn’t remember the exact date, but web search gave this date as the first day of registration in 2013). I booked my test slot for November 11 the last day of the test window. The candidates can change the slot up until a couple of weeks before the start of test window (till someday in September), according to the availability of the desired slot. When I logged in a couple of times since then, and I noticed that the slots in the last days of the test window filled up fast and this was interesting.
I got some data to make some assumptions (I seem to make a lot of them). To make the effort to go register and book the slots on the first day of registration, these candidates must be preparing for CAT quite seriously. And all of them are opting for the last slot available to them within the test window. When a seat becomes available in the last days of the test window (one can see the availability across dates and different centres through the online portal), it gets snapped up within an hour. So most of the candidates preparing for CAT quite seriously are taking the last slots (could be to get more time to prepare, could be just delaying it as long as possible, could be anything).
So who are taking the first slots? Probably, the ones who were not able to get the last slots. When I observed the availability of test slots, over the days, I saw that last available slots filled up first (if my memory serves me right, there was some indicator, colour code, percentage or absolute number, to show how many seats were remaining in a slot at a particular centre). So I assumed that the candidates attempting the test in the first slots were not anxious about getting registered and booking a slot. They were probably attempting the test just for as experience or due to parental/peer pressure with very little preparation.
Voila!
So I could be competing against candidates who haven’t prepared much for the exam or are disinterested in the exam. I changed my test dates to October 16, the very first day of the test window. The testing body either assumed that a largely similar mixture of candidates would be appearing for all the slots or they didn’t think anyone would notice this inefficiency to exploit it. But I had every intention to exploit this inefficient system and it is allowed within the rules of the system.
I had my exam slotted for October 16 (the slot was mostly empty in most centres). Then I thought what if the number of students in a slot is also a factor in the normalization algorithm. So I opted for a date which was just about to be full. I picked the morning slot on October 18. Here is a picture of my admit card (address blacked out).
I attempted the test on October 18, 2013. I was skeptical about how my assumptions would pan out but were fairly confident to score an overall score of 99+ percentile (quite contrasting).
I was in for a pleasant surprise on January 14, 2014, when the CAT results were out. I scored 99.95 percentile overall and 100 percentile in section 2 (verbal ability and logical reasoning).
Fuck yeah!
I don’t have the score sheet. The website for CAT 2013 isn’t functional now. And I had deleted all email communications from IIMs out of spite (reason can be figured out from coming paragraphs). After publishing this (this is an edit) I found an email I sent to take printouts of letters from IIM C, S (score is not mentioned in the file).
A lot of candidates were dissatisfied by the results they got (probably in comparison to their performance in mock tests). They legally moved against IIMs for clarification on the normalization procedure. Two stories in this regard can be found here and here. IIMs removed Prometric as a test partner (their contract ended in 2013). Next year for CAT 2014 the tests were held across just 2 days. And now it is conducted on a single day. I noticed that Dr. Rohit Kapoor, Faculty of IIM Indore, was convenor for both CAT 2013 and 2014. I know all these are only correlation/coincidences and would not be able to point to a causal reason.
I’m sure I’m not the only one who spotted this weakness in the system. For years faculties of coaching institutes have been scoring very good scores and they always took the initial slots to give a picture of exam pattern that particular year to their students (good thing I didn’t pick October 16 slot). I hadn’t noticed this before the results were out.
I was not awarded admission to any of the 13 IIMs (I was waitlisted for IIM Trichy). Could be because of my poor soft skills (poor but not pathetic), IIMs intentionally cut me off (it’s a possibility, I don’t want to file RTI. I was asked questions testing my verbal prowess during interviews. I didn’t perform well for those questions) or because one of the work experience certificates I produced, to fill a break, was fake (but I got waitlisted for IIM Trichy and no inquiries were made at the company which gave me the certificate). I don’t know.
But I got admission FMS, Delhi which accepts CAT score and has significantly lower fees (my score for soft skills were below average and FMS, Delhi didn’t give importance to work experience at all). But I lied to those around me that I had got admission in a couple of IIMs, but chose to pay lower fees. It’s embarrassing and quite difficult to explain why I didn’t get through any of the 13 IIMs with the score I got.
This is an instance wherein I profited from observing and making judgments about the unknown. This is the story of how I cracked CAT. This is the story of how I won the war without trying to win battles.