Weightage tables state what an examination asks about. They do not state what an answer costs to produce, and that second figure is the one that should decide where preparation time goes.
Nearly every published guide to competitive preparation ranks topics by share of marks, and the ranking is not wrong. A reading of 8,267 authentic JEE Main questions, drawn from papers between 2002 and 2025, confirms that the distribution is uneven and stable enough to plan around. What such rankings omit is the other half of the calculation. A share of marks is a numerator. Return on study time is a ratio, and the denominator, the hours a topic costs before it yields a correct answer under examination conditions, appears in almost no published table.
The omission matters because the spread in that denominator is wider than the spread in weight. The heaviest topics sit within a couple of percentage points of one another; the hours they demand do not. Two topics of equal weight can differ greatly in the time required to reach reliable performance, and a ranking by weight alone treats them as interchangeable.
Across the corpus, 117 topics recur often enough for their frequency to describe a pattern rather than the accident of one year's paper. Physics accounts for 48 of them, Chemistry 41 and Mathematics 28. Mathematics concentrates its countable mass on fewer stable topics, so depth within any one of them buys more of the paper. Physics spreads across nearly twice as many, rewarding command of a few recurring principles over topic-by-topic coverage.
The second finding is a ceiling. Matrices and Determinants accounts for 8.8 per cent of JEE Main Mathematics since 2016, 275 questions across the papers in the archive. Coordination Compounds accounts for 8.7 per cent of recent Chemistry, 171 questions. Kinematics in one and two dimensions accounts for 6.6 per cent of recent Physics, 180 questions. That is the heavy end, and none of it approaches a tenth of its subject. A candidate who mastered the single heaviest topic in each subject and nothing else would still meet a paper drawing more than nine questions in ten from elsewhere. Weightage is a tilt in the odds, not a shortcut past the syllabus.
Those three topics are close in weight and nothing alike in cost. Matrices and Determinants is largely procedural. Its rule set is small and closed, and a candidate who practises it moves quickly from nothing to competence before flattening out. Its questions are answered by executing steps, which makes the hours predictable.
Coordination Compounds carries almost the same weight and behaves in the opposite way. Much of it is recall: names, structures, and the associations between them. Acquisition is fast, because material of that kind enters memory in a fraction of the time a mathematical technique takes to become fluent. Retention is the expensive part. What is learnt in one sitting will not be there many months later without deliberate return.
Kinematics is the lightest of the three by count and may be the best-paying by the hour, for a reason no count can show. The quantities it defines are used wherever motion is described, so hours booked against it are partly recovered under other headings. Frequency data attributes a question to one topic. It cannot attribute the understanding that answered it, so topics whose ideas are reused are undervalued by any table built on counts.
A topic learnt in one month must still be available in another. Ebbinghaus established more than a century ago that forgetting is steep at first and then slow. Cepeda and colleagues' 2006 review of distributed practice found that spacing study across sessions outperforms the same total time massed together, and that the advantage grows with the retention interval. Roediger and Karpicke's 2006 experiments on retrieval practice found that testing oneself on material beats rereading it, with the gap widening at longer delays.
The consequence for topic selection is direct. The true cost of a topic is acquisition plus a recurring maintenance charge, and that charge is not uniform. Procedural skill degrades slowly and is largely restored by a handful of problems. Dense factual recall degrades faster and needs scheduled return. A topic that is cheap to learn and expensive to keep can cost more across a preparation cycle than one that was slow to acquire and then stayed acquired.
Marking schemes change the shape of the return curve. The accuracy at which attempting becomes worthwhile is the penalty divided by the sum of reward and penalty. JEE Main and NEET break even at 20 per cent. UPSC Prelims breaks even at exactly 25 per cent, which is the return on a blind pick from four options, making a random guess there neutral in expectation.
Under the JEE scheme a blind pick is already mildly positive, and the first few hours in an unfamiliar topic, enough to recognise the form of a question and rule out one implausible option, lift a candidate well clear of break-even. Early breadth pays under that rule in a way it does not under the UPSC one, where nothing is earned until the knowledge is real. The answer to which topics reward study time most is therefore not fixed. Early, the reward lies in a shallow pass across many of the 117. Later, it lies in depth on the heavy topics and on those whose ideas are reused.
Expected marks per hour is a topic's share of the paper, multiplied by the probability of converting its questions, divided by the hours of acquisition and maintenance it demands. Only the first term is published anywhere. The rest are personal, and can be estimated only by attempting questions and recording what happened. That is unwelcome news for anyone hoping a table will settle the matter, and it is why the heaviest topic is often not the first one a rational candidate should open.
The full set, with progress tracking and five agent perspectives per question, is in the JupiteX app — browse the exam catalogue or browse the Learn library.