Capstone: a research note on an Indian derivatives question

This is the last module of the Expert level, and it has no new market facts in it. It asks you to write one research note on an Indian derivatives question, and it publishes the rubric the note is marked against, so you can mark your own before anyone else sees it. A note is short, about 1,200 words, and built around one question, one dated dataset, one method and one honest list of limits. It is an exercise in method. It is not a research report, it carries no view on where any index is going, and nobody should act on it. What the school marks is the thinking: whether the question can be answered with the data, whether the data is dated and old enough, and whether the limits are as loud as the result.

The idea, plainly

What is the note for?

The Expert modules each teach a method: a model of the quote, a schedule for an order, a deflated Sharpe, a reading of a regulator's order. The note makes you use one of them on a question of your own choosing, the way a desk would. The test of a method is not whether you can state it. It is whether you can say what it would take to be wrong, and then look. A good note makes the reader's work easy: the question is in the first line, the answer is in the second, and the rest is the evidence for the answer and the reasons it might fail.

How do I choose a question?

Pick one that data can answer and that has two sides worth testing. A question that asks for a forecast is out: "where will Nifty close" has no answer in the files. A question that asks what happened, or how a number behaves, is in: "how far did the expiry day close sit from the highest open interest strike on the expiry days in the files" can be counted. Good starting points are the live question pages, such as does max pain work, does Nifty pin to open interest strikes, is Nifty GEX reliable, historical, implied and realised volatility and futures rollover cost. Take one, narrow it, and decide what a "no" would look like before you open the data.

What goes in the note?

Seven parts, in this order, with a budget for each. Going over the cap is a mark against clarity, not a sign of effort.

1. Question and answer, about 150 words. The question in one sentence, the answer in one sentence, and the date range it holds for.
2. Data, about 200 words. Where each file comes from, the dates it covers, its age, and every cleaning step. Market data in a note must be at least 30 days old on the day of writing, as in the rest of the school, and must come from end of day files, never a live screen.
3. Method, about 250 words. The model or test, why it fits the question, and the parameters chosen before the run.
4. Result, about 250 words and at least one chart, with the source and date under it.
5. Costs and rules, in the result or the method: charges, lot sizes, expiry days and settlement rules that applied in the period, named by regime and date.
6. Limits, about 200 words. Biases, the number of variants tried, what the sample cannot show.
7. What would change my mind, about 100 words. The result that would overturn the answer.

How is the note marked?

There are eight criteria and 100 marks. For each criterion, a strong note earns the full weight, an adequate one two thirds, a weak one a third and an absent one nothing. The pass mark proposed here is 70. The list sets out what the strong and the weak ends look like.

• Question, 10 marks. Strong: One answerable question, narrow, with a stated test for a no. Weak: Broad, a forecast or several questions at once.
• Data and dates, 15 marks. Strong: Every file named, dated, at least 30 days old, cleaning steps listed. Weak: Source vague, dates missing or a screen used.
• Method fit, 15 marks. Strong: Method suits the question, parameters fixed before the run. Weak: Method borrowed without a reason, parameters tuned after.
• Evidence and charts, 15 marks. Strong: Result follows from the data, charts carry source and date. Weak: Claims beyond the data, charts without credit.
• Costs and rules, 10 marks. Strong: Charges, lots and settlement regime named by date. Weak: Costs ignored or rules from the wrong period.
• Limits and biases, 15 marks. Strong: Variants counted, biases named, sample limits stated plainly. Weak: Limits missing or buried.
• Wording and compliance, 10 marks. Strong: Education only, no calls, levels or targets, allegations kept apart from findings. Weak: Advice or promise wording, a claim stated as fact without a source.
• Clarity, 10 marks. Strong: Within the word budget, answer first, plain words. Weak: Over budget, answer hard to find.

A strong note on a modest question scores higher than a weak note on a grand one. The marks reward being right about what you know, and being clear about what you do not.

How should a chart in the note look?

Under every chart, the source, the date range and the last date. Chart title that states the takeaway. Axes labelled in the unit. No chart without a number a reader could check. The specimen below is a standard, not a finding: it shows a series of Nifty volatility with its credit line, and says nothing about what the volatility will do.

Which files can a first note start from?

Three inputs from the files in this course, each drawn exactly as a note would draw it. They are starting points for a question, not findings, and nothing is claimed about what they show.

What does a strong Limits section contain?

Five things, each in a sentence or two. The sample: how many observations, over what dates, and what regime of lot sizes, charges and expiry days they sit in. The variants: how many versions of the test were run, including the dropped ones. The definition choices: which ones were made before the run and which after. The thing the data cannot show: a different regime, a crash that did not happen, a rule that changed after the sample ended. And the nearest alternative explanation, written as fairly as you can. A reader who finds the weakness before you named it will trust the rest of the note less. A reader who finds that you named it first will trust it more.

How do I test a claim before I write it?

Take each sentence of the Result and ask which cell of which file it comes from. If you cannot point to one, the sentence is an opinion and belongs in a different place or nowhere. Then take each number and ask whether it is a count, a rate or a rupee figure, and whether its unit is written beside it. Open interest is in units and volume is in contracts, and mixing the two is the commonest slip. Last, read the note as the person who disagrees with it, and write down the one dataset they would ask for. If you already have it, say what it shows.

Worked example: marking one section

Take the Method section of an imaginary note on whether the expiry day close sits near the highest open interest strike. No result is claimed and no market number is used. The point is the marking.

Version A says: "I tried a few ways of finding the strike until one looked right." The rubric gives it a weak mark on method fit and a weak mark on limits. Parameters were tuned after the run, and the variants were not counted.

What should I leave out?

Leave out predictions, levels and anything a reader could treat as a call. A note that ends with a view on the coming session loses marks on wording whatever else it does well. Leave out brand names of brokers and vendors, and put books and papers in the sources and not the body. Leave out screenshots of live screens, and anything that identifies a person who has not agreed to it. And leave out claims about a regulator's case that go beyond what the order says: say "the order alleges" and not "the firm did". When a source is an order, name its date and say whether it is a finding or an allegation.

What happens to the note?

For now, the rubric is published so that you can mark your own note and swap it with a peer. Whether and how RupeeCase will review notes towards the master certificate is not settled in this lesson. The certificates page will say what is offered, and it will say it plainly. A certificate from RupeeCase is a record of completing free education. It is not a NISM certificate and not a licence to advise or to publish research for others.

Try it yourself

Write the first two lines of your note now, before any data: the question and the answer you expect. Then write the result that would show you wrong. Pick one dataset from the files and list its start date, end date and age in days. If you cannot write the second line, the question is not ready.

The common mistake

The belief: A good research note proves its point. A good note says how much its data can show. It is not marked on whether the answer is yes or no, and a clean no is worth as many marks as a clean yes. The failure the rubric looks for is the confident note built on a thin sample, a tuned parameter or an unstated count of what was tried. The second failure is advice dressed as analysis: a levels table, a view on next week, a strategy named as the one to use. Both lose marks, and both are easy to avoid by writing the limits first and the answer last.

Key takeaways

• A suitable question describes a number's behaviour in dated data.
• Write the test, including what a no looks like, before the data.
• Data must be end of day files at least 30 days old.
• Count the variants and state the limits as loudly as the result.
• Allegations stay apart from findings; no calls, levels or targets.
• Eight criteria share the marks, so no single skill carries a note.

Before you trade. Derivatives trading carries a high risk of loss. SEBI's study "Profitability of Individual Traders in the Equity Derivatives Segment (FY25 to FY26)" (August 2026) found that 87.7% of individual traders in equity F&O made net losses in FY26, with aggregate net losses of Rs 91,685 crore. Read the SEBI study