The Black-Scholes-Merton Model

The model answers one narrow question: what is a fair price today for the right to buy or sell something at a fixed level on a fixed future date? Fischer Black, Myron Scholes and Robert Merton solved it in 1973 by refusing to forecast direction at all. Their argument was that if you continuously held the right quantity of the underlying against the option, the two positions would cancel out. A package that carries no risk must earn only the risk-free rate. Work backwards from that constraint and one — and only one — premium survives.

That is why the model contains no view on whether NIFTY goes up or down. Nowhere in the formula is there an expected return. This surprises most people the first time they see it. The model prices the option by pricing the hedge, and the hedge does not care about direction. What it cares about is how far the underlying is likely to travel, because that is what determines how expensive it will be to keep the hedge balanced from now until expiry.

Practically, this makes the model a translator rather than a crystal ball. It converts a volatility assumption into a rupee premium, and it can be run backwards to convert an observed rupee premium into a volatility assumption. That second direction is the one traders use every day. When you look at a NIFTY option chain and read an implied volatility figure of 13.4 against a strike, you are looking at Black-Scholes run in reverse on the price that strike last traded at.

Five numbers go in and one premium comes out. Spot is the live price of the underlying — NIFTY at 24,500, or RELIANCE at whatever it last traded at. Strike is the level written into the contract and never changes. Time to expiry is expressed as a fraction of a year, so 14 calendar days becomes 14 divided by 365, or 0.038. The risk-free rate is the return on money with no credit risk, which in India is normally taken from short-dated government paper such as the 91-day treasury bill yield. Volatility is the annualised standard deviation of returns the model assumes will hold from now until expiry.

The two halves of the formula are easier to read than they look. The first term, spot multiplied by N(d₁), is what you would have to own of the underlying to replicate the option. That quantity is delta, which module 9 covers in detail. The second term, the discounted strike multiplied by N(d₂), is what you would have to borrow to pay for it. The premium is the difference: the value of the shares you would hold minus the money you would owe.

N(d₁) and N(d₂) are both numbers between 0 and 1 drawn from the normal distribution. N(d₂) is the model’s own probability that the option finishes in the money, measured in a risk-neutral world rather than the real one. N(d₁) is that probability adjusted for how much the option pays when it does finish in the money. The gap between them is small for a near at-the-money weekly and widens for long-dated contracts on volatile underlyings.

Four of the five inputs are facts. Spot is on your screen. Strike is printed in the contract specification. Time to expiry is on a calendar. The risk-free rate is published and, over the two-week life of a weekly option, barely matters — moving it by half a percentage point changes a near-dated NIFTY premium by a rupee or two. Two traders with the same terminal will never disagree about these four.

Volatility is different in kind, not just in degree. Nobody can observe how much NIFTY will move between now and expiry, because that has not happened yet. You can measure how much it has moved in the past — realised or historical volatility — but the model does not want the past. It wants the future. So the input is always somebody’s estimate, and the premium you see quoted is the market’s collective estimate expressed in rupees.

This is why almost all real options work is volatility work. Two traders looking at the same 24,700 call can agree on spot, strike, time and rate and still disagree violently on price, because one thinks the coming fortnight will be quiet and the other has read the event calendar. India VIX, published by the NSE, is the market’s aggregated version of this estimate for near-dated NIFTY options. Module 13 covers it in full.

There is no formula that takes a premium and returns a volatility. The equation cannot be rearranged that way. So platforms do the only other thing available: they guess, check, and correct until the model’s output matches the price the option actually traded at. The volatility that makes the two agree is called implied volatility, because it is implied by the price rather than measured from anything.

The search converges quickly because premium rises steadily with volatility — there is exactly one answer, and a handful of iterations finds it to two decimal places. That is why your terminal can display an IV figure against every strike in a NIFTY chain without any noticeable delay. It is running this loop for each row, on every tick.

The consequence is worth sitting with. Implied volatility is not an independent measurement that you can compare the price against to find a bargain. It is a restatement of the price in different units. Saying "this option is expensive because IV is 22" and saying "IV is 22 because this option is expensive" are the same sentence. IV only becomes useful when you compare it to something else — the same underlying’s own history, which is what IV rank and IV percentile do in module 14, or the volatility that actually gets realised.

The model assumes the option is European — exercisable only on expiry day and not before. On the NSE this happens to be true. Index options and single-stock options listed here are both European style, so an Indian trader is not exposed to the early-exercise error at all. It becomes a real problem only for American-style contracts on overseas exchanges, and that is the specific hole the binomial model in module 7 was built to fill.

The model assumes volatility is a single constant number that holds from today until expiry. It is not. Volatility clusters, spikes into policy meetings and results, and collapses the morning after. If a company reports earnings nine sessions from now, the coming fortnight is not one volatility — it is a quiet stretch, a violent gap, and another quiet stretch. Squeezing that into one number is a distortion, and the market corrects for it by quoting different implied volatilities at different strikes and different expiries. That correction is the volatility skew of module 15.

The model assumes returns are lognormally distributed, which is the polite way of saying it treats a crash as almost impossible. Real Indian equity history contains gap-down sessions that a lognormal model would price as events not expected within several human lifetimes. Because the model underprices the extreme, traders bid far out-of-the-money puts above their theoretical value. That persistent gap is not a mispricing to be harvested — it is the market charging correctly for a risk the formula cannot see.

Numbers make this concrete. Suppose NIFTY is at 24,500, you are looking at the 24,500 call, and there are 30 calendar days to expiry. Hold spot, strike, time and rate completely fixed and change nothing but the volatility assumption. The table below uses the standard approximation for an at-the-money option, so your terminal’s exact figure will differ by a few rupees, but the shape is right.

From a 12 volatility to a 21 volatility, the premium roughly doubles. Nothing about the index changed. Nobody made a directional call. The entire difference is the market revising how far it thinks NIFTY will travel in a month. On one lot of 75, that is the difference between paying about ₹25,000 and paying about ₹44,000 for the identical contract.

This is the mechanism behind almost every "I was right and still lost" story. A trader buys a call into an event when the volatility assumption is elevated, the event resolves, the assumption drops back, and the premium falls even though the index rose. Module 11 on vega puts a number on exactly how much premium each volatility point is worth, and module 16 walks through the collapse that follows an event.

The Greeks are not a separate system bolted onto the model. Each one is the answer to "if I nudge one input, how much does the premium move?" Nudge spot and you get delta. Nudge delta itself and you get gamma. Nudge time and you get theta. Nudge volatility and you get vega. Nudge the interest rate and you get rho. They all fall out of the same equation, which is why they are internally consistent and why they must be read together rather than one at a time.

That is also why every Greek on your screen inherits the model’s assumptions. A theta figure assumes volatility stays put between today and tomorrow. A delta figure assumes the lognormal shape holds. When the market gaps, these numbers do not merely become inaccurate — they become the wrong description of what happened, because the move was of a type the model does not admit exists.

For a working trader the practical order is this. Delta tells you how directional you currently are. Gamma tells you how fast that will change. Theta tells you what standing still costs. Vega tells you what a change of mood costs. The next six modules take them one at a time, and each of them is a derivative of the formula in this article.