From Itô to Black-Scholes: How Randomness Became a Pricing Engine

Chicago, 1973.

A new exchange opens its doors with a strange product at the center of the room: standardized listed options. Traders understand stocks. They understand buying low and selling high. But options are different. Their value bends. They decay with time. They react to volatility. They can be used to speculate, hedge, insure, or ruin yourself with leverage.

The market needs prices.

Almost at the same moment, a formula arrives.

Fischer Black and Myron Scholes publish The Pricing of Options and Corporate Liabilities. Robert Merton publishes his own treatment of rational option pricing the same year and helps give the framework the continuous-time depth it deserves. Decades later, the Nobel committee will describe the result as a new method for valuing derivatives, developed by Merton and Scholes in collaboration with the late Fischer Black, and will emphasize how it made risk management more efficient across financial markets.

That is the finance-history version.

But the mathematical story does not begin in Chicago.

It begins in 1900 with Louis Bachelier staring at the Paris stock market and seeing random motion where other people saw only speculation. It passes through Norbert Wiener, who made Brownian motion a rigorous mathematical object. Then it turns through Kiyosi Itô, who did something quieter and stranger: he built a calculus for paths too rough to differentiate.

Black-Scholes did not appear out of nowhere. It was the moment stochastic calculus became a pricing engine.

Quant finance begins when random price paths become objects you can transform, hedge, simulate, and price.

This post is the bridge from the stochastic mathematics series into quantitative finance. If the earlier articles asked "what is randomness?", this one asks a more financial question:

How did randomness become something markets could put a price on?

This is educational research, not investment advice. We are studying the machinery, not recommending trades.

If you are joining this thread here, the earlier story is worth reading in order:

Monte Carlo Methods: Discovery & Evolution follows randomness from physical mystery to computational method.
Stochastic Differential Equations: A Practitioner's Guide turns noisy motion into equations and simulations.
Bachelier's Revenge shows how finance entered the randomness story before physics finished explaining Brownian motion.
Norbert Wiener: Cybernetics and Random Processes makes continuous random paths mathematically real.
The Quiet Genius Who Made Randomness Calculable explains why Itô calculus is the hinge between random paths and finance.

The Road to 1973 Was Longer Than It Looked

The Black-Scholes formula is often introduced as if finance suddenly became mathematical in the early 1970s.

That is too clean.

The real story is messier and more interesting. By 1973, several pieces had been waiting for decades.

The first piece came from Bachelier. In his 1900 thesis, Théorie de la Spéculation, he treated market prices as random. That was a strange move. The respectable mathematical problems of the time were in mechanics, geometry, celestial motion, heat, and physics. Bachelier took the stock market seriously as a mathematical object.

He was early. Too early.

He described a Brownian-style model for speculation before Brownian motion had been fully explained by physics or made rigorous by probability theory. His work did not become the immediate foundation of a new financial industry. It mostly disappeared.

Then physics caught up. Einstein connected Brownian motion to molecular collisions. Wiener later built Brownian motion as a probability measure on continuous paths. The random curve was no longer only a visual mystery under a microscope or a metaphor for market noise. It was a legitimate mathematical object.

But there was still a problem.

Brownian paths are continuous, but almost surely nowhere differentiable. They move without jumps, yet they never settle into a smooth local slope. Ordinary calculus expects curves that behave, at least locally, like lines. Brownian motion refuses.

So by the time Wiener had made Brownian motion rigorous, mathematics had a beautiful object that classical calculus could not handle.

Itô supplied the missing tool.

Itô's Small Correction That Changed Everything

In 1944, Kiyosi Itô published a paper called "Stochastic Integral" in the Proceedings of the Imperial Academy. It was short, technical, and not written for Wall Street.

In 1951, he published "On a Formula Concerning Stochastic Differentials" in the Nagoya Mathematical Journal. Cambridge Core's page for the paper describes its aim plainly: to show the formula concerning stochastic differentials in detail and in a more general form.

That formula became Itô's lemma.

The ordinary chain rule says that if a variable changes, a smooth function of that variable changes according to its first derivative. That works when the input path is smooth enough.

Brownian motion is not smooth enough.

For a Brownian increment over a tiny interval:

d W \sim d t

Now square it:

(d W)^{2} \sim d t

That is the technical hinge.

In ordinary calculus, second-order terms vanish too quickly to matter. With Brownian motion, the square of the random wiggle accumulates at the same scale as time. Curvature no longer disappears. It contributes.

For an Itô process,

d X_{t} = μ (X_{t}, t) d t + σ (X_{t}, t) d W_{t}

Itô's lemma says a smooth function $f (X_{t}, t)$ evolves as:

df = (\frac{\partial f}{\partial t} + μ (X_{t}, t) \frac{\partial f}{\partial x} + \frac{1}{2} σ^{2} (X_{t}, t) \frac{\partial ^{2} f}{\partial x ^{2}}) d t + σ (X_{t}, t) \frac{\partial f}{\partial x} d W_{t}

The extra term is the signature:

\frac{1}{2} σ^{2} (X_{t}, t) \frac{\partial ^{2} f}{\partial x ^{2}}

Randomness and curvature create drift.

That sentence sounds abstract until you apply it to money.

The Stock Price Model That Became the Classroom Doorway

The simplest continuous-time stock model in the Black-Scholes world is geometric Brownian motion:

d S_{t} = μ S_{t} d t + σ S_{t} d W_{t}

Here:

$S_{t}$ is the stock price;
$μ$ is the expected return;
$σ$ is volatility;
$W_{t}$ is Brownian motion.

This model is not reality. Real markets jump, cluster volatility, gap overnight, suffer liquidity shocks, and laugh at constant-volatility assumptions.

But geometric Brownian motion is the right classroom doorway because it captures three ideas at once:

price changes scale with the current price;
volatility is proportional to the current price;
the price stays positive.

Now ask the question that reveals Itô's correction:

How does the log price move?

Let:

Y_{t} = lo g S_{t}

If we used ordinary calculus carelessly, we might write:

d lo g S_{t} \approx \frac{1}{S _{t}} d S_{t}

That would give:

d lo g S_{t} = μ d t + σ d W_{t}

But this misses the curvature term. Applying Itô's lemma correctly gives:

d lo g S_{t} = (μ - \frac{1}{2} σ^{2}) d t + σ d W_{t}

There it is:

- \frac{1}{2} σ^{2}

That little term is volatility drag. The expected arithmetic return and the expected log growth rate are not the same. Randomness changes the growth accounting.

The first finance lesson of Itô calculus is brutal: volatility is not just risk around a trend. It changes the trend you experience.

This is the point where stochastic calculus stops being an elegant mathematical invention and starts looking like finance infrastructure.

If Itô's lemma can tell us how log prices evolve, it can also tell us how any smooth function of price and time evolves.

An option is exactly that.

An Option Is a Function of a Random Path

A European call option gives you the right, but not the obligation, to buy a stock at a fixed strike price $K$ at maturity $T$ .

At maturity, the payoff is simple:

max (S_{T} - K, 0)

Before maturity, the value is not simple. It depends on:

the current stock price;
the strike;
the time remaining;
the risk-free rate;
volatility;
dividends and carry assumptions in richer versions;
the probability distribution of future prices.

In the basic Black-Scholes setup, we write the option value as:

V = V (S, t)

The option value is a surface. It changes with stock price and time. Because $S_{t}$ is random, $V (S_{t}, t)$ is random too.

Now Itô's lemma enters again:

d V = (\frac{\partial V}{\partial t} + μ S \frac{\partial V}{\partial S} + \frac{1}{2} σ^{2} S^{2} \frac{\partial ^{2} V}{\partial S ^{2}}) d t + σ S \frac{\partial V}{\partial S} d W

This equation is the moment the option stops being a mysterious contract and becomes a stochastic process with two parts:

a deterministic drift part multiplied by ( $d t$ );
a random shock part multiplied by $d W$ .

The random part is:

σ S \frac{\partial V}{\partial S} d W

That $\frac{\partial V}{\partial S}$ term is delta: the option's sensitivity to the stock price.

If the stock jumps a tiny random amount, the option moves by roughly delta times that move. The key insight is that the stock itself has the same Brownian shock.

That means the randomness can be canceled.

Locally. Continuously. Under ideal assumptions.

That is the hedging insight.

The Portfolio That Makes the Randomness Disappear

Imagine holding one option and shorting $Δ$ shares of the stock.

The portfolio is:

Π = V - Δ S

The change in the portfolio is:

d Π = d V - Δ d S

The stock evolves as:

d S = μ S d t + σ S d W

The option evolves as:

d V = (\frac{\partial V}{\partial t} + μ S \frac{\partial V}{\partial S} + \frac{1}{2} σ^{2} S^{2} \frac{\partial ^{2} V}{\partial S ^{2}}) d t + σ S \frac{\partial V}{\partial S} d W

Choose:

Δ = \frac{\partial V}{\partial S}

Now compare the random terms.

The option contributes:

σ S \frac{\partial V}{\partial S} d W

The stock hedge removes:

\frac{\partial V}{\partial S} σ S d W

They cancel.

That cancellation is the heart of Black-Scholes-Merton. Not the final formula. Not the normal distribution. Not the famous $d_{1}$ and $d_{2}$ . The heart is the construction of a locally risk-free hedged portfolio from one risky stock and one derivative written on that stock.

If the portfolio has no Brownian risk over the instant, then under the model's assumptions it must earn the risk-free rate. If it earned more, arbitrageurs would buy it. If it earned less, they would sell it. Either way, the mismatch would be traded away.

The Nobel Prize press release explains the economic idea with the same structure: combine options and shares so stock-price risk is eliminated, then adjust the composition as sensitivities change. That dynamic hedge is why the option can be valued without choosing a subjective risk premium.

That last phrase matters.

Before Black-Scholes-Merton, option valuation looked like it required an opinion about expected stock return and investor risk appetite. But in the hedged portfolio argument, $μ$ drops out. The expected return of the stock is not the input that prices the option.

The risk-free rate takes its place.

This is the bridge from stochastic calculus to risk-neutral valuation.

The Pricing Equation

Carrying through the no-arbitrage hedge argument gives the Black-Scholes PDE:

\frac{\partial V}{\partial t} + \frac{1}{2} σ^{2} S^{2} \frac{\partial ^{2} V}{\partial S ^{2}} + r S \frac{\partial V}{\partial S} - r V = 0

This equation is doing something remarkable.

We started with a random stock path:

d S_{t} = μ S_{t} d t + σ S_{t} d W_{t}

We ended with a deterministic partial differential equation for the derivative price.

The randomness has not disappeared from the world. It has been transformed. Under the assumptions, dynamic hedging lets the market price the derivative as if expected returns were replaced by the risk-free rate.

For a European call, solving the PDE with terminal payoff:

V (S, T) = max (S - K, 0)

produces the familiar Black-Scholes formula:

C = S_{0} N (d_{1}) - K e^{- r T} N (d_{2})

with:

d_{1} = \frac{ln ( S _{0} / K ) + ( r + \frac{1}{2} σ ^{2} ) T}{σ T}

and:

d_{2} = d_{1} - σ T

The formula is famous, but the equation is more important.

The formula prices a specific contract under specific assumptions. The method gives finance a language for contingent claims: write the payoff, model the underlying dynamics, apply stochastic calculus, hedge the risk, derive the pricing relationship.

That is why the work became larger than a single option formula.

Why This Was More Than a Formula

The Nobel committee emphasized that Black, Merton, and Scholes laid the foundation for rapid growth in derivative markets and that the method extended beyond stock options to other contingent claims, guarantees, insurance-like contracts, and real investment flexibility.

That breadth is the real breakthrough.

Once you can value a payoff contingent on a future state, you can start seeing optionality everywhere:

a call option on a stock;
a corporate bond exposed to default;
a guarantee embedded in an insurance product;
a project that can be delayed, expanded, or abandoned;
a risk system that decomposes exposure into sensitivities.

The same conceptual pattern appears again and again:

define the uncertain state;
define the payoff;
model the dynamics;
identify what risk can be hedged;
price what remains.

This is the point where stochastic math becomes quant finance.

Not because the model is perfect.

Because it creates a disciplined way to ask imperfect but useful questions.

The Assumptions Are the Beginning of the Next Story

Black-Scholes is not a magic description of markets. It is a model with sharp assumptions:

continuous trading;
frictionless markets;
no transaction costs;
no liquidity constraints;
constant volatility;
no jumps;
lognormal stock dynamics;
European exercise;
idealized borrowing and lending at the risk-free rate.

Real markets violate these assumptions constantly.

Volatility is not constant. It clusters. It smiles. It skews. Prices jump. Liquidity vanishes when you need it. Transaction costs turn elegant hedges into expensive churn. Discrete rebalancing leaves residual risk. The market's implied volatility surface tells you, every day, that the simple model is not the world.

But that does not make the model useless.

It makes it a baseline.

In practice, Black-Scholes is often less a literal belief than a coordinate system. Traders quote implied volatility because the model gives them a shared language. Risk systems compute Greeks because the model tells them how prices move with inputs. More advanced models are often best understood as corrections to the first clean benchmark.

The value of Black-Scholes is not that markets obey it. The value is that markets can disagree with it in measurable ways.

That is exactly where the next phase of this series goes.

If constant volatility is too simple, we study realized volatility, stochastic volatility, jumps, and volatility surfaces. If closed-form formulas stop helping, we use Monte Carlo. If hand-built assumptions strain under high-dimensional data, we bring in machine learning. If market microstructure contains signal at minute-level resolution, we build embeddings and probes.

But the intellectual chain remains the same:

graph LR
    A[Randomness] --> B[Calculus]
    B --> C[Pricing]
    C --> D[Risk]
    D --> E[Estimation]

The Bridge to Quant ML

This is also why the transition from stochastic math to machine learning in finance is natural.

Machine learning does not replace stochastic calculus. It enters where the assumptions become too rigid or the state space becomes too large.

Classical quant finance gives us concepts:

return;
volatility;
drift;
diffusion;
hedging;
calibration;
arbitrage;
risk-neutral valuation;
drawdown;
tail risk.

Machine learning gives us tools for estimation:

nonlinear feature interactions;
regime classification;
time-series embeddings;
volatility forecasting;
probability calibration;
signal fusion;
representation learning from market microstructure.

The danger is pretending that ML turns market noise into certainty. It does not. It gives us another way to estimate structure inside uncertainty.

That is why the next posts in this bridge series will move carefully:

geometric Brownian motion in Python;
Black-Scholes without the magic;
Monte Carlo option pricing beyond closed forms;
Value at Risk and breach testing;
volatility as the real signal;
why AUC is not a trading strategy;
sequence models and foundation models for market data.

The goal is not to abandon the stochastic series. The goal is to let it grow into the place where those ideas have teeth.

Quant finance is not just finance with equations.

It is the engineering discipline of making decisions under uncertainty when money, time, and risk are all part of the same system.

And the story starts with a simple problem:

A random path moves.

What is the fair price of a promise written on that path?

Previous Posts in the Randomness Thread

Solving Definite Integrals with SymPy & Monte Carlo introduces random sampling as a practical computational tool.
Monte Carlo Methods: Discovery & Evolution tells the broader history of randomness becoming a method for solving hard problems.
Stochastic Differential Equations: A Practitioner's Guide shows how noisy systems become equations you can simulate.
Bachelier's Revenge: The Forgotten Thesis of Modern Finance connects Brownian-style randomness directly to market prices.
Norbert Wiener: Cybernetics and Random Processes explains how continuous randomness became a rigorous mathematical object.
Antithetic Variates: Monte Carlo Variance Reduction shows the engineering side of making random simulation more efficient.
The Quiet Genius Who Made Randomness Calculable is the direct predecessor to this post.

References

Kiyosi Itô, "Stochastic Integral", Proceedings of the Imperial Academy, 1944.
Kiyosi Itô, "On a Formula Concerning Stochastic Differentials", Nagoya Mathematical Journal, 1951.
Kyoto Prize profile, Kiyosi Itô, 1998 Kyoto Prize in Basic Sciences.
Nobel Prize, The Prize in Economic Sciences 1997 - Press release, October 14, 1997.
Louis Bachelier, Theory of Speculation, 1900 thesis translation.

Contents