Stochastic Calculus for Quants: A Primer

Amit Kumar Jha, UBS

Contents
1 Probability 2

2 Random Walk 2

3 Sigma Algebra 3

4 Filtrations 3

5 Wiener Process 4

6 Martingale 4

7 Types of Martingales 5

8 Ito Process 5

9 Ito’s Lemma 6

10 P-Q Measures 6

11 Monte Carlo Simulation 6

12 Stochastic Differential Equations (SDEs) 7

13 Geometric Brownian Motion (GBM) 8

14 Local Volatility 9

15 Stochastic Volatility 9

16 Black-Scholes Model 9

17 Girsanov’s Theorem 10

18 Jump Diffusion Models 10

19 Lévy Models 11

1
1 Probability
Probability, in its simplest form, is a way of expressing how likely something is to happen.
Think of it as a scale from 0 to 1. A probability of 0 means something won’t happen,
while a probability of 1 means it’s certain to happen. Everything in between gives a sense
of the likelihood.
In the world of finance, probability plays a crucial role. Financial markets are un-
certain by nature. When investors, traders, or even everyday people think about money
matters, they’re often dealing with probabilities, whether they realize it or not.
For instance, consider a company’s stock. There are numerous factors, both internal
(like the company’s earnings report) and external (like geopolitical events), that can
influence its price. While it’s impossible to predict the stock price with absolute certainty,
analysts and traders use probability to gauge the potential directions the stock might take.
Example: Let’s say an analyst believes there’s a 60% chance (or a probability of
0.6) that a particular stock will go up tomorrow due to a positive earnings report that’s
expected to be released. This doesn’t mean the stock will definitely rise. Instead, out of
many days with similar circumstances, the stock would go up about 60% of the time. If
the earnings report is better than expected, the stock might see a significant rise. If it’s
as expected, there might be a moderate rise. If the earnings are poor, the stock might
even fall. The probability helps the analyst communicate the level of confidence in the
prediction.
Moreover, probability isn’t just about predicting stock movements. It’s used in various
financial instruments, from options pricing to risk assessment in loans. For instance, a
bank might use probability to determine the likelihood of a borrower defaulting on a loan.
If there’s a high probability of default, the bank might charge a higher interest rate to
compensate for the increased risk.
In essence, probability gives a structured way to think about uncertainty, allowing
financial professionals to make more informed decisions in the unpredictable world of
finance.

2 Random Walk
At its core, a random walk is like taking a stroll where you don’t plan your steps in
advance; instead, each step you take is random and independent of the previous one.
Imagine walking blindfolded, taking steps forward or backward based on the flip of a
coin. Over time, your path will zigzag in a seemingly unpredictable manner.
In the world of mathematics, this concept has been formalized as the ”random walk”
model, where each step or movement is determined randomly. It’s a way to capture the
essence of unpredictability in various scenarios.
When it comes to finance, the random walk theory is often associated with stock
prices. Why? Well, stock prices are influenced by countless factors, from company-
specific news like earnings reports to broader events like geopolitical tensions or economic
policy changes. Predicting the exact movement of a stock price based on all these factors
is incredibly challenging. The random walk theory suggests that stock prices move in
a way where each day’s price change is random and independent of the change on the
previous day. Essentially, the theory posits that stocks take a ”random walk,” making it
difficult, if not impossible, to consistently predict their future movements.

2
 This idea has profound implications for investing. If stock prices truly follow a random
walk, then it suggests that strategies based on predicting short-term price movements
are doomed to fail in the long run. Instead, it would mean that the market is efficient,
reflecting all available information in current prices.
Example: Let’s dive deeper into our coin-flipping trader. Suppose this trader flips a
coin every morning. If it lands on heads, he buys a share of a particular stock, pushing
its price slightly up. If it lands on tails, he sells a share, causing a small drop in the stock
price. Over days, weeks, and months, the stock’s price chart would show a series of ups
and downs, each one independent of the last. This price chart would resemble a random
walk, highlighting the unpredictable nature of the stock’s movements based solely on the
trader’s coin flips.
However, it’s essential to note that in the real world, many believe that stock prices
don’t follow a pure random walk. There may be trends, patterns, or anomalies that can
be exploited. But the random walk theory serves as a baseline model, reminding investors
of the inherent unpredictability of financial markets.

3 Sigma Algebra
Imagine you’re trying to organize a vast collection of books in a library. You’d create cat-
egories, subcategories, and maybe even sub-subcategories. Now, think of these categories
as sets. In probability and statistics, we also deal with sets, especially when talking about
events. However, we need a systematic way to organize and handle these sets, especially
when we want to measure probabilities associated with them.
Enter the concept of a sigma algebra (or σ-algebra). It’s like a rulebook for organizing
these sets. This rulebook ensures three things:
1. The entire collection of possible outcomes (think of it as the complete library) is
always included. 2. If a set (or category) is in our rulebook, its complement (all books
not in that category) should also be in the rulebook. 3. If we have a sequence of sets
(like a series of subcategories), their union (combination) should also be in our rulebook,
even if we have infinitely many of them.
Why is this important? Well, when we’re dealing with probabilities, we need to ensure
that we can assign a probability to any set in our sigma algebra. It ensures consistency
and completeness in our probability assignments.

4 Filtrations
In everyday life, our knowledge about things grows over time. Think of it as watching a
movie: at the start, you know nothing about the plot, but as time progresses, the story
unfolds, and your information about it accumulates.
Filtrations capture this idea of accumulating information in a formal mathematical
way. It’s a sequence of nested information sets, where each set contains all the information
up to a specific time. As time moves forward, our set grows, including all the new
information.
In finance, filtrations play a vital role, especially in the world of derivatives pricing and
risk management. Financial professionals need to make decisions based on the information
they have at a particular time. Filtrations help in modeling this available information.

3
 Example: Let’s talk about a stock analyst tracking a company. On the 1st of the
month, she might only have the company’s past performance data. By the 15th, she
might get a press release about a new product launch. By the end of the month, she
might have the company’s quarterly earnings report. The information available to her
keeps growing as the month progresses. If we were to capture all the information she has
at each point in time in sets, we’d get a filtration. The set on the 15th would include the
set from the 1st plus the new product information, and the set at the end of the month
would include all that plus the earnings report.

5 Wiener Process
Imagine you’re watching the erratic motion of a tiny pollen grain floating on water under
a microscope. It darts around, making unpredictable movements in every direction. This
phenomenon, first observed by the botanist Robert Brown, is called Brownian motion.
Now, when mathematicians and physicists started studying this motion, they formalized
it into what’s known as the Wiener Process.
The Wiener Process is essentially a mathematical representation of this seemingly
random motion. It’s like taking our earlier concept of a random walk and making it
continuous, so instead of discrete steps, you have a continuous curve.
In finance, this concept becomes crucial. Stock prices, for example, don’t jump in
fixed intervals; they move every moment the market is open. The Wiener Process helps
model this continuous price movement, especially over short intervals. It’s a foundation
upon which many other financial theories and models are built.
Example: Let’s say you’re observing a stock’s price movement over a single day.
From opening to close, the stock’s price moves up and down, reacting to countless factors.
Over such a short duration, these movements might seem erratic and unpredictable, much
like the Brownian motion of a pollen grain. This is where the Wiener Process comes in,
providing a model that closely resembles the stock’s behavior over that day.

6 Martingale
Imagine you’re at a casino, playing a completely fair coin-toss game. Every time the coin
lands heads, you win a dollar, and every time it’s tails, you lose a dollar. If you were to
track your winnings over time, the pattern you’d see could be described as a martingale.
In a martingale, no matter how much you’ve won or lost so far, your expected winnings
in the next round are always zero.
The idea behind a martingale is that you can’t predict future outcomes based on
past events. It’s like saying, no matter what’s happened before, the future remains
unpredictable.
In finance, this concept is used to model certain types of asset prices or investment
strategies, suggesting that past price movements or returns don’t provide any useful
information to predict future ones.
Example: Let’s consider a hypothetical stock where any news or events affecting its
price are completely random and unforeseeable. Today, the stock might go up due to
positive unexpected news, and tomorrow it might drop due to some negative surprise. If
you were to invest in this stock, your expected return, regardless of its past performance,
remains constant over time. Such a stock’s price behavior can be modeled as a martingale.

4
7 Types of Martingales
Apart from the standard martingale, there are variations that describe different types of
random processes:
• Submartingale: Imagine the coin-toss game, but with a slight bias where heads
(your wins) come up a bit more often. Over time, you’re expected to have more
winning days than losing ones. In this case, your winnings are likely to increase
over time, on average. This scenario can be described as a submartingale.

Figure 1: Visualisation of types of Martingales

• Supermartingale: Now, consider the opposite. The game is slightly biased against
you. Over time, you’re more likely to lose than win. Your expected winnings
decrease over time. This situation is a supermartingale.
In the financial world, these concepts help professionals understand and model the be-
havior of assets or investment strategies under different assumptions about their expected
returns.

8 Ito Process
Picture a boat gently floating down a river. The river’s current (which is always in one di-
rection) represents a steady force called the ”drift,” while the boat’s bobbing due to waves
and wind represents random, unpredictable movements, termed ”diffusion.” The combi-
nation of these two elements—the steady current and the random bobbing—captures the
essence of the Ito Process.
Named after Kiyoshi Ito, the Ito Process is a mathematical model that describes such
phenomena. It expands on the Wiener Process by adding a drift term, which represents
a consistent trend, and a diffusion term, which captures the random fluctuations.
In finance, this model becomes particularly useful. For instance, a stock might have
a general upward growth trend (thanks to the company’s performance, sector growth,
etc.), but day-to-day trading might cause random price fluctuations.

5
 Example: Consider a technology stock that’s been growing steadily over the years
due to consistent innovation and market leadership. However, daily news, trader senti-
ment, or market events lead to price volatility. This stock’s price movement—having a
general upward trend but also experiencing random fluctuations—can be aptly described
using the Ito Process.

9 Ito’s Lemma
Calculus students are likely familiar with the chain rule—a method that helps differentiate
a composite function. Now, what if our function is a bit more unpredictable, with random
movements? That’s where Ito’s Lemma comes in. It’s essentially the chain rule, but for
stochastic processes.
Named after the same Kiyoshi Ito, this lemma is a cornerstone in stochastic calculus.
It provides a way to differentiate functions of stochastic processes, paving the path for
many advanced financial models, especially those concerning option pricing.

10 P-Q Measures
In the world of finance, understanding and measuring risk is crucial. To do this, profes-
sionals often switch between two views or ”measures” of the world: the P measure and
the Q measure.
1. The P measure (or physical measure) is the ”real-world” view. It’s about ac-
tual probabilities, representing how assets like stocks are expected to behave based on
historical data and future predictions.
2. The Q measure (or risk-neutral measure) is more of a hypothetical view. Here, we
assume that all assets grow at the risk-free rate (like the rate of a government bond). It’s
not about what we truly expect to happen but provides a simplified world that’s crucial
for pricing derivatives.
The switch between these measures is more than just a mathematical trick. It’s rooted
in a fundamental finance principle: there shouldn’t be any arbitrage opportunities (free
money) in the market.
Example: Imagine you’re a quant trying to price an option (a financial derivative).
In the real world, the stock associated with that option might have all sorts of expected
growth rates, volatilities, and risks. However, to price the option, you’d switch to the Q
measure, simplifying your calculations and ensuring the price you arrive at doesn’t allow
arbitrage.

11 Monte Carlo Simulation

Picture yourself in a casino, standing before a roulette table. As the wheel spins and
the ball bounces, it seems almost impossible to predict where it will land. But what if
you could spin that wheel thousands or even millions of times and record every outcome?
Over time, you’d start to see patterns or probabilities emerge. This idea of understanding
complex systems through repeated random sampling is at the heart of the Monte Carlo
Simulation. Named after the famous Monaco casino town, the Monte Carlo Simulation
is a computational technique used to estimate the probability of different outcomes.

6
 Figure 2: Visualisation of Monte Carlo Simulation

It’s a bit like conducting a vast number of ”what if” scenarios to understand potential
future events. By repeatedly simulating a process with random inputs, we can obtain
a distribution of outcomes, helping us understand the probabilities and risks associated
with a particular system or decision.
In finance, Monte Carlo Simulation becomes invaluable. Financial markets are inher-
ently complex and filled with uncertainties. Instead of trying to predict the future with
a single deterministic model, the Monte Carlo method allows professionals to explore a
myriad of potential scenarios, understanding not just the most likely outcomes but also
the range of possibilities.
Example: Consider an investment portfolio comprising various assets: stocks, bonds,
commodities, etc. Predicting its future value is challenging due to the countless factors
influencing each asset. Using Monte Carlo Simulation, an investor can simulate thousands
of potential future market scenarios, each with different market returns, interest rates,
and economic conditions. After running these simulations, the investor won’t get a single
predicted value for the portfolio but rather a distribution of potential values, helping
them assess the portfolio’s risk and potential return.

12 Stochastic Differential Equations (SDEs)

At its core, a differential equation is like a puzzle or a riddle. It provides a relation-
ship between something and its rate of change, and solving it gives us insight into the
behavior of the system described by the equation. But what if this system isn’t stable
and predictable? What if, like the weather, it’s subject to random and unforeseeable
influences?
This is where Stochastic Differential Equations (SDEs) come into play. These are
equations that, in addition to the usual terms, have components that behave unpre-
dictably or ”stochastically.” The solutions to these equations aren’t precise trajectories
but rather a myriad of possible paths, each with a certain likelihood.
In finance, many assets and instruments are influenced by a multitude of unpredictable

7
factors: sudden news, geopolitical events, market sentiment shifts, and more. SDEs
provide a framework to model and understand these assets, taking into account both the
deterministic trends and the random fluctuations.
Example: Consider a company’s stock price. While it might generally grow due
to solid performance, it’s also subject to unexpected news—like a sudden merger or a
regulatory hurdle. An SDE can help model this stock price by incorporating both its
general growth trend and the random shocks it might experience.

13 Geometric Brownian Motion (GBM)

Imagine you’re watching a tree grow. Over the years, the tree not only gets taller but also
wider, its branches more spread out. This growth isn’t linear; the bigger the tree gets,
the more it grows each year. Now, add to this growth some randomness—like varying
weather conditions affecting the tree’s growth differently each year.
This combination of consistent growth and randomness is what Geometric Brownian
Motion (GBM) captures. In the mathematical world, GBM is a model that describes a
quantity that grows steadily and is simultaneously subject to random changes. In the

Figure 3: Visualisation of GBM

realm of finance, GBM becomes particularly relevant when talking about stock prices.
While a company’s stock might have a general trend—maybe due to company perfor-
mance or overall market conditions—it also experiences random fluctuations based on
countless factors.
Example: Take a booming tech company. Its stock might generally be on the rise
due to consistent innovation and market demand. However, on any given day, the stock
could go up or down based on various factors: a product announcement, a competitor’s
move, or global market conditions. The GBM model captures this behavior, reflecting
both the stock’s general growth trend and its day-to-day randomness.

8
14 Local Volatility
Imagine you’re on a long road trip. As you drive through different terrains and weather
conditions, the speed of your car varies. On a clear highway, you might speed up, but
in a rainy mountain pass, you’d naturally slow down. This varying speed, depending on
your location and conditions, can be likened to the concept of local volatility.
In finance, local volatility refers to the idea that volatility (or the rate at which a
financial instrument’s price moves) isn’t constant but varies depending on factors like
the current price level and time. Developed as a way to refine the Black-Scholes Model,
which assumes constant volatility, local volatility models provide a more dynamic view
of market behavior, adjusting volatility based on observed market conditions.
Example: Think of a stock that historically sees significant price swings whenever it
approaches a particular price level, maybe due to psychological or historical reasons. A
local volatility model would capture this behavior, adjusting the volatility higher when
the stock is near that price level and possibly lower when it’s far from it.

15 Stochastic Volatility
Now, instead of a road trip, imagine you’re sailing on the open sea. The intensity of
the waves (or the sea’s volatility) isn’t just determined by where you are (like near an
island or in the deep ocean) but also by random and unpredictable factors like sudden
wind gusts or distant storms. This unpredictability in the intensity of waves is akin to
stochastic volatility.
Stochastic volatility models in finance embrace the idea that volatility itself is random
and can change unpredictably over time. While local volatility models adjust volatility
based on factors like price level and time, stochastic volatility models introduce an addi-
tional layer of randomness, acknowledging that markets can be influenced by unforeseen
events or shifts in sentiment.
Example: Consider a global event like a sudden geopolitical conflict. Such an event
might cause markets worldwide to become more volatile, not because of specific asset
prices or historical patterns, but due to the uncertainty and unpredictability introduced
by the event. A stochastic volatility model would capture this kind of random spike in
volatility, providing a more comprehensive view of market risks.

16 Black-Scholes Model
When you’re cooking a dish, knowing the recipe helps you predict the outcome. Similarly,
in the world of finance, especially in options trading, the Black-Scholes Model acts as a
recipe. Developed by Fischer Black, Myron Scholes, and Robert Merton in the early
1970s, this model provides a theoretical estimate of the price of European-style options.
Just as a recipe requires specific ingredients in precise amounts, the Black-Scholes
Model considers various factors to estimate an option’s price. These factors include the
current stock price, the option’s strike price, the time until the option expires, the stock’s
volatility, and the risk-free interest rate.
The beauty of the Black-Scholes Model lies in its ability to boil down these multi-
ple factors into a single formula, offering traders and investors a standardized way to

9
value options. However, it’s essential to understand that like any model, it’s based on
assumptions, some of which might not hold in real-world scenarios.
Example: Imagine an investor trying to decide whether to buy an option on a tech
company’s stock. Using the Black-Scholes Model, they can input the current stock price,
how long until the option expires, and other factors into the formula. The resulting value
gives them an estimate of what the option should be worth, helping guide their decision.

17 Girsanov’s Theorem
Translators help us understand one language in terms of another. In the realm of math-
ematical finance, Girsanov’s Theorem plays a similar role, but for probability measures.
When dealing with financial models, especially those involving stochastic processes, we
often encounter different ”views” or ”measures” of probability. The most common ones
are the P (physical or real-world) measure and the Q (risk-neutral) measure. Girsanov’s

Figure 4: Visualisation of Girsanov

Theorem provides the mathematical framework to transition between these measures.

It’s like a bridge that ensures a smooth passage, preserving the structure and properties
of our models.

18 Jump Diffusion Models

Nature is full of surprises. While things might seem calm and predictable, sudden events
can disrupt the status quo. The same holds true for financial markets. Stock prices,
for instance, might exhibit a steady trend but can experience abrupt changes due to
unexpected news or events. Jump Diffusion Models capture this dual nature. They
combine the usual random walk or diffusion process (the calm river flow) with sudden
jumps (the big fish causing ripples). These jumps can represent sudden market reactions
to major news, like mergers, regulatory changes, or geopolitical events.
Example: Consider a pharmaceutical company’s stock. While its price might exhibit
typical market fluctuations, the announcement of a breakthrough drug or, conversely, a

10
 Figure 5: Jump Diffusion Model for a Pharmaceutical Company’s Stock

failed clinical trial can cause a sudden and significant jump or drop in its stock price.
Jump Diffusion Models can incorporate these abrupt changes, providing a more compre-
hensive view of the stock’s price dynamics.

19 Lévy Models
Imagine you’re watching a serene pond, where occasionally a stone is thrown, causing
ripples to spread across the water. Most of the time, the pond is calm, but these random
disturbances create sudden and noticeable effects. Lévy Models in finance are akin to
this scenario, where asset prices mostly evolve smoothly but can be impacted by sudden
and significant jumps.
Named after the French mathematician Paul Lévy, Lévy Models are a class of stochas-
tic processes that incorporate both continuous paths (like the calm pond) and discontinu-
ous jumps (the ripples from the thrown stones). While traditional models like Geometric
Brownian Motion describe asset prices as smooth paths with some randomness, Lévy
Models introduce the possibility of abrupt changes or jumps in these prices.
These jumps can be due to various reasons: sudden news releases, major geopolitical
events, or any other unexpected occurrences that can drastically affect market sentiment.
By incorporating these jumps, Lévy Models offer a more realistic representation of asset
price dynamics, especially in markets known for their abrupt movements.
Example: Consider a biotechnology company awaiting regulatory approval for a new
drug. For months, the stock might exhibit typical market fluctuations. However, the day
the approval (or rejection) news is released, the stock might experience a significant jump
(or drop) in price. A Lévy Model would be well-suited to describe this stock’s behavior,
accounting for both its usual price movements and the potential for sudden jumps based
on significant news.

11