Chapter 16 · Brownian Bridge & Chapter Summary

Randomness With a Destination

What happens when a random walk is contractually obligated to arrive somewhere specific — and on time?
The pinned random walk

A standard Brownian Motion is a traveller who sets off from a starting point and wanders wherever chance takes them. The further into the future, the more spread out the possible positions become. Now imagine the same traveller — but with a strict contract. They must arrive at a specific destination at a specific time. They can still wander along the way, but the endpoint is fixed. This is the Brownian Bridge: randomness with a known destination.

Brownian Motion vs Brownian Bridge — spreading fan vs pinned paths converging to a fixed endpoint Two panels side by side. Left shows Brownian Motion paths spreading out like a fan as time increases. Right shows Brownian Bridge paths spreading to the middle then all converging back to zero at time T. Standard Brownian Motion Brownian Bridge X(T) = 0 Variance grows → ∞ as t → ∞ X(0)=0 X(T)=0 Variance = 0 at both ends, max at T/2

Left: Brownian Motion fans outward forever. Right: the Bridge spreads to the middle then all paths are pulled back to the pinned endpoint at time T.

Standard Brownian Motion W(t)

Variance: t — grows without bound.
At time T, paths can be anywhere.
No constraint on the endpoint.
Like a traveller with no fixed destination.

Brownian Bridge X(t)

Variance: t(T−t)/T — zero at both ends.
Highest uncertainty exactly at T/2.
Forced to X(T) = 0 (or any target b).
Like a traveller with a contracted arrival.

1 · Building the bridge — the tug formula

The Brownian Bridge is constructed from a standard Brownian Motion by a simple but clever operation. We take a random path W(t) and pull its endpoint back to zero by subtracting a correction proportional to the final value W(T).

X(t) = W(t) − (t/T) · W(T)
W(t) — the underlying Brownian Motion (the raw random path)
(t/T)·W(T) — the correction term: a straight-line "tug" pulling the end back to zero
At t = T: X(T) = W(T) − (T/T)·W(T) = W(T) − W(T) = 0 ✓

The path is free to wander — but the correction term always cancels out the endpoint.
Bridge construction — raw Brownian path W of t, the correction line t over T times W of T, and the resulting bridge X of t Three overlapping curves on the same axes. The raw Brownian path in blue ends high at time T. The correction line in red is a straight diagonal from zero to W of T. The bridge in gold is the difference — starting and ending at zero, wandering in between. 0 Time t → W(t) (t/T)W(T) X(t) bridge X(t) = W(t) − (t/T)·W(T)

The blue dashed path (W) ends high. The red line is the correction. Their difference — the gold bridge — starts at zero, wanders freely, and returns precisely to zero at time T.

2 · The variance profile — maximum uncertainty at the midpoint

The Bridge's defining characteristic is its variance profile. Unlike Brownian Motion (which grows forever), the Bridge's variance is shaped like an arch — rising from zero at the start, peaking at the exact midpoint, then returning to zero at the pinned endpoint.

Var(X(t)) = t(T−t) / T
At t = 0: Var = 0 (pinned start)  |  At t = T/2: Var = T/4 (maximum)  |  At t = T: Var = 0 (pinned end)
Variance profile comparison — Brownian Motion grows linearly while Bridge forms an arch peaking at midpoint Two curves on the same axes. The Brownian Motion variance is a straight line rising from bottom-left to top-right. The Brownian Bridge variance is an arch — rising from zero, peaking at T over 2, and returning to zero at T. Time t → Var BM: Var = t Bridge: max Var = T/4 t = T/2 0 T Bridge Var = t(T−t)/T BM keeps growing →

The Bridge's variance (gold arch) peaks at the midpoint then returns to zero. The Brownian Motion's variance (blue dashed) grows forever with no upper bound.

The covariance function c(s,t) = s∧t − st/T encodes the entire statistical structure of the Bridge. It shows that two positions at times s and t are correlated because they share the same anchor at time T — even if one is early and one is late, both are being pulled toward the same destination.

3 · The SDE — the magnetic drift

There is a second, equivalent way to describe the Brownian Bridge — as a Stochastic Differential Equation (SDE). This view reveals something extraordinary: the drift term acts like a magnet, pulling the process toward the target. And as time approaches the deadline, that magnetic pull becomes infinitely strong.

Theorem 16.1 — The Bridge SDE
dY(t) = − Y(t)/(T−t) · dt + dW(t)
−Y(t)/(T−t) dt
The magnetic drift. Pulls the process toward zero with a force proportional to current distance and inversely proportional to time remaining. Gets stronger as the deadline approaches.
dW(t)
The random shock. Standard Brownian noise — the process can still wander freely at each instant, fighting against the magnetic pull.
The panic at the end — magnetic drift force grows as time to deadline shrinks A series of horizontal bars showing the magnitude of the drift force at different times to deadline. Far from the deadline the force is small. Near the deadline the force is enormous. Strength of magnetic pull −Y/(T−t) as deadline approaches T−t = 1.0 year Force = 0.5× T−t = 0.5 year Force = 1.0× T−t = 0.1 year Force = 5.0× T−t = 0.01 year Force = 50× T−t → 0 → ∞

The magnetic drift force grows without bound as the deadline approaches. At T−t = 0, the pull is infinite — guaranteeing Y(T) = 0.

This "panic" is mathematically elegant: the infinitely strong drift at the deadline is exactly what guarantees the endpoint condition Y(T) = 0. Even if random shocks keep pushing the process away from zero, the magnet always wins at the last moment. It is the continuous-time analogue of "no matter how lost you get, you will arrive at the destination exactly on time."

4 · Why quants use the Brownian Bridge

The Brownian Bridge is not an abstract curiosity. It solves two practical problems that arise constantly in quantitative finance and simulation.

🎲 Bridge Sampling in Monte Carlo

Instead of simulating a price path step-by-step from t=0 to t=T, you can: (1) sample the endpoint S(T) directly, (2) place the bridge between S(0) and S(T), and (3) fill in the interior. This stratified sampling is computationally more efficient — fewer paths are needed to achieve the same accuracy.

📊 Conditional Path Inference

You know an NSE stock opened at ₹1,200 and closed at ₹1,185. A Brownian Bridge lets you reconstruct the most likely intra-day paths consistent with those two endpoints — for risk analysis, detecting intra-day patterns, or back-filling tick data gaps.

⏰ Barrier Option Pricing

When pricing a knock-out option that expires if the stock ever touches a barrier, the Bridge tells you the probability of a path touching the barrier between two observation points — even if neither observation point is at the barrier.

🔀 Path-Dependent Derivatives

Asian options, lookback options, and other path-dependent products require knowing the distribution of the path itself, not just the endpoint. The Bridge provides the conditional distribution of intermediate values, given start and end.

Bridge sampling addresses a key inefficiency of naive Monte Carlo: most paths cluster near the mean and contribute little to rare-event pricing. By conditioning on the endpoint, bridge sampling forces paths to span the full distribution more efficiently — reducing the number of simulations needed by an order of magnitude for tail-risk pricing.

Try it — build your own Brownian Bridge

Simulate multiple Brownian Bridge paths between a start value and an end target of your choosing. Watch the paths spread freely in the middle and then all converge to the same pinned endpoint at time T. Observe how the variance arch (shown below the paths) rises to a peak at T/2 and returns to zero.

0
0
20%
20
Max variance (theory)
Max variance (observed)
All paths end at b?
Paths simulated

Gold = the mean path  |  Coloured = individual bridge paths  |  All paths begin at a and converge to b at time T

Chapter summary — the complete cheat sheet

You have now completed the entire calculus of finance — from the first coin flip of Chapter 1 through to the Brownian Bridge of Chapter 16. Here is the definitive cheat sheet that maps every concept to its key tool and its practical payoff.

Concept Key tool Practical result
Random Walk ±1 coin flip Foundation of all price models — discrete randomness
Brownian Motion W(t) ~ N(0,t) The continuous limit — independent, normal increments
GBM S = S₀ exp{σW + (μ−½σ²)t} Stock prices: log-normal, always positive, drift + noise
Quadratic Variation (dW)² = dt Measures path roughness — the engine of stochastic calculus
Itô Integral ∫ Δ dW Running P&L — always a Martingale (no free lunch)
Itô-Doeblin Formula df = fₜdt + fₓdX + ½fₓₓdt The stochastic chain rule — how functions change with price
Black-Scholes PDE cₜ + rxcₓ + ½σ²x²cₓₓ = rc The equation that prices every option — Theta + Delta + Gamma = r
BSM Formula c = xN(d₁) − Ke⁻ʳᵀN(d₂) Closed-form option price — only σ matters, not α
Lévy's Theorem [M,M] = t + Martingale Identifies if a synthetic strategy is truly Brownian
Multivariable Itô + f_{xy} dX dY cross-term Multi-asset portfolio risk — the correlation interaction term
Brownian Bridge X(t) = W(t) − (t/T)W(T) Randomness with a fixed endpoint — Monte Carlo efficiency

Chapter 16 — the Brownian Bridge in three ideas:

What comes next: Chapter 5 — Risk-Neutral Pricing and Girsanov's Theorem. This is the tool that transforms a real-world probability measure (where stocks have drift α) into a risk-neutral measure (where everything is a Martingale). It is the "magic" that makes the BSM formula independent of α — and the foundation of all modern derivatives pricing.

Next Chapter: Exercises & Proving Ground →