Chapter 11 · Itô Calculus

The Itô-Doeblin Formula

The "Law of Relativity" for finance — how functions change in a random, vibrating world.
From hourly to every microsecond

In Chapter 10, our trader made decisions once per hour and held positions fixed in between. That was a useful starting point. Now we remove that restriction entirely. We allow position size Δ(t) to change continuously — every tick, every microsecond, every instant. This is how a modern algorithmic trading system actually operates. And the remarkable result is that everything we proved in Chapter 10 still holds.

Evolution from step-function trading to continuous trading — staircase becomes smooth curve Three panels showing progressively smoother position size functions: coarse staircase, finer staircase, then a smooth continuous curve Chapter 10 Hourly steps Getting finer Many small steps Chapter 11 Continuous Δ(t) Δ fixed per hour Δ fixed per minute Δ(t) flows smoothly

The limit of faster and faster step-traders becomes a continuous trader. All three Itô theorems from Chapter 10 carry over exactly.

This generalisation is not just a mathematical nicety. It is the foundation of algorithmic trading. Every strategy that adjusts positions in real time — based on price, volatility, or model signals — is operating in this continuous regime. The mathematics of the Itô Integral, extended to continuous Δ(t), is what makes these strategies formally describable and analysable.

2 · A classic trap: ∫W(u)dW(u) and the cost of jaggedness

Here is a trap that catches every student who tries to apply school calculus to stochastic processes. Consider the simplest possible "continuous strategy": at every moment, hold exactly W(t) shares — whatever the current price level is. In ordinary calculus, the integral of x dx is simply ½x². You would expect the same answer here. You would be wrong.

School Calculus
∫₀ᵗ x dx = ½x²
Smooth function.
No extra terms.
Exactly what you expect.
Itô Calculus
∫₀ᵗ W(u) dW(u) = ½W²(t) − ½t
Random path.
An extra term appears.
Where did it come from?

That highlighted −½t is called the Cost of Jaggedness. It appears because of a fundamental asymmetry: in real trading, you must set your position at the start of each tiny interval — before seeing which way the price moves. The path's constant vibration means you are always slightly behind the curve, and that lag accumulates into a real, measurable loss of value equal to exactly half the elapsed time.

The cost of jaggedness — position set at start of interval versus actual price move during interval A zoomed-in view of a price path during one tiny interval, showing the position was set at the beginning but the price zigzagged, creating a gap between expected and actual integral value One tiny interval [tⱼ, tⱼ₊₁] Position set here → Δ(tⱼ) = W(tⱼ) Gap = cost of jaggedness Position locked here (start of interval) Cannot see what will happen next →

Position must be set at the start of each interval. The path's jaggedness creates a gap that accumulates into the −½t correction.

This is also why Itô calculus is chosen over the alternative "Stratonovich calculus" for finance. In Stratonovich's approach, the position is set at the midpoint of each interval, which makes the −½t term vanish — producing a cleaner formula. But in the real world, you cannot see the midpoint of an interval before it has passed. Itô calculus is the only version that respects the non-anticipatory constraint of actual trading.

3 · The Itô-Doeblin formula: the new chain rule

In ordinary calculus, the Chain Rule tells you how a function changes when its input changes: if f depends on x, and x is changing, then df = f'(x) dx. The Itô-Doeblin Formula is the stochastic version of this Chain Rule — valid when x is Brownian Motion and therefore cannot be differentiated in the ordinary sense.

THEOREM 11.1 — THE ITÔ-DOEBLIN FORMULA
The Stochastic Chain Rule
df(W(t)) = f'(W(t)) dW(t) + ½ f''(W(t)) dt
f'(W) dW — the ordinary part: how the function tilts as price moves (the Delta)

½ f''(W) dt — the Itô correction: an extra push from the curvature of the function, caused by the path's constant vibration (the Gamma)

In school calculus, the second term would be zero because smooth functions have no quadratic variation. In stochastic calculus, (dW)² = dt keeps the second term alive and non-zero.
Ordinary Chain Rule
df = f'(x) dx
Only one term.
First derivative only.
Works for smooth paths.
Itô-Doeblin Formula
df = f'(W) dW + ½f''(W) dt
Two terms always.
Second derivative appears.
Required for random paths.

The extra term — the Itô correction — arises from the curvature of the function f. Imagine driving on a road that is violently vibrating. Even if you hold the steering wheel perfectly straight, the jiggling of the car creates extra distance traveled. In finance, if a stock is volatile, the average value of a derivative shifts even when the stock price stays in the same general area — simply because of the curvature of the payoff function.

Convexity and the Ito correction — curved function shifts upward under volatility A curved convex function f of x, with a straight line showing the first-order approximation and a gap between the line and the curve showing the half f-double-prime dt Ito correction f(x) x₀ f'(x)dx — first order only ½f''dt Itô correction x (Price) f Convex curve: f'' > 0 Itô correction pushes value UP

For a convex function (f'' > 0), the Itô correction adds value — volatility is beneficial. For a concave function (f'' < 0), it subtracts value. This asymmetry is the mathematical basis of the convexity advantage in bond pricing and options.

In options pricing, f'' is called Gamma (Γ). A positive Gamma means the option's value is a convex function of the stock price — so the Itô correction ½f''dt is always positive. This is why options buyers benefit from volatility: the curvature of the payoff function means the Itô term is always working in their favour, regardless of which direction the market moves.

4 · Itô processes: the general stock model

Most real assets are not pure Brownian Motion. They have both a trend and volatility that can vary over time. The Itô Process captures both in one compact expression — and the Itô-Doeblin Formula applies to any smooth function of it.

dX(t) = Δ(t) dW(t) + Θ(t) dt
Δ(t) dW(t) — the random part: Brownian shock scaled by the volatility Δ at each moment
Θ(t) dt — the predictable part: the drift or trend, the "escalator" from GBM
Both Δ and Θ can change continuously over time — this is what makes it a general model

When the Itô-Doeblin Formula is applied to a general Itô Process instead of pure Brownian Motion, it extends to account for both the random and the trend components. The result, for a function f(t, x) of both time and an Itô Process, is:

THEOREM 11.2 — ITÔ-DOEBLIN FOR ITÔ PROCESSES
The Full Stochastic Chain Rule
df(t, X(t)) = fₜ dt + fₓ dX + ½fₓₓ (dX)²

= fₜ dt + fₓ (Θ dt + Δ dW) + ½fₓₓ Δ² dt
After applying the Multiplication Table (dW·dW = dt, everything else = 0), this simplifies to:

df = (fₜ + Θfₓ + ½Δ²fₓₓ) dt + Δfₓ dW

A deterministic drift part (dt) and a random shock part (dW) — exactly the structure of an Itô Process itself.

The calculation relies on the same Multiplication Table from Chapter 8. Applied to the Itô Process, it eliminates the small terms and leaves exactly the right structure:

×dWdt
dW dt → survives as Δ²dt 0
dt 0 0
5 · Reading the formula as a price landscape

The full Itô-Doeblin Formula for a surface f(t, x) breaks the change in value into exactly three pieces, each with a precise financial meaning. Together they describe how any derivative — an option, a structured product, a strategy payoff — responds to the passage of time and the movement of the underlying asset.

fₜ
Time Decay
How the value sinks as time passes, holding price fixed. For options, this is the daily erosion of time value.
Θ (Theta)
fₓ
Price Sensitivity
How the value tilts as price moves by one unit. The slope of the surface in the price direction.
Δ (Delta)
fₓₓ
Convexity Boost
How the curvature of the surface interacts with market volatility. The Itô correction — the "vibration push."
Γ (Gamma)
Price landscape surface decomposed into three forces — time decay, price delta, and gamma convexity A 3D-style surface diagram with three labeled arrows: one pointing downward for theta time decay, one pointing along the price axis for delta, and one curving upward for gamma convexity fₜ (Theta) fₓ (Delta) fₓₓ (Gamma) Time → Price → Every point on the surface feels all three forces simultaneously

Theta pulls the surface down over time. Delta tilts it along the price axis. Gamma bends it upward with volatility. All three act at once.

This three-part decomposition is precisely the Greeks framework used by every options desk in the world. Theta is the daily time decay you see on your P&L. Delta is the hedge ratio — how many shares of the underlying you need to hold to neutralise price risk. Gamma is the curvature that makes options profitable in volatile markets. The Itô-Doeblin Formula is the mathematical proof that these three forces are exhaustive — together, they describe all the ways a derivative can change in value.

Try it — see the Itô correction in action

Run the classic example: hold W(t) shares at each moment (Δ = W). The school calculus answer would give ½W²(t). The Itô answer gives ½W²(t) − ½t. Simulate many paths and verify that the −½t correction is real and measurable — not a mathematical abstraction, but an actual average shortfall in your P&L.

1.0 yr
252
Avg ½W²(T) [school]
Avg ∫W dW [Itô]
Difference
Predicted −½T

Purple = school calculus answer (½W²)  |  Gold = Itô integral (½W² − ½t)  |  The gap between them is the −½t correction, confirmed empirically

Chapter 11 in four ideas:

What comes next: The Itô-Doeblin Formula applied to one specific, carefully chosen function — the price of an option — produces the Black-Scholes-Merton equation. That is the formula that transformed Wall Street. You now have all the tools needed to derive it from first principles.

Next Chapter: Working Examples (GBM, Vasicek & CIR) →