LessonThis is a lesson restated in new words. Any figure in it is an example from the source, not this desk's record.
Ito's lemma is the stochastic-calculus rule for the differential of a function that depends on time and on a random process. The lesson applies it to a derivative whose underlying follows geometric Brownian motion, then scales by a fixed holding. The portfolio change that comes out has a time term, a spot term, and a second-derivative term carrying volatility squared. That expansion is the lesson's bridge into the Black-Scholes argument for a European option under no arbitrage.
Why it matters
A portfolio of derivatives moves because the derivative value moves. If the derivative price is a smooth function of calendar time and of a jiggling underlying, ordinary calculus drops the terms that the randomness produces. Ito's lemma keeps the term that matters: the second derivative with respect to the underlying, multiplied by the squared shock, which does not vanish. The lesson stops once that differential is written. Hedging it down to a riskless rate is the next argument, and this session's job is to justify the differential the hedge will use.
The idea in plain words
Let H be the portfolio value, D the value of one derivative, and Q the number of derivative units. The lesson writes H as Q times D. Q is a constant count of contracts. D is not a constant. It depends on time and on the underlying price, and the underlying is taken to follow geometric Brownian motion, the model in which percentage moves have a drift and a volatility times a Brownian shock.
The portfolio differential is Q times dD, because Q factors out. dD is where the lemma enters. A first-order Taylor expansion in time and in the underlying would list partial derivatives times d t and dS, plus second-order pieces. Stochastic calculus then deletes most of the second-order pieces. The square of a time increment vanishes. The product of a time increment and a Brownian increment vanishes. The square of a Brownian increment equals the time increment. Applied to geometric Brownian motion, the square of dS collapses to volatility squared times price squared times d t. Mixed partials and the pure second derivative in time drop out. What survives in dD is the time partial times d t, the spot partial times dS, and one half the second spot partial times volatility squared times price squared times d t.
Multiply by Q and that is dH.
How to use it
Start from the product H = Q D and write dH = Q dD. State the inputs of D: time, and a price that obeys geometric Brownian motion. Expand dD with the lemma. Substitute the multiplication rules above so that every d t squared and every d t dW disappears, and replace the squared Brownian piece by d t. Cancel the price factors that appear when the squared geometric-Brownian-motion increment is expanded. The lesson's collected result is
dH = Q times (the time partial of D, times d t, plus the first spot partial of D, times dS, plus one half the second spot partial of D, times volatility squared, times the price squared, times d t).
Each partial is a partial of the original function D. The spot differential dS is still the geometric-Brownian-motion increment, so this expression is still random. The randomness sits in dS. The lesson's claim is that this is the right random differential to hand to a hedging argument. With a holding in the underlying chosen to cancel the dS term, the remaining movement can be tied to a riskless return, which is the no-arbitrage step behind European prices in the Black-Scholes model. This session derives the differential, and it names that use.
What usually goes wrong
Treating the derivative value as a constant wipes out the portfolio change. The contract count Q stays fixed, so it should not add a second random term. Dropping the one-half second-derivative term is the ordinary-calculus mistake, and it is the term that keeps volatility in the change after a hedge removes the spot shock.
The cancellations use three rules: a squared time increment is zero, a time increment times a Brownian increment is zero, and a squared Brownian increment equals the time increment. The underlying model is geometric Brownian motion. A different price equation leaves a different second-order piece.
Questions people ask
What is held constant in the portfolio?
Q, the number of derivative units. The moving piece is D, the derivative value.
Which second-order terms survive?
The second derivative of D with respect to the underlying, multiplied by one half, by volatility squared, by price squared, and by the time increment. The pure time second derivative and the mixed partial drop out.
What is this differential for?
The lesson presents it as the expansion behind a hedged European option book in the Black-Scholes argument, under a no-arbitrage reading of that hedge.