All option guides
Two conventions for noisy dynamics12 min read

Stratonovich vs Itô Calculus Explained

Learn how Stratonovich and Itô stochastic integrals differ, how to convert their drift terms, and why finance models must state the convention

Prepared by Mark · Primary sources below

Direct answer

Itô and Stratonovich calculus describe the same suitably converted stochastic dynamics, but they evaluate noisy integrands at different points in each interval. The conversion adds a quadratic-variation drift term, so an equation is incomplete until its integral convention is named

The sampling point changes the integral

An Itô integral uses information at the left endpoint of each time interval. This makes its integrand predictable and gives the integral a martingale property under standard assumptions

A Stratonovich integral uses a midpoint-style limit. It follows the ordinary chain rule more closely for smooth coordinate changes

Neither notation is automatically more realistic. The appropriate convention follows from the modeling limit, the information flow, and the quantity being represented

Converting conventions changes drift

For a one-dimensional diffusion, dX = a(X)dt + b(X)∘dW in Stratonovich form becomes dX = [a(X) + 1/2 b(X)b'(X)]dt + b(X)dW in Itô form

The extra term is not a discretionary adjustment. It comes from the quadratic covariation between the state-dependent integrand and Brownian motion

When b is constant, b' is zero and the two forms agree. State-dependent volatility is where silent mixing becomes costly

A multiplicative-noise example exposes the difference

Take b(X) = σX and a zero Stratonovich drift. Its equivalent Itô drift is 1/2σ²X

Using σ = 40% makes that correction 8% of X per year before any separately specified economic drift

That number is a change of representation, not an expected stock return or a tradable alpha. A finance model must still specify its measure, carry, dividends, and constraints

Pricing models normally state Itô dynamics

Continuous-time asset-pricing models commonly use Itô processes because self-financing trading strategies are adapted to information available before the next shock

Itô's lemma, martingale measures, and the standard Black–Scholes derivation are written in that convention

Physical models derived from colored-noise limits can naturally produce Stratonovich equations. Convert them before combining them with an Itô pricing or hedging calculation

The conversion does not fix model risk

An equation may be correctly converted and still omit jumps, stochastic volatility, liquidity, discrete hedging, or misspecified parameters

Multidimensional systems require the full Jacobian and covariance structure, not a scalar half-times-b-times-b-prime shortcut

Always state the SDE, integral convention, probability measure, time unit, and interpretation of the noise before comparing results

Common questions

Is Stratonovich calculus more correct than Itô calculus?

No. They are equivalent descriptions when converted correctly. The right form depends on how the stochastic model arises and how it will be used

Why does the drift change during conversion?

The integrand can co-move with Brownian motion when it depends on the state. Its quadratic covariation survives as a finite drift correction

Can I use the ordinary chain rule with an Itô process?

Not directly. Itô's lemma adds a second-order term from quadratic variation. Stratonovich notation restores the ordinary-looking rule only with its own convention

Sources and further reading

Related guides