Differential Calculus and Optimization Basics › Derivatives and Rates of Change

In finance, calculus derivatives provide local sensitivity measures. Consider a simple example: the profit function of a market maker.

Financial derivatives such as options and futures depend on an underlying asset, rate, index, or other reference. Calculus derivatives are one way to measure how a differentiable pricing model responds to a small change in that reference or another model input. For example, the sensitivity of an option pricing function to the underlying stock price is its partial derivative with respect to that price. Discrete scenarios, finite differences, and lattice methods can measure related changes without requiring an analytic derivative.

Suppose a market maker's daily profit $\Pi$ depends on trading volume $V$:

$$\Pi(V) = 0.001V - 500 - 0.00000001V^2$$

where:

• $\Pi$: daily profit in dollars
• $V$: trading volume in units
• $0.001V$: revenue from bid-ask spread ($0.001 per unit)
• $500$: fixed daily operating costs
• $0.00000001V^2$: market impact costs that grow quadratically with volume

This profit function captures three fundamental aspects of market making. The first term represents the revenue the market maker earns from the bid-ask spread: every time a unit is traded, the market maker captures a small spread of $0.001. If this were the only factor, profit would grow linearly forever. The constant term of 500 represents the fixed costs that must be paid regardless of volume: technology, salaries, rent, and regulatory fees. These costs must be covered before any profit is realized.

The quadratic term is an illustrative market-impact model rather than a universal empirical scaling law. Within this example, total impact cost grows with $V^2$, so marginal impact cost grows linearly with volume and eventually offsets the fixed per-unit spread revenue. Actual impact depends on liquidity, execution horizon, order type, and market conditions; the negative quadratic coefficient is a modeling assumption chosen to create this trade-off.

The first term represents revenue (a small spread earned per unit volume), the constant represents fixed costs, and the quadratic term captures market impact costs that grow with volume. The derivative:

$$\Pi'(V) = 0.001 - 0.00000002V$$

where:

• $\Pi'(V)$: the marginal profit at volume level $V$
• $0.001$: the marginal revenue per unit (the bid-ask spread)
• $0.00000002V$: the marginal market impact cost, which increases linearly with volume

This is the marginal profit: the local rate of profit change per unit of trading volume. A small increment $\Delta V$ changes profit by approximately $\Pi'(V)\Delta V$. For this quadratic, the exact increment is $\Pi'(V)\Delta V-10^{-8}(\Delta V)^2$, so a one-unit increment is not exactly the derivative. When $\Pi'(V) > 0$, a sufficiently small volume increase improves profit. When $\Pi'(V) < 0$, marginal market-impact cost exceeds marginal spread revenue.

The original function tells us total profit at any volume level, while the derivative tells us whether a feasible small increase or decrease in volume improves the modeled result. Both questions matter: "Is this activity profitable?" and "Would trading more or less improve it?" The profit level answers the first; the derivative helps answer the second.

A major application of derivatives in finance is the calculation of option sensitivities, known as "the Greeks." These measure how an option's model price responds locally to changes in specified inputs while the other inputs are held fixed.

The Greeks earned their collective name because most of them are denoted by Greek letters. The five introduced here summarize several important local sensitivities under a specified pricing model and held-fixed inputs. They are useful for model-based pricing and hedging, but they neither guarantee exact dollar changes nor capture every exposure, such as jumps, volatility-surface dynamics, cross-Greeks, higher-order terms, transaction costs, or model error.

Consider a European call option with model price $C(S,T,\sigma,r;K)$, holding strike $K$ and other contract terms fixed. Its varying inputs are stock price $S$, time to expiration $T$, volatility $\sigma$, and risk-free rate $r$. These Greeks are first- or second-order derivatives, with a minus sign for Theta because $T$ measures time remaining:

• Delta ($\Delta$): $\frac{\partial C}{\partial S}$ measures sensitivity to stock price changes
• Gamma ($\Gamma$): $\frac{\partial^2 C}{\partial S^2}$ measures the rate of change of delta
• Theta ($\Theta$): $-\frac{\partial C}{\partial T}$ measures sensitivity to time decay
• Vega ($\nu$): $\frac{\partial C}{\partial \sigma}$ measures sensitivity to volatility changes
• Rho ($\rho$): $\frac{\partial C}{\partial r}$ measures sensitivity to interest rate changes

where:

• $C$: the call option price
• $S$: the current stock price
• $T$: time to expiration in years
• $\sigma$: the model volatility input; when inferred from an option price, it is the implied volatility for that specified contract, strike, maturity, and model
• $r$: annualized risk-free interest rate, expressed as a decimal

The displayed model uses annualized decimal volatility and positive $S,K,T,\sigma$. Expiry and zero-volatility limits need separate treatment. These raw derivatives are per unit of their specified inputs: Vega and Rho quoted per one-percentage-point change are multiplied by $0.01$, while a per-day Theta quote requires a stated days-per-year convention. The Options Industry Council's Greek definitions describe the same local sensitivities; quotation units must still be kept consistent.

Delta is widely used in option risk management. If an option quoted per share has a delta of 0.5, then a small $1 increase in the stock price changes the model price by approximately $0.50 when other inputs are fixed. Traders can offset an option portfolio's delta with stock or other options. A delta-neutral portfolio has net first-order underlying-price sensitivity near zero, so small stock moves have little first-order effect under the model; the hedge must be rebalanced as delta and other inputs change.

Gamma adds depth to the delta picture by measuring how delta itself changes. An option with high gamma has a delta that changes rapidly as the stock price moves. This is particularly important near expiration for near-the-money options. For a dynamically delta-hedged long-gamma position, sufficiently large realized moves can produce rebalancing gains, but those gains are not automatic and must be weighed against theta, transaction costs, jumps, and other exposures. A short-gamma position faces the opposite convexity and can require frequent hedge adjustments.

These Greeks provide a useful but incomplete local sensitivity summary under the chosen model. Delta and Gamma describe first- and second-order underlying-price effects; Theta describes passage-of-time sensitivity; Vega describes model-volatility sensitivity; and Rho describes interest-rate sensitivity. Hedging them addresses those modeled local exposures, not jumps, cross-effects, volatility-surface changes, liquidity costs, or model risk.

A later Black-Scholes chapter presents their explicit formulas and interpretations. For now, the first-order Greeks approximate the price response to a small change in the corresponding input. Gamma measures how Delta changes and contributes the second-order term $\frac{1}{2}\Gamma(\Delta S)^2$ to a smooth stock-price response, with other inputs held fixed.