Derivatives and Differentiation Rules

An option's delta tells you how much its price moves for a tiny move in the underlying. That's not an average over some range of prices — it's the instantaneous rate of change, right at the current price. Average rate of change over a big move is easy: divide the total price change by the total underlying move. But "instantaneous" — shrinking that move down to essentially nothing while still getting a meaningful number out — needs a precise definition, and that's exactly what the derivative provides.

An analogy: a speedometer versus a trip odometer

A car's average speed over a trip is just total distance divided by total time — easy, but it tells you nothing about how fast the car was going at any particular instant, especially if it sped up and slowed down along the way. A speedometer reads something different: the speed right now, at this exact moment. The derivative is the mathematical speedometer — it takes the same "distance over time" idea used for averages and shrinks the time window down toward zero, until what's left is a genuine instantaneous rate rather than an average over some stretch.

The math, one symbol at a time

The derivative of $f$ at a point $x$ is defined as the limit

$$f'(x) = \lim_{h \to 0} \frac{f(x+h) - f(x)}{h}.$$

In words: take the average rate of change between $x$ and a nearby point $x+h$ — exactly "change in output over change in input" — and see what value that average settles down to as the gap $h$ shrinks toward zero. A handful of rules let you compute derivatives without ever using this limit directly: the power rule $\frac{d}{dx}x^n = nx^{n-1}$; the product rule $\frac{d}{dx}[f(x)g(x)] = f'(x)g(x) + f(x)g'(x)$; and the chain rule $\frac{d}{dx}f(g(x)) = f'(g(x))\,g'(x)$, which handles a function nested inside another — differentiate the outer function, then multiply by the derivative of what's inside.

Worked example 1: bond price sensitivity by the power rule

A simplified bond price as a function of yield $y$ is $P(y) = \frac{1000}{(1+y)^{10}}$ (a 10-year zero paying $1,000 at maturity). Rewriting as $P(y) = 1000(1+y)^{-10}$ and applying the power rule (with the chain rule, since the inside is $1+y$, whose derivative is 1): $P'(y) = 1000 \times (-10)(1+y)^{-11} \times 1 = -10000(1+y)^{-11}$. At $y = 0.05$: $P'(0.05) = -10000(1.05)^{-11} \approx -10000 \times 0.5847 \approx -5847$. This says: for a small increase in yield, the price falls by roughly $58.47 per 1% (0.01) move in yield — exactly the intuition behind bond duration, computed directly from the power and chain rules.

Worked example 2: differentiating a product

A trading strategy's daily P&L is modeled as $f(t) = t \cdot e^{-0.1t}$ (a signal that grows linearly but decays exponentially over time $t$, in days). By the product rule, with $u=t$ ($u'=1$) and $v=e^{-0.1t}$ ($v' = -0.1e^{-0.1t}$ by the chain rule): $f'(t) = 1 \cdot e^{-0.1t} + t \cdot (-0.1 e^{-0.1t}) = e^{-0.1t}(1 - 0.1t)$. Setting $f'(t)=0$ gives $t=10$ — the signal's strength peaks at day 10, after which the exponential decay dominates the linear growth and P&L starts falling. Finding that peak took differentiating a product and solving for where the rate of change crosses zero, without ever plotting the function.

Adjust the exponent and watch how the curve's steepness (its derivative) changes at each point — the tangent line's slope at any point is exactly what $f'(x)$ computes there.

What this means in practice

Every "Greek" in options trading — delta, gamma, vega, theta — is a derivative of the option price with respect to some input (underlying price, volatility, time). Bond duration and convexity are the first and second derivatives of price with respect to yield. Anywhere a quant asks "how sensitive is this to a small change in that," the answer is a derivative, computed using exactly these rules chained together for however complicated the pricing formula is.