Risk-neutral pricing in three sentences

"Risk-neutral pricing" is the most intimidating phrase in derivatives. It sounds like it requires belief in a parallel universe where investors don't care about risk. It doesn't. Here's the entire idea in three sentences:

1. Nobody trusts your forecast of where the underlying will be in a year.
2. But everybody agrees on what it costs today to replicate the option payoff using stocks and bonds.
3. So we price the option as the cost of its replication, not as an expectation of what it might pay out.

That's it. The "risk-neutral" part is just the mathematical machinery that makes step 3 work cleanly.

Why it's called "risk-neutral"

When you do the algebra of replication carefully, the answer comes out as if you computed an expected payoff in a fictitious world where:

• Every asset earns exactly the risk-free rate (no risk premium).
• Investors don't demand any extra return for taking risk (= they're "risk-neutral").

That world doesn't exist. Real investors are risk-averse and they expect a premium for taking on volatility. But the option price doesn't care what investors expect — it only cares about the cost of replication, which comes out the same as if you priced under the risk-neutral assumption.

Said differently: the risk-neutral measure is a computational device, not a description of reality. It gets you the right answer because the real-world drift cancels out in the replication argument.

Why this is the whole game

Without replication, derivatives pricing would be a forecasting problem: every trader would price options based on their personal view of the underlying. There would be no agreed price. Buy and sell would never meet.

With replication, the disagreement collapses. Two traders can hold opposite views on the underlying, and yet agree on the option price — because they both agree on what it costs to dynamically hedge it. The option becomes priced off a quantity (the cost of hedging) that's observable and objective, not off a quantity (your forecast) that's subjective and unfalsifiable.

That's the breakthrough Black, Scholes and Merton made in 1973. Before that, no consistent way to price options. After that, an entire industry of derivatives pricing built on top of one idea: price by replication, not by forecast.

The catch

Replication only works when you can actually replicate. When the market is complete— every payoff can be hedged with stocks and bonds — there's a unique replication and a unique price. When markets are incomplete(stochastic vol, jumps, illiquid hedges, transaction costs), there are many possible "risk-neutral measures" — and prices depend on which one you pick. That's where preferences sneak back in.

Black-Scholes assumes complete markets — hence its unique price. Heston doesn't — hence the calibration choices. LSV calibrates to the vanilla smile and then makes a specific choice of risk premium for vol-of-vol. Every time you see a model with "market price of risk" or "variance risk premium" in the parameters, that's where the residual preference choice lives.