Skip to content

Methods

The mathematics is the course, not a preface.

We teach quantitative finance as a set of arguments you can reproduce: a price from a hedge, a risk number from a distribution, a code path from a formula. Fluency without the derivation is treated as unfinished.

A monitor showing a blurred risk heatmap and candlestick chart

Probability at a working depth

You will manipulate conditional expectation, change of measure in the binomial model and then in Black–Scholes, and say clearly when a result needs independence. We introduce the language of filtrations so a stochastic integral is not magic. We do not pretend this is a measure-theory course, and we will tell you when a proof has been sketched rather than closed.

Stochastic calculus as a tool

Itô’s lemma is derived, used, and then used again until the steps are boring. You apply it to geometric Brownian motion, to a variance process in a simple stochastic-volatility sketch, and to the hedging argument behind the Black–Scholes PDE. The point is not to collect formulae. The point is to notice which term is the hedge and which term is the financing.

Three numerical methods, compared on the same contract

Trees

Recombining binomial trees for European and American puts. You will see how many steps it takes before the price stops wandering, and you will write down the early-exercise decision instead of calling a library.

Grids

Finite differences on the Black–Scholes PDE. Explicit schemes that blow up, implicit schemes that do not, and boundary conditions you chose rather than inherited from a screenshot.

Simulation

Monte Carlo with a standard error you report beside the price. Antithetic and control variates, a biased discretisation you can measure, and a barrier you refuse to treat as continuous when your time step is not.

Overhead view of printed charts, a pen, a ruler, and coffee

Statistics that can be refused

Regression is taught with residual plots, leverage, and a habit of writing the claim in a sentence before fitting it. Volatility clustering is shown with a simple model so you can see persistence without worshipping a particular acronym. Factor models are estimated, then stressed by removing a factor and watching the portfolio move.

Risk is taught as a family of questions. What horizon? What confidence, and why is confidence the wrong word? What history did you choose, and what did that history never contain? A value-at-risk number without a backtest is marked as incomplete.

Python is part of the argument

Code is marked when it hides a unit error, a look-ahead, or a random seed you did not record. Style is marked only when another reader would fail to run the file. We care more about a documented seed and a saved figure than about a fashionable framework.

What we will not flatter

A calibrated model is not a true model. An implied volatility surface is a way of storing prices. A backtest that never trades through a cost is a story. We will ask you to say these sentences out loud in the viva, because they are the ones that slip when a meeting is going well.

Reading, not a shrine

Seminar notes are written by Adulateaeris and included in the fee. Where a classic paper or a chapter is worth your time, we name it and say which pages. You are not examined on brand loyalty to a textbook. Optional print books are your own cost if you want them on paper.