Researcher course in high-frequency trading
High-Frequency Trading
The models behind high-frequency strategies: optimal market making with inventory risk, adverse selection, short-horizon signals from the book, latency arbitrage, and how to test strategies whose results depend on queue position.
High-frequency trading, studied seriously, is a branch of applied probability. A market maker posts prices on both sides, earns the spread when both fill, and loses when the price moves while it holds inventory or when it trades with someone who knew more. The course builds the models that balance those forces.
How does an HFT market maker set its quotes?
In the Avellaneda-Stoikov model, the market maker shifts both quotes away from the mid price in the direction that reduces its inventory, and widens or narrows the spread according to volatility, risk aversion and how quickly orders arrive. The result is a reservation price and an optimal spread with closed-form approximations. You will derive them, implement them, and test them in a simulator that respects queue position.
Why can’t HFT strategies be tested with an ordinary backtest?
Because their profit depends on whether a limit order would have filled, and that depends on where it would have sat in the queue and what other traders did. A backtest that fills every order touched by the price overstates profit badly. The course builds a queue-aware simulation and shows the size of the difference.
Who is this course for?
Quantitative researchers and developers working in or moving into market making and HFT. It is the capstone course of the high-frequency track and counts toward the Certificate in High-Frequency Trading.
What you will be able to do
- Derive and implement the Avellaneda-Stoikov market making model.
- Model inventory risk and adverse selection and their effect on quotes.
- Build short-horizon signals from order book features and measure their decay.
- Describe latency arbitrage and the economics of the speed race.
- Design a simulation that models queue position and fill probability.
Syllabus, module by module
The HFT landscape
Strategy types, firm economics, and the role of speed in each.
Market making
Quoting both sides, the spread as compensation, and inventory as the main risk.
Optimal market making
The Avellaneda-Stoikov model: reservation price, optimal spread and their derivation.
Adverse selection
Informed flow, toxicity measures and quoting when the book is about to move.
Short-horizon signals
Imbalance, microprice, trade flow and cross-asset leads; decay and half-life.
Latency arbitrage
Stale quotes, cross-venue races, and the Budish-Cramton-Shim critique of continuous trading.
Simulation
Queue-aware simulators, fill models, and why simple backtests fail for HFT.
Risk and controls
Position and loss limits, fat-finger checks and exchange rules.
Regulation
Market abuse rules, spoofing and layering, and algorithmic trading obligations.
Assessment preparation
Supervised work on the written assessment.
How the certificate is earned
A written paper-style assessment with a simulation component: you extend a market making model with a signal or constraint of your choice, test it in a queue-aware simulator, and write up the derivation and results. On a pass you receive the Course certificate in High-Frequency Trading, issued by the Institute with a verification code. Read how certification works.
Before you start
Stochastic calculus to the level of Ito's lemma, the material of Limit Order Books, and working Python.
This course counts toward
Questions applicants ask
Can you learn high frequency trading online?
You can learn the models, the microstructure and the testing methods, which is what this course teaches. Running a competitive HFT business also needs infrastructure and capital that no course provides.
What maths does high frequency trading use?
Probability, stochastic processes and stochastic control for market making, and statistics for short-horizon signals. The course assumes stochastic calculus to the level of Ito’s lemma.
Does the course include coding?
Yes. Models are implemented and tested in Python. Low-Latency C++ for Trading Systems covers the engineering side.