HFT Quantitative Researcher
Finds very short-horizon signals in order-book data and designs quoting policies that survive fees and adverse selection.
The job
An HFT quantitative researcher finds very short-horizon signals in order-book and trade data, and designs the quoting or taking policies that use them. The desk’s P&L is spread captured and short-term prediction, minus fees, adverse selection and inventory risk. The job is a signal that survives those costs, scored with fill assumptions that match the exchange. Things go wrong when a model predicts well but its edge is smaller than the fee, and when a backtest assumes passive orders fill more often than they do.
- Supports
- High-frequency trading and market-making desks
- Stages
- 5 to learn, then proof of work
- Two routes
- Free resources, or SelfMadeQuant programmes and projects
The roadmap
The same stages either way. The free route uses university courses, lectures, books and papers that are free online. The SelfMadeQuant route maps each stage to our programmes and ends in projects on real market data.
Stage 01 · Foundations
Probability for trading
Most market-making interview questions are probability questions.
- Expected value puzzles
- Markov chains
- Poisson processes
- MIT 6.262 Discrete Stochastic Processes (opens in a new tab)
MIT OpenCourseWare (Robert Gallager)
Lectures on Poisson processes, Markov chains and renewal theory used to model order arrivals.
- Poisson Processes (Chapter 9) (opens in a new tab)
David Pollard / Yale University
Short notes on the Poisson process properties behind arrival and queue models.
- Quant Trading Guide (opens in a new tab)
Callum McDougall
Probability puzzles and mental arithmetic in the style of trading firm interviews.
Stage 02 · Core
Market microstructure
How a limit order book works and who you trade against.
- Limit order books
- Spread and adverse selection
- Kyle and Glosten–Milgrom
- Market Microstructure: A Survey of Microfoundations, Empirical Results and Policy Implications (opens in a new tab)
Bruno Biais, Larry Glosten and Chester Spatt
Reviews Kyle, Glosten-Milgrom and inventory models alongside the evidence for each.
- Limit Order Books (opens in a new tab)
Gould, Porter, Williams, McDonald, Fenn and Howison / arXiv
Survey of order book mechanics, empirical regularities and the main modelling approaches.
- How markets slowly digest changes in supply and demand (opens in a new tab)
Jean-Philippe Bouchaud, J. Doyne Farmer and Fabrizio Lillo / arXiv
Explains price impact, persistent order flow and why liquidity is revealed only gradually.
Stage 03 · Core
Market making and execution
Quoting around inventory and scheduling large orders.
- Avellaneda–Stoikov
- Inventory risk
- Almgren–Chriss
- Dealing with the Inventory Risk: A solution to the market making problem (opens in a new tab)
Guéant, Lehalle and Fernandez-Tapia / arXiv
Closed-form quotes under inventory limits, extending Avellaneda-Stoikov.
- Optimal Execution of Portfolio Transactions (opens in a new tab)
Robert Almgren and Neil Chriss
The trade-off between impact cost and timing risk that defines an execution schedule.
- Algorithmic and High-Frequency Trading (opens in a new tab)
Álvaro Cartea, Sebastian Jaimungal and José Penalva / Cambridge University Press
The standard text on optimal market making, inventory control and optimal execution.
Stage 04 · Specialist
High-frequency econometrics
Why trade prices bounce between bid and ask, and measuring volatility in noisy ticks.
- Bid–ask bounce
- Roll model
- Realised measures
- A Tale of Two Time Scales: Determining Integrated Volatility With Noisy High-Frequency Data (opens in a new tab)
Lan Zhang, Per Mykland and Yacine Aït-Sahalia / Princeton
Why naive realised variance breaks under noise and how subsampling corrects it.
- Designing realised kernels to measure the ex-post variation of equity prices in the presence of noise (opens in a new tab)
Barndorff-Nielsen, Hansen, Lunde and Shephard
Kernel estimators that keep realised variance consistent when quotes contain microstructure noise.
- Empirical Market Microstructure (opens in a new tab)
Joel Hasbrouck / NYU Stern
Covers the Roll model, bid-ask bounce and price discovery with estimation code.
Stage 05 · Specialist
Machine learning on order-book data
Features, models and calibration for signals measured in milliseconds.
- Order-flow imbalance
- Tree models
- Probability calibration
- Fill-aware backtests
- The Price Impact of Order Book Events (opens in a new tab)
Rama Cont, Arseniy Kukanov and Sasha Stoikov / arXiv
Defines order flow imbalance and shows its linear link to short-horizon price changes.
- DeepLOB: Deep Convolutional Neural Networks for Limit Order Books (opens in a new tab)
Zihao Zhang, Stefan Zohren and Stephen Roberts
A benchmark deep model for predicting mid-price moves from raw order book snapshots.
- Predicting Good Probabilities With Supervised Learning (opens in a new tab)
Alexandru Niculescu-Mizil and Rich Caruana / Cornell
Shows boosted trees give distorted probabilities and how Platt and isotonic scaling fix them.
- Probability calibration (opens in a new tab)
scikit-learn documentation
Reliability diagrams and calibration methods to apply to classifier outputs before sizing trades.
Stage 06 · Proof of work
Order flow imbalance signal on crypto order books
Record level 2 order book updates for a liquid crypto pair from a public exchange feed, compute order flow imbalance and fit both a linear model and a gradient-boosted classifier for the next mid-price move. Answer how much predictive power survives after the spread is paid and whether the model's probabilities are well calibrated.
Stage 01 · Foundations
Probability for trading
Most market-making interview questions are probability questions.
- Expected value puzzles
- Markov chains
- Poisson processes
- Probability, Statistics and Financial Data
- Discrete Random Variables
- Continuous Random Variables
- Limit Theorems
Stage 02 · Core
Market microstructure
How a limit order book works and who you trade against.
- Limit order books
- Spread and adverse selection
- Kyle and Glosten–Milgrom
- Market Microstructure and Execution
- Market Data and Trading State
- Liquidity, Spread and Impact
Stage 03 · Core
Market making and execution
Quoting around inventory and scheduling large orders.
- Avellaneda–Stoikov
- Inventory risk
- Almgren–Chriss
- Market Microstructure and Execution
- Orders and Execution Schedules
- Post-Trade Analysis and Monitoring
Stage 04 · Specialist
High-frequency econometrics
Why trade prices bounce between bid and ask, and measuring volatility in noisy ticks.
- Bid–ask bounce
- Roll model
- Realised measures
- Econometrics and Financial Time Series
- Univariate Time Series: ARMA and Box–Jenkins
- Volatility Estimation Theory
Stage 05 · Specialist
Machine learning on order-book data
Features, models and calibration for signals measured in milliseconds.
- Order-flow imbalance
- Tree models
- Probability calibration
- Fill-aware backtests
- Financial Machine Learning
- Financial labels, weights and memory
- Nonparametric and Tree-Based Methods
- Importance, calibration and sizing
- Quantitative Research, Data and Backtesting
- Labels and Validation Design
- Economic Backtest Simulation
Stage 06 · Proof of work
Build the projects
Each project is built on real market data and reports what the data said, including when the idea failed.
- Q8Order-Book and Slippage CalculatorA 0.5 BTC market order walks 12 levels for 0.66 bp of impact, while the taker fee is 4–60 bp.
- MMInventory-aware market-making researchEvery quoting policy lost money at a zero maker fee because of adverse selection; a 1 bp rebate would turn almost all of them positive.
- OBMachine-Learning Signals on Limit-Order-Book DataGradient boosting beat the linear baseline on every day, but kept 0.46 bp per round trip against 8 bp of assumed fees.
Practise the interview
Questions this role gets asked, each with a hint and a worked solution.
- A new model lowers prediction error but produces lower net returns. How do you investigate?
- A classifier is well calibrated overall but you only trade when it says above 0.8. What do you check?
- A painting is worth an unknown amount up to 100,000 and the seller knows more than you. What should you bid?
Start the SelfMadeQuant route free: a free account opens one complete topic from each primer and the interview question bank. No card needed.
Start free