Quantitative Trading

• One weekly meeting where every project in progress is discussed and gets feedback. It sets the goals for the week ahead and keeps the research pointed in the right direction.
• Members meet again separately during the week to move a project forward together, which is where most of the actual modelling happens.
• Occasional meetings with Strategy Application, when a strategy has survived validation and is ready to be taken live.

• Quantitative portfolios and algorithmic strategies, handed to Strategy Application to run live.
• The HRP strategy, published with its live metrics on this site.
• Research articles for the TIC Journal documenting the models behind them.

Onboarding begins with a detailed explanation of the projects currently in progress. You then join one, study the material it requires and take on weekly tasks with support from the rest of the team. Those tasks are part of the project itself, not a separate exercise.

Most quantitative research ends in a discarded idea, because it did not beat the baseline. That makes genuine interest in markets and plain resilience the things that matter most, alongside creativity: the unusual approach is usually what moves a project. Maths, physics, programming and market knowledge all help, but none is required before joining.

The team develops and validates quantitative strategies. A project can use any family of models as long as its foundation is mathematical. Portfolio Optimisation does not put the strategy into production; that work belongs to Strategy Application.

vs Machine Learning →

Machine Learning builds every strategy with machine learning and studies those methods in greater depth. Portfolio Optimisation can use machine learning when it fits, but it can also choose another mathematical approach.

vs Strategy Application →

Portfolio Optimisation develops and validates the strategy. Strategy Application receives it afterwards and builds the infrastructure needed to run it live.

vs Risk →

Both teams research portfolio models and work in Python. Risk measures and controls portfolios managed by the fund teams. Portfolio Optimisation constructs and validates the allocation or strategy itself.

A project starts with a possible market inefficiency. The team then collects and transforms historical data, builds the model, optimises it and validates it with stress testing. Most ideas are rejected during validation, which is the purpose of that stage.

Two projects are currently running. One is an improvement to the Black-Litterman framework. Its MVP was released so the team could collect feedback, find bugs and decide which features are worth adding. The other project is building a mean-reverting basket of financial instruments and trading it with mean-reversion strategies based on a Kalman filter.

• The week opens with a follow-up meeting where each member presents the state of their project and the next objectives are set.
• During the week members study books and papers, write code, train models, analyse experimental results and review each other's work.
• Research is part of the job rather than preparation for it: it is how the team decides whether a model still deserves to be in a project.

• Predictive models built on data processing pipelines.
• Content for the TIC website and articles for the TIC Journal, based on the work.
• Entries in international machine learning competitions applied to finance.
• Events about the use of machine learning in finance.

New members normally join an existing project. Your first tasks are to understand the current code and, when useful, explain your reading of the models to the team so that everyone agrees on the project's status. You should be contributing to the project within your first few weeks.

This team suits students who enjoy programming, mathematics, machine learning and problem solving. Members are expected to work autonomously, question their results, learn quickly and pay attention to detail. Previous finance experience is not required; willingness to learn and contribute matters more.

Machine learning is the basis of every project in this team, not one possible method among several. Members spend the year on technical finance projects, write code used by the team and take part in international competitions.

vs Portfolio Optimisation →

Portfolio Optimisation draws on the wider quantitative literature and uses machine learning when it fits the problem. Machine Learning uses it in every strategy and studies it in greater depth.

vs Strategy Application →

Machine Learning builds and validates the model. Strategy Application takes the completed model and builds the live trading infrastructure around it.

Every project uses Python, Git and code review. The rest of the stack depends on the problem: it may involve PyTorch, TensorFlow or scikit-learn. Members learn the tools required by their project instead of following one fixed list.

• A deployment starts with scripts from Portfolio Optimisation or Machine Learning. The team connects the required data sources, integrates the strategy into the trading engine and maps the model output to specific buy and sell orders.
• Engine work may involve fixing a parsing bug that makes a ticker unavailable for trading or adding a feature such as the API integration for a new broker.
• Another member reviews every change before it reaches production.

• End-to-end algorithmic trading engines, running on the club's own servers.
• The live pipeline behind the strategy performance published on this site.
• Open source code: every engine is public on GitHub, including the low-latency C++ one still in development.

You contribute to the codebase from the start, beginning with smaller bugs and features that get you familiar with the system. Feedback is immediate: another member reviews your code before it ships, which is how the standard gets passed on rather than written down.

This team is for people who enjoy programming and want to see how their code behaves once it is running unattended against a real broker. An interest in trading helps, but previous finance knowledge matters far less than caring about performance, reliability and problem solving. It ships more code into production than any other team in the club, which is why the bar on code quality sits higher.

Strategy Application handles the production side of Quant. The team does not design the strategies. It receives validated research and builds systems that run on schedule, recover safely and place paper-trading orders without manual intervention.

vs Portfolio Optimisation →

Portfolio Optimisation develops and validates the strategy. Strategy Application takes the script, connects it to live data and a broker, and makes it run inside the trading engine.

vs Machine Learning →

Machine Learning trains the models. Strategy Application builds the engine that runs their output as scheduled orders without supervision.

vs Risk →

Both teams build production systems around portfolio data. Risk produces controls and recommendations for a fund manager. Strategy Application automatically executes the orders from a validated strategy once it is live.

One Python engine is in production and a second is in development. The team is also building a C++ engine for low-latency, high-frequency trading. This project involves demanding low-level optimisation. Once the engine is ready, the team plans to write its own high-frequency strategies so that members work on both the infrastructure and the strategy.