AlphaBook
A live exchange in your browser: real order book, real market data, structured trading games. Open-source, built at Oxford, free for any student, society, or classroom.
alphabook.uk — sign up and trade in under a minute.
Why
The skills trading firms actually test — market making, probability calibration, pricing under pressure — are taught almost nowhere outside the firms themselves. AlphaBook makes them trainable: a real price-time-priority matching engine, live market data, and games modelled on trading-floor exercises, all with virtual balances and zero real-money risk.
We use it to run live trading bootcamps and competitions for the Oxford student quant community, including the Oxford Alpha Fund: an instructor creates a game, scripts the news flow, and a room of students competes on a shared order book with a real-time P&L leaderboard.
The platform
- Market Simulation — trade AAPL, MSFT, NVDA, AMZN, GOOGL, META, and TSLA on a live limit order book anchored to real prices; an automated market-maker bot keeps the book liquid
- Market Simulation Coding — write a trading bot, run it on your own machine, and trade four synthetic items over a rate-limited order API against everyone else's bots for ten minutes; no code runs on the server, 1000-lot position limits, P&L marked at hidden fair value. Everything your bot earns becomes credits on a shared map, where you build an empire and fight over territory
- 5Os — estimate the five statistics (min, quartiles, median, max) of a hidden hand of cards; best calibration wins
- Headline Trading — news hits the wire, prices move; trade a futures market on the story before time runs out
- Poker Auction — sealed-bid second-price auctions for cards, then trade hands in a post-auction market
- Mental Math — timed arithmetic drills with configurable types and difficulty
Every mode has a live leaderboard. Works solo too — the market-maker bot is always on the other side.
Roadmap
We are seeking grant funding to turn the simulator into an adaptive teaching platform:
- LLM-driven agents — market-maker bots that adapt their quoting to the student's skill level
- AI strategy coach — post-session review of a student's fills: what a professional would have done differently
- Session replay — full order-flow recording so instructors can dissect a session tick by tick
Team
Xue (Snow) Yan (project lead) · Siyu (Sylvia) Li · Nick Wang · Fiona (Yiran) Zhang · Xinyun Jiang · Yiluan (Eylan) Zeng
Architecture
| Layer | Technology |
|---|
| Backend | Python 3.11 / FastAPI / Uvicorn |
| Frontend | Jinja2 + vanilla JavaScript |
| Data & auth | Cloud Firestore (async) + Firebase Auth |
| Real-time | WebSockets (book snapshot broadcast on every trade) |
| Market data | Stooq quotes + synthetic tick engine |
| Deploy | Docker → Cloud Run + Firebase Hosting |
Key modules in app/: order_book.py (matching engine: price-time priority, partial fills), market_maker.py (two-sided quoting with realistic depth and delayed sweeps), market_data.py (real quotes blended with synthetic ticks), fiveos.py / headline.py / poker_auction.py / mental_math.py (game engines), admin.py (games, news, leaderboard, users). Tests live in tests/; CI runs ruff + pytest on every push.
Market Simulation Coding (algo_engine.py + market_sim_py.py + world.py) is client-side: players run their bot on their own machine and reach the market only through an authenticated, rate-limited order gateway (algo_ratelimit.py caps it at ~10 orders/sec). No player code runs on the server, so there is nothing to sandbox — the engine only validates and matches orders on arrival. The same is true of the world layer (world.py), where an empire action is validated JSON just like an order; it holds the map, the buildings, the units and the combat, and is kept pure and seeded so the whole ruleset is testable without standing up a market. A player's trading P&L is their build budget, which is the only coupling between the two halves. Both keep their own books and fictional items, separate from the live stock market. Full write-up — architecture, the API, and a strategy-writing guide with worked examples — in docs/market_sim_py.md.
Run it locally
Prerequisites: Python 3.11+, a free Firebase project (Auth + Firestore).
git clone https://github.com/snowy615/alphaBook.git
cd alphaBook
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env # fill in Firebase config — see SETUP.md
python -m uvicorn app.main:app --reload
Open http://localhost:8000. Full Firebase setup and Cloud Run deployment: SETUP.md.
Contributing & license
PRs welcome — see CONTRIBUTING.md. MIT licensed (LICENSE).
