QuantDesk

QuantDesk is a quant research and paper-trading console built on Spring Boot. It backtests strategies on historical price data, pulls live market data and news sentiment, scans a US large-cap universe for momentum, and routes signals through a risk-checked trading engine — in simulation or against an IBKR paper account, never with real money. Backtests and all tests run fully offline against bundled CSV data.

SAFETY

QuantDesk never trades real money. The only trading modes are OFF (default — the engine does nothing), DRY_RUN (signals are logged, no orders are placed anywhere), and PAPER (orders go to a simulated broker or an IBKR paper account). There is no real-money mode in the codebase. Every order — paper or simulated — must first pass the risk manager (position and exposure caps) and the news sentiment veto, and the dashboard kill switch (POST /trading/kill) flips the engine to OFF instantly. Nothing in this project is investment advice.

Module flow

flowchart LR
    marketdata[marketdata] --> indicator[indicator]
    marketdata --> strategy[strategy]
    marketdata --> universe[universe scanner]
    indicator --> strategy
    strategy --> backtest[backtest]
    strategy --> engine[trading engine]
    universe --> engine
    news[news + sentiment] --> engine
    risk[risk] --> engine
    engine --> execution[execution: simulated / IBKR paper]
    backtest --> web[web + dashboard]
    engine --> web

Loading

The universe scanner ranks the whole universe by momentum and hands the top names to the trading engine, which sizes and risk-checks every order before it reaches the (paper-only) execution layer.

Features

• Pluggable Strategy interface — drop in a new strategy by implementing a single-method contract; the engine does not care how the signal is produced.
• Built-in strategies — SMA crossover, RSI, Momentum, and Buy-and-Hold.
• Universe scanner — cross-sectional momentum ranking over ~100 US large caps, feeding an automatic top-5 rotation portfolio.
• News sentiment — headlines are scored and negative sentiment vetoes buys.
• Performance metrics — total return, maximum drawdown, Sharpe ratio, Sortino ratio, and CAGR, plus trade count and win rate.
• Sample data included — bundled CSV price series so the engine and all tests run offline with no data setup.
• REST API + dashboard — backtests, scan results, positions, trade log, and trading controls as JSON and as a single-page dashboard.

Universe scanner & momentum strategy

The universe lives in data/universe.csv — roughly 100 US large caps as a plain CSV that you can edit to change what gets scanned.

Each scan computes cross-sectional momentum: every symbol's trailing return over a lookback of 63 trading days (about three months), ranked best to worst. The trading engine holds the top 5 and rebalances automatically: it sells positions that drop out of the top 5, buys new entrants, and targets equal weight per position. Every rebalance order still passes the news sentiment veto and the risk caps before execution.

Endpoints:

Endpoint	Purpose
GET /scan	Latest ranking (rank, symbol, momentum score, last price, data availability) and lastScanAt
POST /scan/run	Trigger a scan immediately
GET /backtest/portfolio	Backtest the cross-sectional momentum portfolio on historical data

An honest note on momentum. Cross-sectional momentum is one of the best-documented historical anomalies in equity markets (Jegadeesh & Titman, Returns to Buying Winners and Selling Losers, Journal of Finance, 1993). It has also suffered long droughts and violent crashes, and a documented historical edge is exactly that — historical. Past performance guarantees nothing about the future.

Prefer trading a fixed watchlist instead of the scanned universe? Set quantdesk.trading.universe-mode=false and the engine trades the configured watchlist as before.

Tech stack

Component	Version
Java	21
Spring Boot	3.3
Build tool	Maven

Run locally

mvn spring-boot:run

Open the dashboard at http://localhost:8080/ or run a backtest against the bundled sample series:

GET http://localhost:8080/backtest?symbol=SAMPLE&fast=10&slow=30

symbol selects the price series, and fast / slow set the SMA crossover periods. The response is a JSON document with the symbol, strategy name, and the computed performance metrics.

Roadmap

Shipped so far: the backtesting engine, pluggable live market data, LLM news sentiment, risk limits, the universe scanner with automatic momentum rotation, and paper-only execution (simulated broker and IBKR Client Portal paper gateway). Next up: richer portfolio analytics and more data providers.

Out of scope, permanently for now: live and real-money trading. Execution targets paper trading only and requires a broker paper account plus a compliance review before any orders — simulated or otherwise — are placed.