Research-first quantitative trading system
Overview
Quant Labs is not a trading bot. It is a strategy factory designed to automate the full research pipeline: idea generation, rigorous backtesting, robustness validation, experiment tracking, and live signal generation.
The system has been fed 6 years of historical crypto data and trained through extensive backtesting. It continuously scans cryptocurrency markets, ranks trading opportunities, and posts signals to the Telegram channel when criteria are met.
Signal Bot
The automated signal bot (@quantlabs_signals) operates in a continuous loop. It scans for signals, validates against trained models, posts to Telegram when entry criteria matches, tracks results, and self-learns from every trade outcome.
6 Years of Historical Data
The system processes 6 years of historical crypto data across 50+ trading pairs. This massive dataset is used to train and validate strategies, ensuring signals are grounded in proven historical performance.
Real-Time Market Scanning
Connects to Binance WebSocket API for real-time market data. Scans across multiple timeframes (1h, 4h, 1d) and applies regime-aware filtering to only post signals when market conditions are favorable.
Signal Ranking
Signals are ranked from 1-10 based on: historical win rate on similar setups, current market regime alignment, volume confirmation, and cross-timeframe confluence. Only signals meeting minimum threshold are posted.
Telegram Signals
Each signal includes: Entry price, Stop Loss, Take Profit levels, Position size as percentage of portfolio, Confidence score (1-10), Timeframe, and time until signal expiration.
Auto-Tracking
The system autonomously tracks and reports all signal outcomes:
- -OUT Markers
When invalidation conditions are met (price crosses stop loss, signal expires, or market regime shifts), the system automatically marks the signal as OUT with the loss percentage.
- -WIN/LOSE Results
When a trade completes (hit TP or stopped out), the system posts the result autonomously with full trade breakdown including hold time, entry/exit prices, and percentage gain/loss.
- -Self-Learning
After each trade, the outcome feeds back into the training model. The system continuously refines its strategy rankings based on actual results, improving signal quality over time.
Research Engine
Strategy Factory
JSON-based strategy definitions with structured entry/exit rules. Strategies are versioned with lineage tracking. The optimization engine mutates parameters and combines successful traits through genetic algorithms.
Backtesting Engine
Event-driven simulation with realistic trade execution. Models fees, slippage, partial fills, and latency. Walk-forward validation ensures strategies generalize to unseen data.
Regime Detection
Markets are classified as trending (ADX greater than 25), ranging (ADX less than 25), or volatile (rolling volatility above threshold). Signals are only posted when regime aligns with strategy conditions.
ML Integration
Python-based research layer for data analysis and ML models. TypeScript execution core for fast, deterministic signal generation. TensorFlow integration for future predictive models.
Strategy Evaluation
Strategies are scored using normalized metrics with weighted scoring:
Hard filters before promotion:
- - Minimum 30 trades
- - Maximum 25% drawdown
- - Positive profit factor