Faray Traders - IMC Trading Bot

Inspiration

The IMC trading competition presented a unique challenge: trading financial derivatives whose values depend on real-world events like flight traffic at Munich Airport and water levels in the Isar River. We were inspired to build a sophisticated algorithmic trading system that could:

• Leverage real-time external data to gain an edge over pure market-making strategies
• Combine multiple data sources (flight data, water levels) to calculate fair values

The opportunity to integrate live flight data from Munich Airport with market-making strategies excited us, as it represented a real-world application of quantitative trading principles.

What it does

Our system is a hybrid algorithmic trading bot that combines market-making with fundamental valuation to trade financial derivatives. Here's what it does:

Valuation Engine (Value Box):

• Scrapes real-time water level and flow rate data from the HND Bayern website (95 data points)
• Fetches flight data from Munich Airport via Aerodatabox API
• Calculates fair values for different products:
  • Product 1: flow_rate × water_level (real-time)
  • Product 2: Range-based calculation using max/min water dynamics
  • Flight Products: Settlement = 3 × (Total Arrivals + Total Departures)

Flight Data Fetcher (Flight Catcher):

• Retrieves and processes flight arrivals/departures from Munich Airport
• Groups flights into 30-minute time slots for analysis
• Automatically handles API limitations (splits requests > 12 hours)
• Provides real-time settlement estimates for trading decisions

Trading Strategy:

• Market Making: Quotes both bid and ask prices simultaneously with configurable spreads
• Position-Based Skewing: Adjusts prices based on inventory (skews down when long, up when short)
• Settlement-Based Trading: Blends expected settlement (70% weight) with market price (30%) to determine fair value
• Risk Management: Position limits (±50) prevent losses

Settlement Price Estimation:

• Integrates flight data over market periods (typically 24 hours)
• Calculates settlement using the formula: 3 × (Arrivals + Departures)
• Updates every 5 minutes to reflect new flight data
• Falls back gracefully when API is unavailable

How we built it

Architecture: We built a modular Python system with clear separation of concerns:

1. ValuationEngine (valuation_engine.py): Central valuation system that aggregates data from multiple sources

   • Web scraping for water level data using BeautifulSoup
   • API integration for flight data via Aerodatabox

2. FlightDataFetcher (flight_data_fetcher.py): Specialized component for flight data processing

   • Handles complex API response parsing with multiple fallback strategies
   • Automatic time range splitting for periods exceeding API limits
   • Robust datetime parsing supporting multiple formats

3. Trading Bots:

   • FairValueBot (fair_value_bot.py): Pure market-making with position skewing
   • FlightDataBot (flight_data_bot.py): Market-making + settlement-based trading with background data refresh

4. Base Infrastructure (imcity_template.py): Provided trading framework with SSE streaming for real-time market data

Technical Implementation:

• Real-time Data Streaming: Server-Sent Events (SSE) for live orderbook updates
• Concurrent Processing: Thread-based architecture for simultaneous data fetching and trading
• Fair Value Calculation: fair_value = (0.7 × settlement) + (0.3 × market_price) - (position × skew_factor)
• Order Management: Cancels existing orders before placing new ones to avoid duplicates
• Error Handling: Graceful degradation when APIs are unavailable

Built With

• python