Pairs Trading Project

Team Members: Yunkai Gao, Jiajun Huang

What is Pairs Trading?

Pairs trading is a market-neutral strategy that matches a long position with a short position in two historically correlated stocks. It profits from temporary mispricings while being relatively hedged against market movements.

Why Choose Pairs Trading?

  • Market neutral
  • Statistical arbitrage
  • Diversified and controllable risk
  • Proven historical profitability

Pair Stock Selection

  • Same industry:
    • Similar fundamentals, Business models, Financial profiles
  • Statistical method: Suppose P denotes the closing price of stock i at time t, where formation period t = 0,1,…,T
    • Calculate standardized prices for stocks
    • Calculate SSD of the standardized prices for stocks X and Y
    • Select pairs with minimum SSD

Setting Trading Rules

  • Calculate 𝜇 and 𝜎
    • Compute the mean and standard deviation of the standardized stock price spread during the formation period
  • Open position
    • Standardized price spread between the two stocks < 𝜇−1.5𝜎 or > 𝜇+1.5𝜎
  • Close position
    • If < 𝜇−1.5𝜎 , close position when the Standardized price spread back to 𝜇−0.2𝜎
    • If > 𝜇+1.5𝜎 , close position when the Standardized price spread back to 𝜇+0.2𝜎
  • Stop-loss line
    • If standardized price spread < 𝜇−2.5𝜎 or > 𝜇+2.5𝜎 after open position, we close position to stop loss.
  • Draw Standardized Price Spread Series Chart
    • Normalized spreads, averages, opening lines, closing lines, and stop-loss lines.

Refinements and Advanced Methods

  • Pair Identifier
    • Clustering: K-means, how to select hyperparam
    • GNN: Stocks as nodes, pair relation as line, sector as subgraph
    • Attention combined
  • Spread Construction
    • Move beyond linear spread
    • Use multi-factor approaches (fundamentals, technical indicators)
  • Signal Generation
    • Reinforcement Learning (RL) for end-to-end decision making
    • Ensemble learning for robustness
  • Capital Allocation
    • Multi-objective optimization
  • Dynamic portfolio
    • Rerank to fix unstable Pair relation
    • Risk-contribution clear out