Algorithmic Market Making

What is Algorithmic Market Making?

Market making is the business of providing liquidity by simultaneously quoting bid and ask prices, earning the spread on each round-trip trade. Algorithmic market making automates this process, managing thousands of quotes across multiple securities with microsecond latency. Market makers account for 50%+ of US equity volume and are essential for efficient price discovery.

The core economics: spread capture minus adverse selection minus inventory costs equals profit. A typical equity market maker captures 1–2 basis points per trade but faces adverse selection of 0.5–1.5 bps (informed traders earn 50–70% of the spread). The profit margin is thin but scales with volume — a market maker earning 0.5 bps net on 100M shares/day generates $500K daily P&L.

The Avellaneda-Stoikov model provides the theoretical foundation for optimal quotes under inventory risk. The key insight: quotes should skew toward reducing inventory. If holding 1000 shares long, lower the bid to discourage buying and raise the ask to encourage selling. The skew magnitude depends on inventory size, volatility, and time horizon.

Mathematical Foundation

Optimal Bid-Ask Spread (Avellaneda-Stoikov):

Inventory Skew:

Where:

• — risk aversion
• — volatility
• — inventory position
• — time to close
• — order arrival intensity

P&L Attribution:

Model Architecture

Performance Results

Real-World Case Study

Citadel Securities processes 26% of US equity volume (500M+ annually), co-location at 40+ exchanges, and ML-driven signal generation. Key insight: the majority of market making profit comes from adverse selection management — knowing which orders are from informed traders and adjusting quotes accordingly. Their inventory management system maintains <1000 shares average position per stock while quoting 50+ bps spreads.

Deployment

Common Pitfalls

1. Adverse selection: Informed traders exploit stale quotes — use queue position modeling
2. Inventory accumulation: Holding too much inventory during volatility — implement hard limits
3. Spread war: Competing market makers widen spreads during stress — monitor competitor quotes
4. Regulatory risk: Minimum quote lifetime rules limit flexibility — implement compliance checks
5. Technology risk: Latency spikes cause missed fills — use lock-free data structures

Summary with Key Takeaways

This project built an Avellaneda-Stoikov market making system achieving $1,400 daily P&L with 65% fill rate and 18% adverse selection. The inventory management system maintains positions within risk limits while capturing 5.2 bps average spread. Key principles: adverse selection management is the primary profit driver; inventory skew must adapt to volatility; and spread width should reflect competition and order flow toxicity.