Avellaneda-Stoikov Market Making with Order Flow Imbalance (OFI) Signal

How It Works The Avellaneda-Stoikov model is a classical market making framework that computes optimal bid/ask quotes as a function of current inventory, residual time, and volatility, targeting a reservation price that accounts for inventory risk. The OFI (Order Flow Imbalance) signal measures the imbalance between buy and sell order flow in the orderbook to predict short-term price direction, allowing the MM to skew quotes and reduce adverse selection.

Entry/Exit Conditions The bot continuously posts two-sided quotes (bid and ask) around the A-S reservation price. When OFI is positive (more buy pressure), the bot shifts quotes upward to avoid accumulating short inventory into rising prices, and vice versa. Quotes are refreshed continuously as market conditions change. Inventory limits define maximum position size before the bot stops quoting on one side.

Verified Results The repo has 9 stars and 1 fork, indicating some real-world interest and review. The combination of A-S and OFI is a well-documented academic approach applied to crypto and prediction markets. No specific backtested PnL is mentioned in the description.

Implementation Notes The bot requires real-time access to Polymarket's CLOB (Central Limit Order Book) API for orderbook data and order placement. The A-S model requires calibrated volatility and inventory risk-aversion parameters, which must be tuned for each market's liquidity profile. OFI computation requires tick-level order event data.

Risks Prediction market events can gap to 0 or 1 instantly upon news, creating catastrophic inventory losses that the A-S model does not anticipate. Thin liquidity on many Polymarket markets makes it difficult to exit inventory positions. The OFI signal can be gamed by other participants placing and canceling large orders. Parameter miscalibration can cause the bot to systematically quote on the wrong side.