Order flow imbalance

Order flow imbalance (OFI) measures the net pressure hitting the book: aggressive buys minus aggressive sells over a short window. The intuition is simple — when liquidity is being consumed far more on one side, the mid-price tends to move that way before it fully reprices.

I built an OFI signal engine in Rust for prediction-market trading, consuming exchange WebSocket trade streams for a handful of assets and computing imbalance in real time. The hard parts were never the maths:

• Latency budget. A signal that's right but late is a loss. The engine is a multi-task async runtime so feeds, signal, and execution don't block each other.
• Edge decays. OFI is a fleeting effect; size it with Kelly criterion discipline and respect that the edge can vanish between backtest and live.
• Prove it first. Forward-test against live quotes before risking anything — validate before automating.

A signal is a hypothesis about expected value. Treat it like one.