Order Book Imbalance (OBI) is one of the simplest live microstructure signals you can compute: how much resting buy volume sits at the top of the book compared to resting sell volume. It’s cheap to calculate and available the instant a WebSocket order book update arrives.
The formula
OBI over the top N price levels is (bid_size − ask_size) / (bid_size + ask_size), expressed as a percentage from -100% (entirely ask-heavy) to +100% (entirely bid-heavy). Hyperliquid’s public l2Book WebSocket channel sends a full order book snapshot on every update — no diff reconciliation needed — which makes computing this in real time straightforward.
What it actually tells you
A strongly bid-heavy book means more resting buy interest than sell interest at the top of book right now. That’s a real, measurable fact about current order flow — not a prediction. It shifts constantly as orders are placed and pulled, sometimes within single-digit seconds on liquid assets.
Where it breaks down as a signal
OBI on its own is easy to distort. A single large resting order — genuine or a spoof pulled before it would ever fill — can swing the imbalance dramatically without reflecting real directional pressure. Serious use of OBI pairs it with other signals: trade flow, spread, depth beyond the top levels, and time-weighted persistence rather than a single snapshot read.
Summary
OBI is a genuinely useful, cheap-to-compute signal — but it’s one input, easily distorted by a single large order, and most useful when watched over time and combined with other order flow context rather than read as a standalone prediction.