Academy — How to Read Crypto Derivatives Data
Learn open interest, funding, liquidations and order flow: what each number means, how to read it, and the trap that makes it lie. Plus why this data is built for bots and AI agents.
The one idea
Price tells you what already happened. Derivatives data tells you who is exposed, which is what decides what happens next. Open interest counts positions, funding is the price of holding them, liquidations are where they break, and microstructure is how much size the book can absorb right now.
Open interest and positioning regime
Open interest counts positions, not trades, so it separates a rally new money is buying from one that is only shorts covering. Price up with OI up is new longs; price up with OI down is short covering; price down with OI up is new shorts; price down with OI down is a long flush.
Funding
Funding keeps a perpetual tethered to spot. The settlement cycle belongs to the contract, not the exchange: on Bitget, Gate, BingX and KuCoin many perps settle every 4 hours and some hourly. Annualizing everything as 8h understates a 4h contract by half and an hourly one eightfold.
Liquidations and cascades
A liquidation is forced, price-insensitive closing, which is why it clusters. Cascades group the raw feed into events with a start, a duration, a size relative to that coin's own normal minute, the price move during it and what price did in the hour after.
Order flow and microstructure
Tick-level footprint, CVD split by trade size, and cross-venue spread, top-of-book depth and basis.
For bots and AI agents
Every number rendered on these pages is reachable as structured JSON, and the analytics are computed server-side, so an agent receives the conclusion rather than a pile of ticks. MCP server at /mcp, REST at /api/v1, WebSocket at /ws, machine-readable index at /llms.txt. Tools include get_positioning_regime, get_liquidation_cascades, get_funding_dispersion, get_microstructure and get_market_brief.
What we do not claim
Liquidation maps are models, not measurements: exchanges do not publish where open positions sit. We do not model slippage, because the public book streams carry only the innermost levels. Coverage is finite and every board shows how many venues contributed. None of it is financial advice.