Positions force-closed in this coin over the last 24 hours, split by side. A long liquidation is forced selling; a short liquidation is forced buying.

A liquidation is a trader being forced out of a bet because the price moved too far against them. Longs being liquidated means forced selling; shorts being liquidated means forced buying. Judge the size against what is normal for this coin: $5M is enormous for a small coin and barely noticeable for Bitcoin.

Liquidation-cascade fades are built on this: forced selling overshoots, so the sharpest bounces tend to follow the biggest one-sided prints. Long and short squeeze plays work the same mechanism in reverse, and stop-hunt reversals look for price being pushed into a cluster of stops precisely to trigger them.

Design a new strategy for AGI that trades around forced liquidations, using /api/v1/market-intelligence/liquidations?symbol=AGI live. Define what counts as a cascade, which side you take, how quickly you exit, and how you avoid being liquidated yourself. Validate on the per-event tape at /api/v1/backtesting/hl-liquidations plus /api/v1/backtesting/klines, and report the worst single loss, not just the average.

/backtesting/liquidations gives the aggregated series, and /backtesting/hl-liquidations gives the exact per-event Hyperliquid tape — every individual fill, with its venue. For the price levels where liquidations cluster, use /backtesting/snapshots?data_type=liquidation_map.

Using CryptoDataAPI, backtest the liquidation-cascade fade on AGI. Pull the per-event tape from /api/v1/backtesting/hl-liquidations and price bars from /api/v1/backtesting/klines. Define a cascade as a 5-minute bucket in the top 1% of one-sided liquidation size, then measure forward returns at 15m, 1h and 4h from taking the opposite side. Charge taker fees and report the hit rate and worst loss.