Liquidation Bursts

Busiest-hour share = 100 × recorded USD in the busiest hour ÷ recorded USD in the full window. Both numerator and denominator are recorded observations. Sampling can distort concentration; this is not a measure of all liquidations.

Binance, Bitget and Aster publish sampled liquidations, so their totals and the combined totals are lower bounds during cascades. Binance publishes only the latest liquidation per symbol per 1000 milliseconds, so its totals are a lower bound in busy moments.

• Binance liquidations sampled: Binance publishes only the latest liquidation per symbol per 1000 milliseconds.
• Bybit Every liquidation (allLiquidation stream).
• Bitget Sampled liquidations (largest per coin and side each second).
• Gate Every liquidation order on USDT perps (futures.public_liquidates, batched each second).
• HTX Every liquidation on USDT swaps (public liquidation_orders stream).
• Aster liquidations sampled, at most one per symbol per second. Added 2026-09-28.
• Lighter Every liquidation on crypto perps (liquidation fills on the public trade stream).

An hour with no events is not evidence of uninterrupted feed coverage. Missing coins are not zero. This ranking does not measure open-interest washout or imply a cascade is over. Not included (no registered reported feed): Hyperliquid, MEXC, KuCoin.

Coins are grouped by base ticker across registered reported-feed perpetual markets. Only valid long/short events with finite nonnegative USD notional count. Peak ties use the earliest hour; share ties use recorded total then ticker. Refreshed every two minutes, and at each UTC hour.