When every crypto exchange offers roughly the same leveraged trading product, competition moves to places most traders never bother looking. That is exactly what has happened across the onchain perpetual futures market, where venues like Hyperliquid, Drift, Paradex, and a long tail of newer entrants all offer essentially identical bets settled onchain. According to Margie Feng, marketing lead at Solayer, whose company operates the Solana-based venue Margin Trade, the real battle now comes down to two things nobody advertises on their homepage: what an exchange does to you when you win too much, and which obscure markets it will list that competitors won’t touch.
The mechanism at the center of that first fight is called auto-deleveraging, or ADL, and it is exactly as jarring as it sounds. When a leveraged position blows up so catastrophically that neither the trader’s collateral nor the exchange’s insurance fund can cover what is owed, someone has to eat the loss. So the exchange reaches across its order book and forcibly closes out people who were on the right side of the trade. Winners get liquidated to pay for losers they never traded against directly. Most of the time this plumbing stays invisible, but on October 10, 2025, it became impossible to ignore. The crypto market suffered its largest single-day liquidation event ever, roughly nineteen billion dollars wiped out market-wide, with about ten point three billion concentrated on Hyperliquid alone. In twelve minutes, Hyperliquid’s system force-closed two point one billion dollars worth of winning positions.
How those closures are chosen has become one of the more bitter technical arguments in crypto. The dominant approach traces back to a design from Huobi circa 2015 and has been copied across the industry since. It works through an industry shorthand known simply as the Queue, employed by Binance, Hyperliquid, and others, which sorts profitable traders by gains and leverage and starts shutting them down from the top until the shortfall is covered. Tarun Chitra, founder of risk-modeling firm Gauntlet, published a paper arguing that during October’s chaos the Queue was wildly inefficient — closing far more than necessary and concentrating pain on a small group of big winners while leaving excess profits destroyed somewhere between forty-five and fifty-two million dollars. Dan Robinson, a research partner at Paradigm, publicly attacked both the methodology and the conclusion, writing that Chitra had described an algorithm far crazier than what Hyperliquid actually runs and accusing him of inflating costs from twenty-three million in actual deficit to six hundred fifty-three million
