Hyperliquid’s internal backstop absorbed most forced selling during the sharpest minute of the October 2025 crypto crash, according to a new research preprint that examined liquidation mechanics on the on-chain perpetual futures venue. The paper found that about $641 million was force-sold on Hyperliquid at 21:19 UTC on Oct. 10, with roughly $576 million routed to the venue’s backstop and about $64 million reaching the public order book. That split matters because a thin order book can accelerate price declines and trigger additional leveraged liquidations, while an internal backstop can absorb orders without pushing as much volume into visible market depth. The preprint, which has not completed peer review, focuses on Hyperliquid rather than the broader market. Under Hyperliquid’s liquidation rules, positions are first closed through market orders, but under specified conditions a liquidator vault can take over. The paper describes that vault as a component strategy within the Hyperliquidity Provider (HLP) protocol vault. The study found the backstop absorbed 62.6% of forced-sale value off-book after onset. Forced selling was also highly concentrated in time, with 87.8% of post-onset activity occurring within 30 minutes and 96.5% within one hour. Across a 15.7-hour post-onset window, the paper tracked $733 million of book-directed forced-sale value, including $644 million during the initial nucleation phase. It said the 62.6% backstop share was reported as a separate off-book series rather than part of a single combined liquidation total. To gauge whether the selloff became self-reinforcing, the authors modeled a branching ratio (average additional liquidations linked to each forced sale). Hyperliquid’s structural estimate stayed below 0.2 in every measured regime, reaching 0.195 during nucleation and easing to 0.140 at the peak, while a separate amplification calculation implied 0.122. The authors say that suggests the backstop damped feedback at the venue’s climax, though cross-exchange price links may still have amplified liquidations across the wider market. The paper presents that as a venue-specific finding and suggests future cross-venue testing, especially on platforms without a comparable backstop.