Study: No single signal predicts Binance Bitcoin crashes

A new single-author preprint reports that no single price, leverage or order-flow metric consistently warned before seven large Bitcoin liquidation cascades on Binance between May 2022 and October 2025.

The paper was submitted to arXiv on July 29, 2026, and has not been peer reviewed.

The author analyzed the USD-margined BTCUSDT perpetual market on Binance using one-minute price bars and five-minute measures of open interest, trader positioning and taker buy/sell flow. The study examined seven large liquidation events, using roughly two-month windows surrounding each cascade.

The analysis tested statistical indicators associated with critical slowing down, the idea that a market nearing a sudden transition recovers more slowly from disturbances and therefore shows rising variance or autocorrelation. The researcher computed rolling variance and lag-1 autocorrelation on detrended residuals across 39 combinations of analysis windows for each variable and event.

Price displayed the slowing-down pattern in five of the seven cascades. The two events that did not show that pattern were linked in the paper to abrupt tariff news in February and October 2025. The author treats the difference between those two events and the other five as a tentative observation, noting the small sample for sudden external shocks.

Order-flow measures produced mixed results. Variance in the taker buy/sell ratio tightened before every cascade with usable data, covering six events. All six observations fell in the left tail of a 300-onset placebo distribution and four were below its fifth percentile. Two events nevertheless overlapped the ordinary-market range, so the paper classifies the taker-flow compression as a population-level precursor rather than a consistent alarm for any single crash.

The study found that warning signals moved among variables across events. Leverage and order-flow carried the signature in the October 2025 event, while price showed the stronger signal in an August 2024 cascade. No tested metric produced the same positive critical-slowing-down signature across all seven cases, and out-of-sample tests produced inverted patterns between events.

The author lists several limitations: the sample is limited to one exchange and seven cascades; some 2022 series are incomplete; public measures of leverage and flow are proxies because direct intraday liquidation snapshots were unavailable. The paper did not test other public gauges such as basis, ETF flows or collateral settings.

The study’s sample closed before a later liquidation on June 25, 2026 that involved roughly $1 billion in forced derivatives closures, an event the author could not include.

The paper does not provide a single predictive metric for traders or risk managers. It reports that statistical warning signs appeared in some variables and shifted across events and recommends further testing with more events, direct liquidation data and additional public measures to evaluate indicator combinations.