The $2.75 billion figure taken from CoinGlass liquidation data after the August 19 rally counted short positions across the crypto market, not Bitcoin alone. Bitcoin contributed about $1.6 billion to $1.7 billion, as the corrected FinanceFeeds report from August 20 now states. The distinction matters because a market-wide liquidation total and an asset-level total answer different questions.
What the $2.75 Billion Number Actually Counts
A short liquidation occurs when rising prices reduce a leveraged trader’s margin below an exchange’s maintenance requirement. The exchange then closes the position, normally by buying the contract back, which can add more upward pressure and trigger further closures. The underlying mechanism is explained in FinanceFeeds’ guide to leverage and forced liquidation.
In this event, $2.75 billion referred to those forced short closures across tracked crypto derivatives venues and assets. It did not represent Bitcoin purchased in the spot market, trader deposits lost or total liquidations on both sides. Long liquidations added roughly another $250 million, taking the wider forced-closure total toward $3 billion.
Bitcoin Was About $1.67 Billion of the Market Total
CoinDesk’s later snapshot described about $3 billion of market-wide shorts being forced to cover, with Bitcoin at $1.67 billion and Ether at $1.14 billion. That framing places Bitcoin at a little over half of the squeeze rather than assigning the entire figure to it.
The apparent movement from $2.75 billion to $3 billion does not describe a second event. Tracker totals can change as a rolling 24-hour window advances, prices move and delayed observations enter the calculation. Ether’s contribution was material, consistent with the separate Ether squeeze recorded during the same rally.
Why Liquidation Trackers Disagree
There is no consolidated tape for crypto liquidations. CoinGlass says its figures aggregate perpetual and delivery-contract data from major exchanges, but the total is limited by what each venue publishes. Binance’s liquidation stream documentation says only one liquidation order for each symbol is pushed during each 1,000-millisecond snapshot, which can omit events when closures cluster. That makes an aggregate tracker total a lower-bound estimate rather than an audited market-wide figure.
Coinalyze warns that Binance stopped providing full liquidation data in April 2021 and that Bybit’s feed was incomplete from November 2021 until full reporting resumed in February 2025. Coverage also differs by contracts, exchanges and conversion method. Amberdata distinguishes liquidation orders from liquidation trades when an exchange publishes both, another methodological choice that can produce different totals.
What Record Means in This Case
The defensible claim is that August 19 produced the largest short-liquidation event in CoinGlass records beginning in 2021. It was not the largest overall liquidation event in crypto history. The October 2025 selloff generated about $19 billion in combined long and short liquidations, as the market’s subsequent rebound showed.
Even that $19 billion figure carried a data boundary. In its 2025 market report, CoinGlass said the actual nominal scale may have approached $30 billion to $40 billion after allowing for platform disclosure timing and market-maker feedback. A record based on the tracker is therefore a record within its sample, coverage and history.
How to Read the Next Liquidation Headline
Readers should first check whether a number covers Bitcoin, one exchange or the whole crypto market. They should then identify whether it counts longs, shorts or both, the length and cutoff of the measurement window, the tracker used and the starting date behind any record claim.
Liquidation figures remain useful for showing where leverage was concentrated and how much forced flow accompanied a move. They are less reliable as exact measures of trader losses or genuine spot demand. Forecast tools such as a Bitcoin liquidation heatmap are different again because they estimate where future closures may occur rather than count reported closures that have already happened.