Forced Liquidation
Cascades
Perpetual futures markets have no closing bell, no circuit breaker, and no regulator. When leverage breaks, automated margin calls create self-reinforcing destruction loops. One year of real liquidation data — March 2025 to March 2026 — dissected.
Perpetual futures: leverage with no expiry.
When it breaks, it breaks fast.
Traders on unregulated perpetual futures markets take leveraged positions — sometimes 50× or 100× their capital. A 1% adverse move against a 100× position wipes the entire stake. The exchange does not request more margin. It liquidates automatically, at market price, immediately.
This creates the cascade. Force-selling a liquidated long depresses price. Lower price triggers the next band of longs. Those get liquidated. Their selling depresses price further. The loop is mechanical, not behavioural. The underlying thesis of each position is irrelevant — only the price level matters to the engine.
Price falls → longs liquidated → force-sold into market → price falls further → next band of longs triggered. Self-reinforcing until leverage is exhausted. Cascade Efficiency near 1.0 — the session never recovers.
The identical mechanism in reverse. Price rises → shorts force-bought out → buying pressure accelerates the rally → next band of shorts triggered. July 10, 2025: $963M in shorts liquidated in a single session. Price +4.3%. No news required.
High open interest is fuel. Every narrative cycle — Uptober, halving hype, ETF inflows — recruits fresh leveraged longs. The more OI stacked at adjacent levels, the larger the cascade when the trigger is hit. The market builds its own accelerant.
Regulated equity markets halt trading during extreme moves. Crypto perpetuals do not. The engine runs 24/7, 365 days. The October 10, 2025 cascade — $16.78B in long liquidations — occurred and completed within a single Asian session. No intervention. No floor.
BTC price against daily forced exits.
The machine is always running.
Above 1 = shorts squeezed · Below 1 = longs cascading
Spikes precede largest cascade events
∗ Mar 2025 – Mar 2026 · All figures USD · Short liq = forced buybacks on short positions · Long liq = forced sales on long positions
$16.78 billion in long liquidations.
In one session. No news. Pure mechanics.
A Cascade Efficiency of 0.956 means the session closed near its absolute low — the entire intraday range was consumed by the cascade, with no recovery. This is the fingerprint of mechanical liquidation, not informed selling. The market was not repricing fundamental value. It was executing margin calls until it ran out of leveraged longs to kill.
Every major crypto asset fell simultaneously. The correlation is structural — all share the same leveraged derivatives infrastructure. Meme coins lost 35–42% in the same session BTC fell 7%. The losses were not proportional to any thesis. They were proportional to how much leverage each asset’s market was carrying.
∗ Source: 10OctIntraday sheet · Perpetual futures market close prices · 41 assets tracked
CE = (Open-to-Low) ÷ Intraday Range.
Values near 1.0 are the machine’s signature.
A Cascade Efficiency of 1.0 means the session low was the close — price fell without interruption and stayed there. These are not volatile trading days. They are mechanical events: each liquidation selling into the next. The market never had a chance to recover because every attempted bounce triggered the next wave of forced selling.
The largest events — documented.
| Date | Total Liquidations | CE | Type |
|---|---|---|---|
| 10 Oct 2025 |
$19.25 billion
$16.78B long liq · 17.3% intraday range · Low $101,516
|
0.956 | Long Cascade |
| 24 Aug 2025 |
$1.194 billion
$841M long liq · BTC $113,446
|
0.949 | Long Cascade |
| 14 Nov 2025 |
$1.119 billion
$942M long liq · Vol at 41% annualised
|
0.965 | Long Cascade |
| 31 Jan 2026 |
$2.562 billion
$2.99B short side · Cascade Efficiency 0.997
|
0.997 | Long Cascade |
| 10 Jul 2025 |
$963 million
Short:Long ratio 9.98× · Shorts obliterated
|
0.062 | Short Squeeze |
| 6 Feb 2026 |
$1.278 billion
Largest upside session in dataset
|
0.258 | Short Squeeze |
∗ CE = Cascade Efficiency = (Open-to-Low move) ÷ (Total Intraday Range). Near 1.0 = price fell to session low and stayed. Near 0 = full intraday recovery.
182 gain days. 183 loss days.
Statistically indistinguishable from a coin flip.
In 380 trading days, BTC gained on 182 and lost on 183. The market did not trend directionally. It recycled capital. Every Uptober narrative, every halving cycle, every ETF inflow headline recruited fresh leveraged longs. The machine loaded. At some arbitrary trigger, it unloaded. The survivors were cautious. FOMO returned. The machine reloaded.
Every price high recruits new leveraged longs. Seasonal narratives — Uptober, halving cycles, ETF inflows — are deployed precisely when open interest is already building. The market is given a story to explain why it is safe to add more leverage. It is never safe.
The cascade does not require news. It requires a price at which the first liquidation band is breached. That can be a large market sell, a negative funding rate spike, or pure stochastic noise. Once the first band breaks, the mechanism supplies its own momentum.
Cascade Efficiency values above 0.95 — common in the dataset — confirm that these sessions do not recover intraday. The market closes near its low. The leverage is gone. The positions are gone. The capital has transferred to whoever was on the other side.
After a cascade, survivors are cautious — briefly. Then FOMO returns. Open interest climbs. The next cohort enters, largely unaware of the previous cycle. The machine does not need the same traders. It needs the same psychology. Both are reliably renewable.
182 gains. 183 losses. And one $19 billion afternoon.
Perpetual futures markets are not price discovery mechanisms. They are leverage amplifiers with no safety valve. The largest single-day wipeout in this dataset — October 10, 2025 — did not happen because Bitcoin changed fundamentally. It happened because $16.78 billion in leveraged long positions were stacked at adjacent price levels and a sufficient downward move detonated them sequentially. The underlying asset was irrelevant. The structure was everything.
Data: perpetual futures liquidation feed · Mar 2025 – Mar 2026 · 380 daily sessions · BTC OHLC, total liquidations (all coins + BTC-specific), cascade efficiency, rolling volatility · Not financial advice · This is arithmetic
paulfaulkner.com
