7 days Pro+ free · no cardStart my free trial

Liquidity, order books and liquidations · 4 / 5

Evaluate a liquidation cascade without calling a bottom

A liquidation cascade can accelerate a move when forced position reductions meet limited liquidity. It can also continue farther than expected. Evidence that positions are being liquidated is not evidence that the last forced seller has finished.

Athenum8 minUpdated:

Connect the mechanism to observable data

An adverse mark-price move can breach maintenance requirements. The venue's liquidation process then reduces or closes exposure under its own rules. If the resulting trading pressure meets a thin book, price can move further and put other accounts under stress. Insurance funds, partial liquidation and other safeguards affect the mechanics; do not assume every venue uses the same sequence.

Look for an aligned combination of price movement, normalised OI contraction, executed flow and reported liquidation events. An OI drop alone includes voluntary closures. Liquidation reports have specific coverage and field meanings: some report bankruptcy prices, which should not be labelled exact market execution prices. A missing event is not proof that no liquidation occurred anywhere.

Separate a flush from a recovery

A reversal hypothesis needs its own observable condition. You might study a reclaim of a prior boundary, sustained trading above it and a retest that holds. Those are example research criteria, not universally profitable settings. They create a point at which the proposed recovery can fail.

Also define a continuation scenario. Repeated failed reclaims, new aggressive selling and further loss of nearby bids can keep the move active even after a large OI reduction. The amount already liquidated is a historical fact; it does not set a floor under the next price. Do not widen risk simply because the move looks extreme.

Two paths after the same initial flush

An illustrative asset falls from 100 to 96 while normalised OI contracts from 10,000 to 8,500 contracts and the covered feed reports long liquidations. Before looking ahead, define a recovery test: two completed five-minute closes above 97.20, followed by a retest that holds that boundary under the stated close rule.

Path A closes at 97.40 and 97.30, then retests without closing below 97.20. It qualifies for further evaluation under the recovery rule. Path B rebounds only to 96.70 and then closes at 96.20. It does not qualify, although both paths began with the same dramatic flush. Execution costs and position sizing still need separate evaluation in Path A.

Identical initial event, different subsequent evidence
StagePath APath B
Initial low96.0096.00
First rebound close97.4096.70
Next close97.3096.20
97.20 acceptance testQualifies if retest holdsFails to qualify
  1. 1Forced-close evidence
  2. 2Predefined reclaim
  3. 3Retest / failure
  4. 4Risk and execution check
Reported forced closes establish the event. A separate reclaim and execution test determines whether the recovery hypothesis qualifies.

A large liquidation number is not a complete loss total

Reported notional describes covered liquidated exposure under the provider's method, not necessarily the collateral lost or every event across the market. Adding inconsistent venues or comparing a sampled feed with a fuller feed can create false records. Use consistent coverage before labelling a flush unusually large, and keep notional exposure separate from account losses.

Before acting

  • Verify OI units and the liquidation feed's side and price definitions.
  • Align the price move with the event window and venue set.
  • Declare recovery and continuation conditions before choosing a direction.
  • Estimate stressed execution and keep liquidation distinct from the intended stop.
  • Retain nonqualifying flushes in the research sample.

Check your understanding

OI falls sharply and reported long liquidations surge, but price repeatedly fails below the planned reclaim level. Has the recovery setup triggered?

Show the explained answer

No. The forced-close event occurred, but the separate recovery condition did not. Entering solely because the liquidation total looks large would change the rule after observing the event and leave the original hypothesis untested.

Sources and further reading

Continue with Athenum