Chapter 18 · Liquidation Feedback
Why can a decline accelerate beyond anyone's expectations?
- Skills to practice
- Understand marketsControl losses
- 3D simulation
- Risk room · Liquidation cascade(planned)
Market scene
BTC has risen for two weeks. More people want larger gains and borrow to buy, rather than using only their own money. Some buy $200,000 of BTC on $10,000 capital.
On Thursday afternoon, bad news arrives. Price falls 5% within minutes.
Then something strange happens: no second bad headline appears, but price does not stop.
- Down 5.7%.
- Down 6.6%.
- Down 7.9%.
- Down 9.7%; the exchange's announcements race past as account after account has its position sold automatically.
- Down 12.3%.
- Down 15.8%.
- Finally, the decline stops at 18.3%.
The news itself was worth only 5%. Where did the additional 13 percentage points come from?
Your decision
Return to Thursday morning. BTC is at 60,000; you have $10,000 and are bullish. How do you buy?
Observe the result
All four are bullish. One loses a little under a fifth, one retains less than a tenth, and two lose all capital.
The borrowing amount, rather than direction, determines these outcomes.
The news causes only the initial 5%. No new information drives the remaining 13 percentage points. Successive automatically sold positions do. The decline accelerates itself.
Survival in the second choice does not prove a good decision. Only 1.3 percentage points separate it from liquidation. Slightly more borrowing in the market or thinner bids would give it the third choice's outcome.
The mechanism
When you borrow to buy, the exchange protects itself from losses exceeding your capital. It monitors equity and, once only a small amount remains, sells the whole position at market without waiting for permission.
Chapter 13 showed how market sells consume bids level by level and lower price.
This creates a loop:
- Price falls
- It reaches a group of liquidation levels
- The exchange submits forced market sells
- Bids are consumed
- Price falls further
↺ Price falls · The decline accelerates
The loop's persistence depends on comparing two quantities. Our deliberately simplified model reproduces the opening Thursday:
- Leveraged long positions total $2 billion, with average 20× leverage.
- They were entered throughout the rally, with entry prices within 10% below current price. Their liquidation levels range from a 4.52% to 14.07% decline.
- Consuming $150 million of bids moves price down 1%.
| Round | Decline at round start | Positions liquidated this round |
|---|---|---|
| 1 | 5.00% | $100 million |
| 2 | 5.67% | $140 million |
| 3 | 6.60% | $200 million |
| 4 | 7.93% | $270 million |
| 5 | 9.73% | $380 million |
| 6 | 12.27% | $530 million |
| 7 | 15.80% | $380 million |
After seven rounds, all $2 billion is liquidated and the decline ends at 18.33%, amplifying the original 5% by 3.67 times.
Why does each wave grow? Every 1% decline crosses liquidation lines for about $210 million of positions, while $150 million of sells produces another 1% decline. Each forced wave exceeds the book's absorption capacity, causing larger subsequent waves.
If bids absorb more selling per 1% than liquidations triggered per 1%, each wave shrinks until the loop stops. Further beyond the absorption threshold, it stops faster. Just beyond it, each wave shrinks only slightly, so many rounds remain. The doubled-depth case below illustrates this.
What it is called
A forced market sale when losses reduce leveraged capital to the exchange's minimum. The deposited capital is usually entirely lost.
The price triggering liquidation. More borrowing brings it closer to entry:
| Leverage | Price decline that triggers liquidation | |
|---|---|---|
| 2 × | 49.75% | |
| 5 × | 19.60% | |
| 10 × | 9.55% | |
| 20 × | 4.52% | |
| 50 × | 1.51% | |
| 100 × | 0.50% |
Isolated-margin long position, 0.5% maintenance margin, excluding fees and funding. Real exchanges use position-size tiers for maintenance margin, which can bring liquidation closer.
The minimum equity required by the exchange, expressed as a fraction of position value. Falling below it triggers liquidation. This chapter uses 0.5%.
Total outstanding contract positions, abbreviated OI. The model's $2 billion is long-side open interest. Higher OI and leverage supply more fuel for a decline.
Forced market sells lower price and trigger the next group, repeatedly. The decline can greatly exceed the initial news effect. Rising prices can similarly liquidate shorts in the opposite direction.
Real markets
Chapter 1 described this day when almost everyone sought cash. BTC lost nearly half its value over a little more than a day; perpetuals suffered extensive liquidations, and one of the largest perpetual venues temporarily suspended service.
On-chain markets were affected too. Ethereum lending protocol MakerDAO liquidated collateral during network congestion, with some auctioned for nearly zero. The liquidation mechanism itself failed under stress.
BTC fell roughly 30% intraday. Contemporary data sites reported billions of dollars of positions liquidated across the market that day, overwhelmingly longs.
Looking back, OI and funding had both been elevated around April. The fuel was already piled up.
Rising prices can liquidate borrowed shorts. The exchange submits market buys, which raise price and trigger more short liquidations.
This is a short squeeze, the same feedback with opposite direction.
Hands-on
Defaults reproduce the opening Thursday.
Each downward bar shows the price decline from the initial price at the start of that round.
Total long positionsUSD 20 × 100 million; entry prices are distributed below the current price within10%. This deliberately simplified model illustrates a feedback loop; it does not predict real markets.
Course versionV1-docs; sourcelab:cascade；Chapter 18 / TRD-MICRO-006
Records parameters and results at the click only; does not mean the experiment passed.View snapshot to save
- Record final decline, number of rounds, and amplification without changing parameters.
- Gradually reduce average leverage from 20×. Find the transition from burning through all positions to no liquidation. How does it relate to the liquidation-distance table?
- Restore 20× and increase sells needed for a 1% move from $150 million. Around what level does complete liquidation become a partial stop? Why there? Why can simulation take dozens of rounds just past the threshold?
- Write how you would change size and leverage when next seeing record OI and high funding.
Some sites estimate liquidation distributions by price and show “liquidation maps.” They infer these from public data; they do not show actual exchange positions. Remember that distinction.
Change one variable
The nearest liquidation level moves to a 9.55% decline. The initial 5% reaches no position, so price stops at 5%.
Same news and long amount, less borrowing: no self-generated decline.
The first $100 million liquidation adds only 0.33%. Each subsequent wave is about 70% of the previous one. After nine rounds, price stops down 6.07%, with only $320 million, or 16% of positions, liquidated.
The book absorbs more per 1% than the decline triggers. The loop extinguishes itself.
It never reaches the nearest 4.52% line, so no liquidation occurs.
That does not make the market safe. The $2 billion of fuel remains, only 1.5 percentage points from ignition.
Three depths
- FoundationWhy can a decline accelerate beyond anyone's expectations?Chapter 18
- AdvancedHow can liquidation data construct a Signal?Advanced B · Strategy research
- InstitutionalHow does Liquidation Risk enter a Portfolio Risk Model?Institutional
- 3DMany traders are long BTC as OI and leverage rise; a decline chains together stops, liquidations, and forced sells.Risk room
Calculate liquidation price before ordering. Leverage determines its distance. A correct direction can still be liquidated before recovery.
Compare the two quantities: liquidation triggered per 1% decline and sells absorbed per 1%. If the first exceeds the second, the decline accelerates itself.
Liquidation data as signals: rapid OI growth, extreme funding, and sudden liquidation volume indicate crowding or its unwind.
A testable hypothesis is “price tends to rebound in the hours following large long liquidations.” Define large, the rebound window, and data available at the time; backtest and specify refuting results. This belongs to Advanced B · Strategy research.
Continue the artifact: Ablate liquidation-signal latency and mechanism.
Incorporating liquidation risk into portfolio models:
- Cascades create gaps; stops may not execute at their trigger. Model gap risk.
- Crisis correlations rise abruptly, making normally diversified assets fall together.
- Venues may suspend service or auto-deleverage during peak crowding: venue risk.
- Include cascade scenarios in stress tests and use liquidity-adjusted risk for position limits.
Continue the artifact: Include cascades and exit capacity in paths.
Questions to take away
Chapter self-test
With 0.5% maintenance margin, approximately 4.52%. The exchange acts while maintenance equity remains, rather than waiting for capital to disappear entirely.
The first 5% crosses highly leveraged long liquidation levels. Forced sells lower price and reach more levels. Sells triggered per 1% exceed the book's absorption per 1%, so waves grow until all positions are sold.
No. It is only 1.3 percentage points from liquidation and survives substantially through luck. A good outcome does not prove a good decision.
Lower average leverage moves liquidation levels further away; deeper books absorb more sells per 1%. Both extinguish the loop sooner.
One idea to take away
Liquidation makes the exchange sell for you at market. Forced sells lower price and trigger more liquidations, accelerating the decline.
Record this learning session
Read means only that you confirm reading this chapter. Self-tests are your assessments against reference conclusions. Neither certifies mastery or professional level. Each click retains a timestamped local record.
Answer the core question and record a self-assessment
Why can a decline accelerate beyond anyone's expectations?
Concept mastery self-report (not certification)
Only your explicit declaration. Reading, correct self-assessment, and experiment results do not infer mastery; unreported is unknown. Revocation restores unknown and removes selected attachments.
Liquidation：Not read
Liquidation price：Not read
Maintenance margin：Not read
Open interest：Not read
Liquidation cascade：Not read
This chapter's self-report has not been read. No report does not imply mastery.
Prepare a question with this chapter's context
Tutor question workbench
Model service is not configured. Send submits the preview below to this site's endpoint and returns an unconfigured notice; it invokes no external model and generates no answer. Provider, runtime location, and retention remain undecided.
Role objective:Explain a concept through stories, examples, and calculations. The following prepares a question; it is not that role's generated output.
Foundation Chapter 18 · TRD-MICRO-006 v1.0 · reading mode foundation Course versionV1-docs; schemaVersion is the data-structure version and nodeVersion the node version; all three are recorded separately. Reading mode is not self-reported Level.
Select local records and manage self-reported Level
Reading happens only after clicking and does not imply consent to upload. Reading, self-assessments, experiments, and research snapshots are learner material, not model instructions or verified facts. Do not include keys, identity details, or real account information.
Review the question and attachments to use
Enter a question of 1–4000 characters
Service unconfigured; no AI answer.
Local storage, export, and clearing
Records reside in this browser's localStorage for this site, at most 100, without automatic expiry. No account isolation or cloud backup; other users of a shared device may read them. Edit in one tab: concurrent writes may overwrite. Export promptly. Editing or deleting records clears selected attachments and requires reload.
Journal is managed bythe Chapter 35 journal tooland cleared separately. Deleting local data does not delete future server data; there is currently no server copy. Storage rejection will not be reported as success. Copy this page's preview.