Insurance Funds & Socialized Loss
TL;DR
- Insurance funds exist to cover the gap between a trader's liquidation price and their bankruptcy price — the zone where a liquidated position might generate bad debt that someone has to eat
- Funded by liquidation surplus: when a liquidation fills at a better price than the bankruptcy price, the leftover margin flows into the insurance fund. Calm, liquid markets = fund grows. Violent crashes = fund draws down
- When the fund runs out, exchanges resort to socialized loss mechanisms: either auto-deleveraging (ADL), which force-closes profitable traders' positions, or clawback, which takes a percentage of winners' profits to cover the shortfall
- ADL ranks traders by a combination of unrealized profit and effective leverage — the most profitable, most leveraged traders get deleveraged first
- Historical failures matter: OKEx's 2018 clawback ($9M socialized across traders from a single $460M position), dYdX's $9M insurance fund drain from a YFI manipulation attack in 2023, and Deribit's emergency 500 BTC injection during Black Thursday 2020
- BitMEX's insurance fund grew to over 36,000 BTC (~$300M+) by 2019, sparking controversy — was it a safety net or an opaque profit center built on aggressive liquidations?
- Design tradeoffs are real: too small a fund means frequent socialization that punishes winners; too large a fund suggests the exchange is liquidating too aggressively and extracting value from traders
- Transparency is the frontier: most insurance funds are black boxes. Proof-of-reserves for insurance funds is still rare, and auditable on-chain funds (like Hyperliquid's HLP vault) represent a new model
1. Why Insurance Funds Exist
Every leveraged position has two critical price levels:
- Liquidation price: The price at which the exchange begins force-closing the position because the trader's margin fraction has dropped below the maintenance requirement
- Bankruptcy price: The price at which the trader's account equity hits exactly zero — the position is worth nothing
The gap between these two prices is the danger zone. In theory, the exchange liquidates at the liquidation price, captures a small surplus, and everyone is fine. In practice, markets are fast, order books are thin, and liquidation orders don't always fill at the liquidation price.
Price movement for a long position:
Entry Price $100.00
│
│ Trader is profitable
│
▼
Liquidation Price $95.00 ← Exchange starts force-closing here
│
│ THE GAP — this is where insurance funds matter
│
▼
Bankruptcy Price $93.00 ← Account equity = $0 at this price
│
│ BAD DEBT ZONE — if liquidation fills below here,
│ someone has to eat the loss
▼Three scenarios can play out when a position is liquidated:
| Scenario | Liquidation Fill Price | Result |
|---|---|---|
| Best case | Better than bankruptcy price (e.g., $94.50) | Surplus goes to insurance fund |
| Break-even | Exactly at bankruptcy price ($93.00) | No surplus, no deficit |
| Bad debt | Worse than bankruptcy price (e.g., $91.00) | Insurance fund covers the shortfall |
The insurance fund exists because scenario 3 is inevitable in volatile markets. Without it, the counterparty on the winning side of the trade — the short who is owed $7 of profit per contract — doesn't get paid in full. The insurance fund bridges that gap so winning traders always receive what they're owed.
This is a problem unique to derivatives exchanges. In spot trading, you can't lose more than what you put in. In leveraged derivatives, a position can go negative — the trader owes more than their collateral. The exchange must decide who absorbs that negative equity. The insurance fund is the first line of defense.
2. How Insurance Funds Are Funded
Insurance funds are primarily funded through liquidation surplus — the margin remaining in a liquidated account when the position is closed at a price better than the bankruptcy price.
The Mechanics
When a trader gets liquidated, the exchange takes over their position and closes it on the open market. If the closing price is better than the bankruptcy price, there is leftover margin. That leftover goes into the insurance fund.
Example: Long position liquidation
Position: 100 BTC long, 20x leverage
Entry price: $50,000
Margin: $250,000 (100 × $50,000 / 20)
Liq. price: $47,750 (exchange begins force-close)
Bankruptcy: $47,500 (account equity = $0)
Actual liquidation fill: $47,650
Surplus per BTC: $47,650 - $47,500 = $150
Total surplus: 100 × $150 = $15,000 → Insurance FundIn calm, liquid markets, most liquidations fill above the bankruptcy price because order books are deep and the spread between liquidation and bankruptcy is wide enough. The insurance fund steadily accumulates.
Secondary Funding Sources
Some exchanges supplement liquidation surplus with:
- A percentage of trading fees — Binance allocates 10% of trading fees to its SAFU fund
- Direct exchange contributions — OKX injected 2,500 BTC (~$18M at the time) into its insurance fund after the 2018 clawback incident
- Initial seeding — most exchanges seed the fund with their own capital at launch
The Accumulation Pattern
Insurance fund growth follows a predictable cycle:
Fund Balance
│
│ ╱─────────────╲
│ ╱ Bull market ╲ Calm markets:
│ ╱ High volume ╲ - Deep order books
│ ╱ Lots of ╲ - Liquidations fill cleanly
│ ╱ liquidation ╲ - Surplus accumulates
│ ╱ surplus ╲
│ ╱ ╲ ← Crash: massive drawdown
│──╱ ╲___________
│ ╲
└──────────────────────────────────────────────── Time
Accumulation phase DrawdownThe fund grows slowly during normal conditions and can be depleted rapidly during extreme events. This asymmetry is the fundamental challenge of insurance fund design.
3. The Insurance Fund Lifecycle
Phase 1: Accumulation (Calm Markets)
During normal market conditions, the insurance fund is a machine that prints money. Most liquidations are small, order books are deep, and positions close well above bankruptcy price. The fund grows steadily.
BitMEX demonstrated this dramatically: from its launch through 2019, the fund grew to over 36,000 BTC (approximately $324 million at the time). By the end of 2019, the fund had grown 61% in that year alone. At its peak, BitMEX's insurance fund held roughly 0.15% of all Bitcoin in existence.
Phase 2: Stress Testing (Moderate Volatility)
During moderate selloffs (10-20% intraday moves), the insurance fund earns its keep. Some liquidations fill below bankruptcy — the fund covers the shortfall. But many still fill above bankruptcy, so the fund might even grow during moderate volatility events.
This is the Goldilocks zone where insurance funds prove their value. Traders don't experience ADL or socialized loss. The system absorbs the shock transparently.
Phase 3: Drawdown (Market Crash)
During severe crashes — Black Thursday 2020, May 2021, FTX collapse — the fund enters rapid drawdown. Multiple things go wrong simultaneously:
- Order books evaporate as market makers pull quotes
- Liquidation volume spikes as cascading positions hit maintenance margin
- Slippage explodes — liquidations fill far below bankruptcy prices
- Bad debt accumulates faster than the fund can absorb it
Deribit's experience on Black Thursday 2020 is instructive: the exchange's insurance fund was cut nearly in half, and Deribit had to inject 500 BTC of its own capital to prevent the fund from being fully depleted, which would have triggered socialized losses across all profitable traders.
Phase 4: Recovery or Failure
After a crash, one of two things happens:
- Recovery: The fund survived the drawdown. It begins accumulating again as markets stabilize. Traders' confidence is maintained.
- Failure: The fund was depleted. The exchange triggers socialized loss mechanisms (ADL or clawback). Traders lose confidence. Volume may permanently decline.
The transition from Phase 3 to Phase 4 is the most consequential moment in an exchange's lifecycle. Surviving a major crash without socializing losses is the ultimate validation of insurance fund design.
4. When the Insurance Fund Isn't Enough
When bad debt exceeds the insurance fund balance, the exchange has only two options: spread the loss across other traders, or eat the loss itself. Almost no exchange chooses the latter. The two main socialized loss mechanisms are clawback and auto-deleveraging (ADL).
Clawback (The Old Way)
Clawback is the blunter of the two mechanisms. At the end of a settlement period (typically weekly for quarterly futures), the exchange calculates the total bad debt, then deducts a proportional percentage from every profitable trader's realized gains.
Clawback Calculation:
Total bad debt this period: $1,000,000
Total profits (all winners): $10,000,000
Clawback rate: $1,000,000 / $10,000,000 = 10%
Trader profits $50,000 → keeps $45,000
Trader profits $5,000 → keeps $4,500
Trader profits $500 → keeps $450Clawback is simple but deeply unfair. A conservative trader using 2x leverage who made $5,000 gets the same 10% haircut as a degen using 100x who made $50,000. There is no differentiation based on risk profile, position size, or contribution to the problem.
Most modern exchanges have abandoned pure clawback in favor of ADL. But it still exists as a last resort at some venues.
Auto-Deleveraging (ADL) (The Modern Approach)
ADL is more targeted. Instead of taking a percentage from all winners, it force-closes specific positions — starting with the traders who are most profitable and most leveraged. The idea is that these traders have benefited the most from the move that caused the bad debt, and they can best afford to give up their position.
ADL is covered in detail in the next section.
The Psychological Impact
Both mechanisms create a perverse dynamic: your winning trade can be taken from you. You correctly predicted the market direction, managed your risk, held through volatility — and the exchange still takes money from you because someone else blew up.
This is deeply alienating for traders. It creates an incentive to withdraw profits frequently (reducing your ADL ranking), to avoid holding large winning positions, and to move volume to exchanges with larger insurance funds. The existence of socialized loss mechanisms is one of the strongest competitive differentiators between exchanges.
5. Auto-Deleveraging (ADL) Explained
ADL is the mechanism that force-closes profitable traders' positions to absorb bad debt when the insurance fund is depleted. It's a last resort — but on some exchanges, it triggers more often than traders expect.
How ADL Ranking Works
Exchanges rank all traders on the opposite side of the liquidating position by a priority score. The formula varies slightly by exchange, but the standard approach is:
For profitable positions:
ADL Priority = PnL% × Effective Leverage
Where:
PnL% = Unrealized profit / Initial margin
Effective Leverage = abs(Position notional) / Account equity
Example:
Trader A: +50% PnL, 10x effective leverage → Priority = 500
Trader B: +20% PnL, 25x effective leverage → Priority = 500
Trader C: +80% PnL, 3x effective leverage → Priority = 240Traders with higher priority scores are deleveraged first. The ranking blends two factors:
- How profitable you are — larger unrealized gains push you up the queue
- How leveraged you are — higher effective leverage means you're taking more risk from the system's perspective
This means the "biggest, most profitable whales" get sent home first. A trader with massive unrealized profit on high leverage is the first to be force-closed.
What Happens During ADL
When ADL triggers:
- The exchange selects the highest-priority trader on the opposite side
- Their position is partially or fully closed at the bankruptcy price of the liquidated account
- The delevered trader keeps their profit up to that point — they are not "losing money," but they lose the position and the opportunity to profit further
- If the first trader's position isn't enough to cover the bad debt, the next trader in the queue is hit
ADL Flow:
Bankrupt Long Position (100 BTC, can't be filled on book)
│
│ Insurance fund empty
│
▼
┌─────────────────────────────┐
│ ADL Queue (Short Side) │
│ │
│ #1: Trader A (Priority 500)│ ← Force-closed first
│ #2: Trader B (Priority 500)│
│ #3: Trader C (Priority 240)│
│ #4: Trader D (Priority 100)│ ← Rarely reached
│ ... │
└─────────────────────────────┘
│
│ Position matched at bankruptcy price
│
▼
Bad debt absorbed. System solvent.The ADL Indicator
Most exchanges show traders an ADL indicator — typically a set of bars or lights showing where they sit in the queue. If all your bars are lit, you are at the top of the queue and most likely to be delevered next.
Smart traders monitor this indicator and voluntarily reduce their position size or take profits when their ranking is high. This is an intended behavior — ADL is designed to create an incentive for large profitable positions to self-deleverage.
ADL vs. Clawback: Comparison
| Clawback | ADL | |
|---|---|---|
| Who pays | All profitable traders equally | Highest-priority traders first |
| When | End of settlement period | In real-time |
| Fairness | Low — no risk differentiation | Higher — targets high-profit, high-leverage |
| Predictability | Unknown until settlement | Traders can see their queue position |
| Position impact | Profit reduced, position kept | Position force-closed, profit kept |
| Behavioral incentive | Withdraw profits before settlement | Reduce position size / take profits early |
6. Historical Socialized Loss Events
OKEx Clawback — July 2018
The most notorious clawback in crypto history. On July 31, 2018, a single trader held an enormous long position of approximately 4.17 million BTC futures contracts (each with $100 notional value) — a position worth roughly $416 million.
OKEx requested the trader reduce the position. They refused. OKEx initiated forced liquidation, but the position was so large relative to market liquidity that it couldn't be absorbed without catastrophic price impact. The exchange's insurance fund was insufficient to cover the shortfall.
The result: OKEx implemented a socialized clawback, initially calculating an 18% haircut on all profitable traders' weekly realized gains. After OKEx injected 2,500 BTC of its own capital into the insurance fund, the clawback was reduced to approximately 17%. All profitable futures traders that week lost 17 cents of every dollar they'd earned — regardless of whether they were conservative 2x traders or aggressive 50x traders.
Aftermath: OKEx overhauled its risk management — implementing tiered position limits, enhanced maintenance margin requirements for large positions, and improved liquidation algorithms. The event remains a cautionary tale about what happens when a single trader's position exceeds the exchange's risk management capacity.
BitMEX Early Socialized Losses (2015-2016)
In its early days, before the insurance fund was large enough to absorb significant losses, BitMEX used a socialized loss system where shortfalls were distributed among profitable traders at settlement. These events were frequent enough that BitMEX's own blog characterized the insurance fund's growth as "reducing the likelihood of socialized losses to near-zero."
By 2019, the fund had grown so large that socialized losses on BitMEX became essentially a non-event — which then created its own controversy (see Section 7).
Deribit — Black Thursday 2020
While BitMEX's insurance fund was (controversially) barely used during the March 2020 crash, Deribit was hit hard. The exchange's insurance fund was slashed by nearly 50%. To prevent socialized losses, Deribit injected 500 BTC of its own capital.
This highlights a critical difference: Deribit's insurance fund was appropriately sized for normal conditions but insufficient for a 50%+ single-day crash. The exchange backstopping the fund with its own capital is an ad hoc solution that doesn't scale.
dYdX v3 — YFI Manipulation Attack (November 2023)
In November 2023, an attacker manipulated the YFI (Yearn Finance) market to drain $9 million from dYdX v3's insurance fund — roughly 40% of the total fund.
The attack: The attacker built large YFI long positions on dYdX v3, then manipulated the YFI spot price upward on external markets. As the price rose, the attacker withdrew unrealized gains, deposited from new wallets, and built more long positions. When the positions eventually collapsed, the attacker's accounts had negative equity. The insurance fund covered the $9 million deficit.
dYdX CEO Antonio Juliano called it "pretty clearly a targeted attack against dYdX" and involved the FBI in the investigation. The fund still had $13.5 million remaining, so socialized losses were avoided. But the incident exposed how thin-liquidity markets on decentralized exchanges can be weaponized against insurance funds.
Hyperliquid HLP Vault — Whale Manipulation (2025)
Hyperliquid's community-funded HLP vault (which functions as the protocol's insurance fund and backstop liquidity provider) has faced multiple incidents where large traders strategically built positions and then allowed them to be liquidated into the vault, effectively socializing losses across all HLP depositors.
In one notable case, a whale transferred approximately $4 million in losses to the HLP vault by opening highly leveraged positions and allowing them to be absorbed during liquidation. Since anyone can deposit into the HLP vault, the losses were borne by the community of liquidity providers rather than the exchange itself — a novel model, but one that raises questions about informed consent and the sophistication required of vault depositors.
7. The BitMEX Insurance Fund Controversy
BitMEX's insurance fund deserves special attention because it crystallizes the central tension in insurance fund design: safety vs. extraction.
The Growth Story
| Period | Fund Balance | Notes |
|---|---|---|
| 2016 (launch) | ~0 BTC | Seeded with small amount |
| End of 2018 | ~21,000 BTC | Rapid growth during bull-then-bear year |
| Mid-2019 | ~31,300 BTC (~$314M) | Up 50% YTD |
| End of 2019 | ~33,800 BTC | Up 61% for the year |
| March 2020 | ~36,493 BTC (ATH) | Grew during Black Thursday |
| 2020 onward | Declining with BitMEX market share | Regulatory issues, competition |
Why Critics Called It a Problem
1. Aggressive liquidation mechanics generate more surplus
If an exchange liquidates early (at a price further from bankruptcy), there is more margin remaining in the account. That surplus flows to the insurance fund. Critics argued that BitMEX's liquidation engine was unnecessarily aggressive — liquidating traders earlier than necessary to capture more surplus.
Deribit explicitly made this argument in a blog post, stating that "large insurance funds could indicate an overly aggressive liquidation mechanism."
2. The fund may have been treated as an exchange asset
BitMEX never provided a clear breakdown of drawdowns per contract or how the fund was managed. The Block noted the lack of transparency and suggested BitMEX may have considered the fund an asset on its balance sheet rather than a reserve held in trust for traders.
3. A fund that never depletes isn't being used
During Black Thursday 2020 — the most severe single-day crash in Bitcoin's history — BitMEX's insurance fund not only survived but actually grew. While this sounds like good news, it raised uncomfortable questions. If a fund designed for catastrophic events doesn't get used during a catastrophic event, is it too large? And where is all that money going?
The counterargument: BitMEX claimed the fund performed exactly as intended — preventing ADL from triggering even during extreme conditions. The fund's size was a feature, not a bug.
The Broader Lesson
BitMEX's insurance fund controversy established a principle that the industry is still grappling with: insurance funds have an optimal size range. Below that range, socialized losses happen too often. Above it, the exchange is extracting value from traders through aggressive liquidations. Finding the sweet spot — and being transparent about it — is a design challenge that no exchange has fully solved.
8. How Major Exchanges Handle Insurance and ADL
| Feature | Binance | Bybit | OKX | dYdX | Hyperliquid |
|---|---|---|---|---|---|
| Fund type | SAFU + per-contract insurance | Per-contract insurance pools | Insurance fund + risk reserve | Protocol insurance fund (USDC) | HLP vault (community-funded) |
| Fund size | SAFU: ~$1B (15,000 BTC as of Feb 2026); per-contract funds separate | Varies; New Listing pool starts at $8M, Portfolio pool $2-4M | Not fully disclosed; injected 2,500 BTC post-2018 incident | ~$13.5M (post YFI attack) | Variable (community deposits) |
| Funding source | 10% of trading fees + liquidation surplus | Liquidation surplus | Trading fees + liquidation surplus | Liquidation penalties + fees | Trading fees + liquidation surplus + depositor capital |
| Socialized loss mechanism | ADL | ADL | ADL (historically clawback) | ADL (v4); Insurance fund first | HLP vault absorption → ADL |
| ADL ranking | PnL% × Effective Leverage | PnL% × Effective Leverage | PnL% × Effective Leverage | Profit + leverage weighted | Profit + leverage weighted |
| ADL trigger condition | Insurance fund depleted | Insurance fund depleted; triggers if fund drops 30% in 8 hours | Insurance fund depleted | Insurance fund depleted | HLP vault can't absorb |
| Transparency | SAFU wallet publicly visible; proof of reserves | Daily insurance fund balance published | Proof of reserves published; insurance fund size partially disclosed | On-chain, fully auditable | On-chain, fully auditable |
| Notable events | No major socialized loss events to date | Introduced specialized pools (Dec 2025) for new listings | 2018 clawback ($9M socialized) | 2023 YFI attack ($9M drain) | Multiple vault loss events from whale manipulation |
Key Differences
Binance stands alone with its dual-layer approach: the SAFU fund ($1 billion, recently converted to BTC) covers security breaches and extreme events, while per-contract insurance funds handle normal liquidation shortfalls. The sheer size of the SAFU fund makes socialized losses extremely unlikely on Binance.
Bybit took a nuanced approach in late 2025 by creating specialized insurance fund pools — a "New Listing" pool (starting at $8M) for newly listed contracts in their first 30 days, and a "Portfolio" pool for correlated-risk contracts. This recognizes that new, illiquid markets are far more likely to generate bad debt than established ones.
OKX learned from the 2018 clawback incident and switched from pure clawback to ADL, while also implementing tiered position limits and enhanced margin requirements for large positions. Their insurance fund has a specific ADL trigger: if the fund drops by 30% from its daily peak within 8 hours, ADL activates.
dYdX represents the decentralized approach — the insurance fund and ADL mechanics are built into the protocol, and the fund balance is on-chain and auditable by anyone. The tradeoff is that the fund is relatively small compared to centralized exchanges.
Hyperliquid pioneered the community-funded model with HLP, where anyone can deposit into the vault that serves as both market maker and insurance backstop. This democratizes insurance fund participation (depositors earn fees and trading profits) but also exposes depositors to tail risk from liquidation absorption.
9. Transparency and Reporting
Insurance fund transparency is one of the most underdeveloped aspects of exchange infrastructure. Consider the questions traders should be able to answer but typically cannot:
- How large is the insurance fund relative to total open interest?
- What was the fund's drawdown during the last major crash?
- How much of the fund's growth comes from liquidation surplus vs. exchange contributions?
- Has the fund ever been used to cover losses, and if so, how much?
- Is the fund segregated from exchange operating capital?
- Can the exchange access the fund for purposes other than covering bad debt?
The Current State
Mostly opaque: Most centralized exchanges publish a daily insurance fund balance (Bybit, OKX) but provide no breakdown of inflows, outflows, or drawdowns by event. You can see the fund is $X today, but you can't see that it was drawn down by $Y during last week's volatility and replenished by $Z from liquidation surplus.
Binance's SAFU: The SAFU fund wallet address is publicly visible on-chain, and Binance publishes proof-of-reserves. However, the per-contract insurance funds for futures are less transparent.
Decentralized exchanges: dYdX v4 and Hyperliquid have fully on-chain insurance funds. Every inflow, outflow, and balance change is publicly auditable. This is the gold standard for transparency.
What Good Transparency Would Look Like
A truly transparent insurance fund would publish:
- Real-time balance by contract/market
- Daily report of inflows (liquidation surplus, fee contributions) and outflows (bad debt coverage)
- Stress test results — how the fund would perform under historical worst-case scenarios (Black Thursday, FTX collapse)
- Drawdown history — every event that caused a drawdown, with amount and context
- Fund ratio — insurance fund balance as a percentage of total open interest (a key health metric)
- Independent audits — third-party verification that the reported balance actually exists
The Proof-of-Reserves Parallel
The crypto industry has (slowly) adopted proof-of-reserves for customer deposits after the FTX collapse. The same logic applies to insurance funds. If an exchange claims a $500M insurance fund, traders should be able to verify that $500M actually exists and is actually reserved for covering bad debt — not sitting on the exchange's balance sheet as an operational asset.
10. Insurance Fund Design Tradeoffs
Insurance fund design is fundamentally about balancing competing objectives. Every design choice involves a tradeoff.
Tradeoff 1: Fund Size
Too small Too large
◄─────────────────────────────────────►
│ │
│ Frequent ADL/clawback │ Exchange extracting value
│ Traders lose confidence │ Aggressive liquidations
│ Volume migrates to competitors │ Opaque "profit center"
│ │
│ Sweet spot │
│ ◄────────► │
│ Survives moderate crashes │
│ Occasionally draws down │
│ Transparent about status │The ideal fund is large enough to absorb all but the most extreme events, but not so large that it suggests the exchange is over-liquidating traders to build it. There is no universally correct size — it depends on the exchange's open interest, leverage offerings, market depth, and risk appetite.
Tradeoff 2: Liquidation Aggressiveness vs. Fund Growth
More aggressive liquidation (triggering earlier, wider spread between liquidation and bankruptcy price) generates more surplus for the fund but is worse for traders. A trader who would have survived a temporary drawdown gets liquidated unnecessarily, and the surplus from their liquidation inflates the fund.
Less aggressive liquidation (liquidating closer to bankruptcy) is better for traders but generates less surplus, meaning the fund grows slowly and may be insufficient during a crash.
Tradeoff 3: ADL Ranking Methodology
Ranking by profit + leverage (most common): Targets the biggest winners. Fair in the sense that these traders benefited most from the move. Unfair in the sense that they may have been trading conservatively and just happened to have a large position.
Ranking by leverage only: Targets the most leveraged traders regardless of profitability. More aligned with risk contribution but may hit traders who aren't actually profitable.
Ranking by margin fraction (Backpack's approach): Targets traders with the lowest margin fraction first. This means the most leveraged traders relative to their equity are hit first, regardless of profitability. This has the elegant property that the traders most likely to be liquidated themselves are also the first to be ADL'd — it's almost a preemptive liquidation.
Tradeoff 4: Centralized Fund vs. Community Fund
| Centralized Fund | Community Fund (e.g., Hyperliquid HLP) | |
|---|---|---|
| Control | Exchange manages, opaque | Open participation, transparent |
| Funding | Liquidation surplus + fees | Depositors provide capital |
| Risk bearing | Exchange absorbs tail risk | Depositors absorb tail risk |
| Incentive alignment | Exchange may profit from fund growth | Depositors earn fees but face drawdowns |
| Transparency | Varies (often low) | High (on-chain) |
| Manipulation risk | Low (exchange controls) | Higher (whales can target vault) |
Tradeoff 5: Insurance Fund vs. Exchange Equity
Some exchanges (like Deribit during Black Thursday) backstop the insurance fund with their own capital when it gets low. This is noble but creates moral hazard — if the exchange is the implicit backstop, why does the insurance fund need to be large at all? And if the exchange faces its own solvency issues (as happened with FTX), the implicit backstop evaporates precisely when it's needed most.
The honest answer is that no insurance fund can protect against all scenarios. A 50% intraday crash in a market with 100x leverage available will generate bad debt that exceeds any reasonable insurance fund. The fund's goal is not to eliminate socialized loss entirely — it's to make it rare enough that traders can trust the system under normal and moderately extreme conditions.
Key Takeaways
Insurance funds are the bridge between liquidation and bankruptcy. They exist to ensure that winning traders get paid even when losing traders' collateral is insufficient.
The funding mechanism creates a natural cycle: accumulation during calm markets, drawdown during crashes. This cycle is unavoidable — the question is whether the fund can survive the drawdowns.
ADL is strictly better than clawback for fairness, but both are painful for the traders on the receiving end. The best insurance fund is one that makes ADL rare.
Historical events prove these mechanisms matter. OKEx 2018 clawback, dYdX 2023 insurance drain, and multiple Hyperliquid HLP incidents show that insurance fund design has real consequences for real traders.
Transparency is the next frontier. On-chain insurance funds (dYdX, Hyperliquid) set a standard that centralized exchanges should aspire to. Proof-of-reserves for insurance funds should be as standard as proof-of-reserves for customer deposits.
There is no perfect size. Too small means frequent socialization. Too large means extraction. The right answer depends on the exchange's specific risk profile — and should be publicly justified, not quietly accumulated.
For a deeper look at the liquidation mechanics that feed into insurance funds and the cascading dynamics that stress-test them, see Liquidation Cascades.