CDP liquidation ratios: 4 factors driving DeFi health
A collateralized debt position does not become dangerous only when its collateral falls in price.

The liquidation ratio is the visible boundary, but the actual solvency of a decentralized stablecoin system depends on the interaction between collateral quality, oracle design, gas costs, and liquidation incentives.
A vault with a 150% collateralization ratio can still become undercollateralized if its oracle reports late, liquidators cannot execute during a gas spike, or the market cannot absorb the collateral being sold. Conversely, a protocol with a lower minimum ratio can remain solvent if it has faster liquidation paths and a stronger debt-absorption mechanism. Liquity, for example, allows LUSD to be minted at a minimum collateralization ratio of 110% and uses a Stability Pool rather than relying only on traditional keeper auctions.
The practical question is therefore not simply how much collateral a protocol requires. It is how the entire liquidation machine behaves when prices move faster than the smart contracts, keepers, and liquidity venues can respond.
The liquidation ratio is a solvency boundary, not a safety guarantee
A collateralized debt position allows a user to deposit crypto assets and mint a decentralized stablecoin against them. The protocol records three principal values:
- the market value of the collateral;
- the outstanding stablecoin debt;
- the liquidation threshold or minimum collateralization ratio.
The basic collateralization formula is:
Collateralization ratio = collateral value ÷ debt value × 100
If a user deposits $150 worth of ETH and mints $100 of DAI, the position is collateralized at 150%. If ETH falls and the collateral value declines to $140, the ratio becomes 140%. Whether the vault remains active depends on the protocol’s minimum requirement.
In a standard MakerDAO ETH vault, the minimum collateralization ratio is 150%. A position below that level becomes eligible for liquidation under the relevant vault parameters. That does not mean the protocol waits for collateral to reach zero. The system is designed to act before the collateral value equals the debt, because liquidation requires time, execution capacity, and a penalty buffer.
A simpler model has three states:
1. Healthy position: the collateralization ratio remains above the liquidation threshold, with sufficient room for normal market volatility.
2. Liquidatable position: the ratio falls below the protocol-defined boundary, allowing the liquidation process to begin.
3. Underwater position: the collateral value no longer covers the debt and associated liquidation costs.
Different protocols express this condition differently. Lending markets often use a health factor, with Health Factor < 1.0 indicating that a position has become liquidatable. CDP systems may instead compare the collateralization ratio directly with a minimum percentage.
The distinction matters because a liquidation ratio is a protocol parameter, while health is a dynamic property. A vault at 151% in a 150% system is technically active but operationally fragile. A modest price movement, an oracle update, or an accrued fee can push it across the boundary.
The liquidation ratio defines when the protocol may act. It does not define how safely the protocol can act.
1. Collateral quality determines how quickly the ratio deteriorates
The first driver of CDP liquidation risk is the collateral asset itself. A protocol that accepts ETH, wrapped tokens, liquid staking derivatives, or thinly traded governance assets is not managing the same risk across all vault types.
The liquidation ratio must compensate for the volatility and liquidity profile of each collateral class. A highly liquid asset with deep spot and derivatives markets can generally be liquidated more efficiently than an asset whose price depends on a small number of decentralized exchange pools. But even ETH is not a uniform risk input. Network congestion, staking-related basis movements, and correlations across DeFi collateral can change the liquidation path during a market decline.
The relevant mechanism is straightforward:
- If collateral volatility rises, the time between a healthy ratio and a liquidatable ratio becomes shorter.
- If market depth falls, liquidators receive less predictable execution.
- If the collateral is correlated with other assets in the protocol, several vault categories may approach liquidation simultaneously.
- If collateral is itself a derivative token, the oracle must capture both the underlying asset price and the possibility of a discount or depeg in the derivative.
This is why typical CDP collateralization ratios can range from roughly 110% to 200% or more depending on the protocol and asset. Liquity’s 110% minimum applies within a specific design that combines its collateral rules with the Stability Pool. MakerDAO’s standard ETH vault requirement is materially higher at 150%. Neither percentage can be evaluated in isolation from the liquidation architecture behind it.
The same ratio can represent different levels of risk
Consider two vaults, each with a 150% collateralization ratio.
The first is backed by an asset with deep liquidity and an oracle that updates promptly across multiple sources. The second is backed by a volatile token whose primary market is a shallow DEX pool. Their numerical ratios are identical, but their liquidation risk is not.
The second vault faces a wider execution gap. The market price may move materially between the moment a position becomes liquidatable and the moment a liquidator acquires the collateral. The liquidation penalty may not cover that gap. If the protocol absorbs the debt through an auction, the auction may clear below the expected valuation. If it uses a pool, the pool may be forced to absorb an unusually large inventory.
The important variable is not only the distance to the threshold. It is the distance relative to the asset’s expected price movement during the liquidation window.
For protocol analysts, collateral quality includes:
- spot-market depth across the venues used by liquidators;
- historical and stress-period volatility;
- correlation with other accepted collateral;
- redemption or unwrap liquidity for derivative assets;
- exposure to bridge, smart contract, and governance risk;
- the ability to price the asset when primary markets are impaired.
Collateral that looks productive in stable conditions can become a source of balance-sheet risk when many positions need to exit at once.
2. Oracle design turns market prices into liquidation decisions
A liquidation engine cannot act on an abstract market price. It needs an oracle value that determines whether the collateralization ratio has crossed the threshold. This creates the second major risk: the quality and timing of price information.
The oracle is part of the liquidation mechanism, not an external reporting layer. If the oracle updates too slowly, unsafe positions remain active while collateral value is falling. If it updates too aggressively or relies on a manipulable venue, healthy positions can be liquidated at an artificial price.
The failure modes are different but economically similar. In both cases, the protocol makes an incorrect decision about solvency.
Latency creates a timing gap
Suppose ETH falls sharply in external markets, but the protocol’s oracle updates with a delay. During that interval, borrowers may continue to withdraw collateral or mint additional stablecoins against a value that no longer exists. Liquidators see the true market price but cannot yet trigger the position under the protocol’s accounting price.
When the oracle eventually updates, multiple vaults may become liquidatable at the same time. The system then faces a concentrated execution burden rather than a gradual flow of liquidations.
The opposite sequence is also possible. A temporary price shock reaches the oracle while the broader market has already recovered. If the oracle records the low before the recovery is reflected in the protocol, it can trigger premature liquidations of positions that would otherwise have remained solvent.
Manipulation changes the liquidation threshold itself
An oracle that draws heavily from a shallow DEX pool can be attacked through temporary liquidity distortion. Flash loans make it possible to borrow capital, trade against a pool, move its price, and unwind the transaction within one atomic execution. If the oracle treats that pool as a reliable price source without sufficient resistance to manipulation, the attacker may influence collateral valuation.
The consequences depend on the protocol’s permissions and update rules. A manipulated price can:
- force a healthy vault below its liquidation ratio;
- allow an attacker to mint more stablecoins than the collateral should support;
- delay liquidation by overstating collateral value;
- create an arbitrage opportunity between the protocol’s internal price and external markets.
Decentralized oracle design therefore requires more than multiple data feeds. It requires a clear definition of how prices are aggregated, how outliers are handled, how often values update, and what happens when the reference markets become unavailable.
A robust design may use time-weighted prices, multiple venues, delayed updates, circuit breakers, or governance-controlled emergency procedures. Each method introduces trade-offs. Time weighting reduces the effect of a single transaction but increases latency. Faster updates improve responsiveness but can expose the system to short-lived price manipulation. Circuit breakers reduce cascading liquidations but may leave the protocol temporarily unable to price new positions.
For anyone assessing makerdao cdp liquidation triggers or another DeFi lending protocol liquidation threshold, the oracle update path deserves the same scrutiny as the headline collateral ratio.
3. Gas fees determine whether liquidation is executable
The third factor is network execution. A liquidation can be perfectly specified in the smart contract and still fail economically if the transaction cannot be executed in time.
Liquidators typically pay gas first and receive compensation through a liquidation penalty, collateral discount, auction spread, or another protocol-defined reward. This incentive must cover the cost of execution and the risk that the transaction will fail, be outbid, or clear at an unfavorable price.
During ordinary market conditions, this mechanism may operate with a wide margin. During a rapid sell-off, the margin compresses precisely when the system needs liquidations most.
The sequence is mechanical:
1. The collateral price falls toward the liquidation threshold.
2. The oracle marks the position as liquidatable.
3. Multiple keepers submit liquidation transactions.
4. Network demand increases, raising gas prices and confirmation times.
5. Some transactions fail, remain pending, or become unprofitable.
6. Unsafe positions stay open while the collateral value declines further.
The resulting risk is not theoretical. On-chain gas fee spikes during market volatility can delay liquidator execution and leave a protocol undercollateralized if debt cannot be liquidated in time.
Liquidation profitability is a function, not a fixed reward
A liquidation penalty is intended to compensate the liquidator and protect the protocol. Across DeFi systems, penalties can range from approximately 0.1% to 15%, with many parameters falling in the 3% to 13% range. The correct level depends on collateral volatility, gas costs, market depth, and the liquidation method.
A low penalty may fail to attract keepers during a congested blockspace market. A high penalty may protect the protocol’s accounting but impose a steep cost on borrowers and encourage aggressive liquidation behavior near the threshold.
The penalty must also be evaluated against the size of the position. A 3% reward on a large vault may cover execution costs, while the same percentage on a small vault may not justify the transaction. If liquidators prioritize only the largest or safest opportunities, smaller unsafe positions can accumulate as residual risk.
This creates a practical distinction between liquidatability and liquidation capacity:
- A position is liquidatable when the smart contract permits action.
- A protocol has liquidation capacity when enough economically motivated actors can execute that action under current network and market conditions.
A system can satisfy the first condition while failing the second.
Auctions, pools, and continuous liquidation
The liquidation method changes how gas and liquidity risk are distributed.
Traditional auction systems transfer collateral to liquidators through a bidding process. They can discover a market price, but auctions require active participants and may become difficult to coordinate during a sharp market move.
Liquity uses a Stability Pool. Eligible depositors provide LUSD, and the pool can absorb debt and receive collateral when positions are liquidated. This removes some dependence on a sequence of external keeper auctions, although it creates its own requirements: the pool must have sufficient capital, and depositors must accept exposure to liquidated collateral.
Curve’s crvUSD uses LLAMMA, or Lending-Liquidating AMM Algorithm. Instead of waiting for one abrupt liquidation threshold, LLAMMA performs soft liquidation across price bands. As the collateral price moves through those bands, the mechanism progressively converts exposure between collateral and stablecoin liquidity.
This design changes the shape of the risk. A single cliff may be reduced, but the protocol still depends on the AMM’s liquidity, price-band calibration, and ability to rebalance while the market is moving. Continuous liquidation is not the absence of liquidation risk. It is a different execution model.
| Liquidation model | Primary execution path | Main advantage | Main stress risk |
|---|---|---|---|
| Keeper auction | External liquidators bid for collateral | Flexible price discovery and broad participation | Auctions may clear slowly or below expected value |
| Stability Pool | Deposited stablecoin absorbs debt and receives collateral | Reduces reliance on individual auction keepers | Pool capacity may be insufficient during correlated liquidations |
| Continuous AMM liquidation | Collateral is progressively rebalanced across price bands | Reduces the effect of a single liquidation cliff | AMM liquidity and band design may fail during rapid price moves |
| Direct collateral sale | Liquidator purchases collateral under defined contract rules | Simple and fast when markets are liquid | Execution price can deteriorate sharply in thin markets |
The operational question is always the same: who absorbs the collateral, at what price, and with what available capital?
4. Liquidation incentives decide whether bad debt is contained
The fourth factor is incentive design. A CDP protocol relies on independent actors to identify unsafe positions and perform the transaction. Those actors respond to expected profit, not to the protocol’s solvency objective.
The liquidation penalty is the most visible incentive, but it is not the only one. Keeper competition, auction rules, collateral discounts, gas reimbursement, and the ability to bundle transactions all influence whether liquidation occurs promptly.
If the expected reward exceeds gas and execution risk, liquidators compete to close the position. If expected revenue falls below those costs, the position may remain open even though the protocol marks it as unsafe.
This is especially important when collateral is volatile and liquidation penalties are capped. A protocol can limit the penalty to protect borrowers, but then it must accept that liquidators may require deeper market liquidity or lower position sizes to remain active during stress.
Liquidation penalties transfer risk; they do not remove it
When a vault is liquidated, the borrower typically loses collateral beyond the amount required to repay the debt. That additional amount is the liquidation penalty. It can be routed to the liquidator, the protocol treasury, a stability mechanism, or a combination of these destinations.
The penalty serves three functions:
- it encourages borrowers to maintain a buffer above the liquidation threshold;
- it compensates liquidators for gas and execution risk;
- it creates a first-loss margin against adverse price movement during liquidation.
But the penalty has limits. If the collateral falls faster than the liquidation process can execute, a 5% or 10% penalty may not cover the realized loss. If the market is manipulated, the penalty may instead amplify the damage by making a false liquidation more profitable.
The protocol must therefore align incentives with the full execution path. A keeper should be rewarded for closing a position when it is unsafe, but not for exploiting a temporary oracle error or extracting value from a market that has not actually crossed the economic solvency boundary.
Position concentration creates a second-order problem
Even with well-designed incentives, a protocol becomes more difficult to liquidate when debt is concentrated in a small number of large vaults. One position can exceed the capacity of available buyers, a Stability Pool, or an AMM band.
Large vaults also increase the chance that liquidation itself moves the market. A sale that is individually rational for the keeper can push the collateral price lower, causing other positions to fall below their thresholds. The protocol then enters a liquidation cascade.
This is one reason risk teams monitor not only total collateral but also:
- the largest individual debt positions;
- debt concentration by collateral type;
- the distribution of collateralization ratios;
- the amount of stablecoin liquidity available to absorb liquidations;
- the depth of the venues used for collateral disposal;
- the proportion of positions clustered just above the liquidation threshold.
A protocol with high aggregate collateralization can still be fragile if a large share of its debt is attached to narrowly buffered positions.
How the four factors interact during a market decline
The four drivers rarely fail independently. Stress usually moves through the system as a chain.
Assume an accepted collateral asset declines quickly. Volatility increases and market depth falls. The oracle updates, but with some latency. Once the new price is posted, a group of vaults crosses the liquidation threshold. Liquidators submit transactions at the same time, gas prices rise, and some executions fail. The remaining positions lose more value before they can be closed. If the liquidation penalty is too small to cover the new execution gap, keeper participation falls. The protocol then faces a larger debt shortfall.
The key variable is the time required to move from price recognition to debt absorption.
A system can tolerate substantial volatility if that time is short and the absorption mechanism has capacity. It can fail under a smaller price move if the oracle is slow, blockspace is congested, or the liquidation market is thin.
This also explains why historical performance in calm markets is an incomplete measure of protocol health. A CDP can maintain a stable peg and low liquidation losses for months while accumulating structural exposure to one collateral asset, one oracle route, or one keeper group.
For broader European market and policy context around financial infrastructure, English-language coverage of France can provide a separate macro reference point. It should not be confused with an on-chain solvency metric: the CDP’s immediate risk remains determined by collateral value, oracle execution, and the liquidation path encoded in the protocol.
A practical way to compare CDP designs
A useful evaluation begins with the liquidation ratio but does not end there. The following sequence exposes the main engineering dependencies.
1. Calculate the distance to liquidation.
Do not look only at the minimum ratio. Measure how much collateral value can be lost before the position becomes liquidatable, and compare that buffer with the asset’s expected intraday volatility.
2. Trace the oracle update path.
Identify which markets supply the price, how often the value changes, whether the system uses a time-weighted calculation, and what happens when one venue diverges from the others.
3. Estimate execution time under congestion.
A liquidation that settles in one block during normal conditions may require several blocks during a market shock. The protocol’s safety margin must cover that delay.
4. Map the liquidator’s economics.
Calculate the expected penalty or discount after gas, failed transactions, slippage, and capital costs. If liquidation is unprofitable for realistic position sizes, the contract’s permission to liquidate is not enough.
5. Identify the debt-absorption mechanism.
Determine whether the system uses auctions, a Stability Pool, an AMM, direct sales, or a combination. Then assess who provides liquidity when many positions become unsafe simultaneously.
6. Review concentration and correlation.
A portfolio of different collateral assets may still behave like one asset if those assets share the same market, bridge, oracle, or liquidity dependency.
7. Test the threshold against oracle error.
If a small deviation in the reported price can move a large volume of debt across the boundary, the protocol is exposed to premature liquidation or delayed recognition of bad debt.
This process produces a more accurate assessment than comparing headline ratios. A 110% system and a 150% system may have different failure modes rather than a simple ranking of safer and riskier designs.
Theoretical limits of the liquidation model
Every CDP system assumes that collateral can be valued and converted into debt repayment before losses exceed the available buffer. That assumption has a practical limit.
If the collateral market becomes illiquid, the oracle may report a price that cannot be realized at scale. If blockspace becomes unavailable, a liquidatable position can remain open while its collateral loses value. If all major collateral assets decline together, a Stability Pool may absorb debt but receive assets whose market value is falling faster than the pool can rebalance. If an AMM performs continuous liquidation, the process can still transfer losses through its price bands rather than eliminate them.
The model is therefore not designed to guarantee a fixed dollar value under every market condition. It is designed to make insolvency less likely by imposing overcollateralization and creating automated paths for loss realization.
That distinction matters for decentralized stablecoins. A collateral ratio is a pre-funded buffer. It is not an insurance policy, and it cannot compensate for an oracle that is wrong, a network that is congested, or a market that has no effective bid.
The stress-test questions that matter
A serious protocol review should ask:
- How far can the collateral price move between oracle updates?
- How many blocks can liquidators miss before the collateral buffer is exhausted?
- What percentage of debt can the Stability Pool, auction system, or AMM absorb in one event?
- Which collateral assets become correlated during a broad DeFi sell-off?
- Can a flash-loan attacker influence the price source used for liquidation?
- Does the liquidation penalty remain profitable after gas and slippage?
- What happens when the protocol’s stablecoin itself trades below its target value?
- Can governance change collateral parameters quickly enough to respond, without creating an additional centralization or execution risk?
The answers define the protocol’s actual liquidation capacity. The nominal ratio is only the starting parameter.
Conclusion
CDP liquidation ratio drivers in DeFi are structural rather than cosmetic. Collateral quality determines how quickly solvency deteriorates. Oracle design determines when the protocol recognizes that deterioration. Gas fees determine whether liquidators can act. Incentives determine whether they have a reason to act.
MakerDAO’s 150% standard ETH requirement, Liquity’s 110% minimum with Stability Pool absorption, and crvUSD’s LLAMMA soft-liquidation model illustrate three different approaches to the same problem: converting volatile crypto collateral into a stable unit of account without relying on a centralized balance sheet.
None of these approaches removes liquidation risk. They distribute it across different mechanisms. The relevant analysis is therefore conditional: if prices move this fast, if the oracle updates this way, and if network costs reach this level, can the protocol absorb the resulting debt before collateral value disappears?
That is the engineering boundary behind every decentralized stablecoin peg.