DeFi Liquidators: How Much of the Bonus Becomes Profit?

The signal

The commercial edge to investigate is conversion efficiency: how much of a documented incentive survives financing, execution and unsuccessful competition. A dashboard showing only successful liquidation bonuses cannot answer that question.

Define the opportunity before estimating its margin

Begin with a specific chain, protocol version, market, borrower, debt asset, collateral asset and block. The amount a borrower owes is not necessarily the amount a particular liquidation can repay. A calculator that starts with the entire loan without checking the permitted transaction size can exaggerate both the bonus and the capital requirement.

Aave’s current help page describes a 50% debt limit when health factor exceeds 0.95 and both collateral and debt are at least $2,000; it describes up to 100% when those conditions fail. It also specifies minimum residual-balance rules. These are documented conditions, not a licence to assume identical behaviour across every Aave deployment. [1]

Aave V4’s overview instead describes repayment towards a Spoke-specific target health factor and a bonus that varies with health factor. [2] The immediate analytical consequence is that a strategy must retrieve the applicable rules before forecasting a trade. A familiar protocol name is insufficient identification.

This scope excludes trading perpetual futures around anticipated liquidation clusters. Here the operator actually repays a lending position’s debt and receives its collateral. Combining the two activities in one performance series would mix different cash flows and risks.

Separate the gross incentive from the liquidator’s share

Aave V3’s view-contract reference exposes a liquidation protocol fee on the bonus. Its LiquidationLogic source derives the bonus component and deducts the protocol’s fee from collateral delivered to the liquidator. [3] [4] A 10% share of a 5% bonus is therefore different from a 10% charge on the full collateral value.

For the simplified example, repaying $100,000 with a 5% gross bonus produces $105,000 of collateral at the assumed reference valuation. A protocol share of 10% of the $5,000 bonus removes $500. The liquidator receives $104,500 before sale costs. These are synthetic inputs, not current parameters for a named Aave reserve.

The cash-flow bridge should preserve the fee’s denominator. If the contract uses a different fee basis, the tool is not applicable without adjustment. Token decimals, integer rounding, collateral availability and repayment constraints can also make the actual settlement differ from the continuous arithmetic used here.

A good accounting record shows debt paid, collateral units received and any protocol deduction separately. Recording only the final wallet value makes it harder to distinguish a parameter change from poor execution.

Morpho illustrates why the exact formula matters

Morpho Blue’s documentation permits full or partial liquidation once a position breaches its market’s liquidation loan-to-value threshold. It gives an incentive factor of the smaller of 1.15 and 1 divided by (0.3 × LLTV + 0.7), and states that the protocol takes no share of that incentive. [5]

The core source independently exposes the corresponding calculation, with fixed-point rounding and a current-oracle health check. [6] Applying the documented formula to an assumed 86% LLTV gives a factor of approximately 1.04384: a gross uplift of about 4.384%. That is DN arithmetic before costs, not an observed trade or annual yield.

The documentation’s worked example uses approximately 5% at that threshold. The difference matters when the remaining execution margin is small. We use the stated formula, cross-checked against the source structure, rather than carry that rounded example into the calculator.

The comparison does not identify a superior venue. A smaller nominal incentive with a deeper executable exit can be more attractive than a larger incentive with difficult collateral. A fee-free incentive split also does not make financing, transaction inclusion or inventory management free. Those costs sit elsewhere in the bridge.

Collateral valuation is not the sale receipt

The central stress variable is the gap between the collateral’s reference value and the cash or debt tokens actually recovered. Apply that gap to all collateral sold. Applying a 1% haircut only to a $5,000 bonus would materially understate the cost of selling a position worth more than $100,000.

In the synthetic example, the $104,500 received after the protocol share becomes $103,455 after a 1% all-in sale discount. The $1,045 reduction is larger than many small execution fees. The calculator combines price impact, swap charges and other sale friction into this single discount; users should not enter the same cost again in another field.

The reference and sale valuations must use the same unit and time convention. If debt repayment is valued in dollars while sale proceeds are measured in a volatile token, convert both consistently. A stablecoin symbol is not itself a guarantee that the dollar value remains fixed.

An operator who retains collateral has not yet demonstrated the modelled sale proceeds. Report realised operating results separately from inventory mark-to-market. Retention can be a rational strategy, but subsequent price gains or losses should not be mistaken for evidence that the liquidation execution itself was efficient.

Financing and unsuccessful attempts belong in the same ledger

Aave’s flash-loan guide says the current premium should be read from the contract because governance can change it. It also explains that a flash-loan route requiring repayment must complete that repayment within the transaction or revert. [7] The tool’s 5-basis-point financing input is hypothetical and must be replaced with the actual route’s charge.

An atomic transaction can protect the lending operation from an incomplete repayment, but it does not make every unsuccessful attempt costless. Ethereum’s gas documentation distinguishes rejection before inclusion from execution that consumes gas before reverting. [8] Charge the latter according to receipts; do not bill every submitted or replaced request as if it was mined.

The example assigns $100 to the successful transaction and $20 to each of three included failed attempts. It adds $100 of other attributable costs and $50 of financing. Total costs beyond the sale discount and protocol share are therefore $310. Subtracting debt repayment leaves $3,145.

Own-capital financing can be entered as zero for a direct cash-margin calculation, but that does not establish an attractive return on committed capital. Idle balances, hedging and infrastructure may still have economic costs. State whether the ledger measures contribution margin, fully allocated business profit or capital efficiency; none is interchangeable with the other.

Use the break-even discount to test a proposed trade

The calculator follows one successful liquidation plus the entered number of associated failed attempts. It first calculates collateral remaining after the protocol’s assumed share of the bonus. It then applies the sale discount, subtracts repaid debt and financing, and deducts execution and other attributable costs.

With the default inputs, the break-even sale discount is approximately 4.01%. That is the maximum all-in reduction from the model’s post-fee collateral value before contribution falls to zero, holding every other input fixed. It is not a forecast of available liquidity or a safe slippage setting.

Increasing the sale discount from 1% to 5% turns the example into a $1,035 loss. The gross incentive is unchanged. This is why headline bonus comparisons can reverse once an executable exit is considered. The same output could deteriorate if a smaller permitted repayment leaves fixed costs spread across fewer dollars.

A negative break-even discount means the model already loses money even with no sale discount; it would need a sale premium or lower costs. The tool displays that result rather than replacing it with zero. A zero-cost, zero-bonus boundary case returns zero contribution, which is a mathematical check rather than a plausible business forecast.

A real performance panel must include the losing attempts

DN’s proposed empirical unit is an operator’s complete opportunity cycle. Join the market and borrower identifier to transaction hashes, debt units repaid, collateral units received, protocol deductions, actual sale receipts and native-token fees. Preserve the block and conversion price used for each cash-flow valuation.

Join unsuccessful included transactions to the same strategy ledger where attribution is defensible. Do not silently discard them because they lack a successful liquidation event. Conversely, do not assign a rival operator’s failed transaction to the winner’s costs. A public chain can reveal activity without proving that two addresses belong to the same business.

The hardest missing fields can be private inclusion payments, off-chain infrastructure expense, hedging and unsold inventory. Mark them unavailable rather than infer zero. Revenue calculated from an event’s oracle valuation should be labelled reference-value revenue, not realised cash.

Selection bias also starts before the transaction. An operator may monitor thousands of opportunities and act on very few. The share of attempts that succeed, the size of successes and the capacity to repeat them all affect the business. This article has not collected that dataset; the bridge specifies what would make a future performance claim credible.

The strongest countercase is an efficient, repeatable operator

A well-capitalised operator with reliable infrastructure, inexpensive financing and a liquid exit may retain much more of the incentive than the stressed example. Partial liquidation can also reduce market impact where the applicable rules allow it. The correct conclusion is not that liquidation is inherently unprofitable.

However, the operator’s advantage must survive competition and changing position size. A backtest that awards every eligible position to the same bot without modelling its execution record is measuring theoretical availability. A realised record should demonstrate both the successful settlement and the costs needed to obtain it.

For a prospective operator, the decision is whether the expected conversion margin justifies development, monitoring and capital commitments. For a protocol researcher, it is whether incentives still cover the difficult collateral and stressed execution conditions that matter most. For a borrower, it is a reminder that eligibility rules and collateral-sale conditions can interact sharply during a market move.

The most useful next step is a small reconciled sample from the exact deployment, with successes and attributable failures reported together. A larger collection of unreconciled bonus events would add apparent precision while leaving the central economic question unanswered.

The evidence, side by side

DN Liquidation Cash-Flow Bridge

Synthetic one-success cycle, including the entered failed attempts; not live protocol parameters, a profitability forecast or a backtest.

Finite bounded inputs; repayment positive; bonus, protocol share and sale discount between 0% and 100%; failed attempts a nonnegative whole count; all costs nonnegative.

Contribution excludes tax and overhead unless explicitly entered. No model of win probability, valid liquidation size, live liquidity, inclusion or annualisation. Negative break-even discount is retained: a premium or lower costs would be required.

Scenarios to watch

Default hypothetical cycle

$100,000 repaid; 5% bonus; 10% protocol share of bonus; 1% sale discount; $310 other modelled costs.

$3,145 contribution; approximately 4.01% break-even sale discount.

Collateral exit worsens

Default inputs except 5% sale discount.

$1,035 loss despite the same $5,000 gross bonus.

Zero-friction mathematical boundary

No bonus, no discounts and no costs.

Zero contribution; not a practical expected outcome.

Method and limitations

We read official Aave help, V3 contract references and source, V4 documentation, Morpho Blue documentation and source, and Ethereum gas guidance. This is a mechanism comparison, not a live market or operator performance sample.

Morpho’s 86% LLTV illustration is recalculated from the published formula rather than its approximately 5% prose example. Integer rounding and actual settlement units are outside the simplified calculator.

All cash-flow defaults are synthetic. Runtime checks cover a profitable default, a higher-sale-discount loss, a zero-friction boundary and invalid zero repayment.

Documentation and repository main branches do not establish the code or parameters active at every deployed address. Borrower eligibility, collateral availability and permitted repayment must be verified before applying the bridge.

The model assumes a completed collateral sale and consistent valuation units. It does not estimate opportunity frequency, winning probability, debt depegs, inventory gains, tax or hidden costs.

Primary sources and evidence register

Source dates and reporting periods differ. Several documents from one issuer are one source family, not independent confirmations. The evidence download contains the observations used in this article.

Frequently asked questions

Is the liquidation bonus the return on my capital?

It is an input to the collateral exchange. Capital return also depends on realised sale proceeds, costs, capital committed and time; the tool does not annualise it.

Why apply the discount to all collateral?

The sale covers the collateral received, including the value replacing repaid principal. Discounting only the bonus leaves most sale friction out of the calculation.

Should every failed submission count as a paid attempt?

No. Use costs actually incurred. An unmined request and an included transaction that reverts are different accounting events.

Does the comparison recommend Aave or Morpho?

No. It identifies different rules that a deployment-specific analysis must use. Neither the documentary panel nor the synthetic example establishes which route offers the best realised economics.