What Matters
A liquidation bot does not get paid for predicting that a borrower is risky. It must identify an actually eligible position, submit a valid transaction and win enough execution priority for the liquidation bonus to exceed gas, failed attempts and inventory risk. The opportunity is therefore an execution problem as much as a model problem.
DN Evidence Block
Reviewed: 14 November 2026. Method: current primary-source documentation plus DN operational analysis. Evidence boundary: no live-money performance test is claimed unless explicitly stated. Synthetic examples and worksheet fields are illustrative.
- Aave defines health factor below 1 as liquidation eligibility.
- Aave describes liquidation as permissionless and highly competitive, with bots monitoring collateral, debt, oracle changes and protocol parameters.
- Newer Aave documentation also emphasizes previewing position health before actions and redesigned liquidation mechanics in V4.
The decision behind the headline
A technical, non-hype guide to DeFi liquidation bots: health factors, oracle changes, gas competition, failed transactions and execution economics.
The useful question is not whether an AI system or automation can perform the task once. It is whether the workflow preserves identity, permissions, timing, account state, failure handling and a recoverable record when conditions change.
How to build the workflow safely
- 1. Read the protocol-specific liquidation threshold rather than using a generic leverage rule.
- 2. Track oracle and debt updates at the cadence relevant to the protocol.
- 3. Confirm the exact close-factor or liquidation-size rule for the deployed version.
- 4. Estimate bonus minus gas, priority fees and failed attempts.
- 5. Model inventory and swap risk if seized collateral must be converted.
- 6. Simulate the transaction immediately before submission.
- 7. Record reverted and outbid attempts as real costs.
- 8. Treat protocol upgrades as a trigger for strategy revalidation.
DN Liquidation Economics Worksheet
Use this as a pre-deployment or incident-review record. Do not fill missing evidence with assumptions.
| # | Field | What to record |
|---|---|---|
| 1 | Protocol/version | Record evidence or mark unknown. |
| 2 | Health factor | Record evidence or mark unknown. |
| 3 | Eligible debt | Record evidence or mark unknown. |
| 4 | Liquidation bonus | Record evidence or mark unknown. |
| 5 | Gas / priority fee | Record evidence or mark unknown. |
| 6 | Expected failure rate | Record evidence or mark unknown. |
| 7 | Collateral conversion cost | Record evidence or mark unknown. |
| 8 | Net expected margin | Record evidence or mark unknown. |
DN instrument version 1.0.0. Local worksheet; no credential collection and no claim of independent certification.
What this framework does not prove
Passing a checklist does not prove profitability, legal compliance, security or future performance. It shows that the specific evidence fields were reviewed. Platform rules, protocol parameters and regulations can change, and the current account state always outranks an old screenshot or cached response.
Methodology and limitations
DN reviewed the primary sources listed below on 14 November 2026 and translated the documented mechanics into a decision record. Provider documentation establishes documented features and rules, not independent real-world performance. Where an example is used, it is illustrative and should not be read as observed market data.
Frequently Asked Questions
What is the main risk this article addresses?
A liquidation bot does not get paid for predicting that a borrower is risky. It must identify an actually eligible position, submit a valid transaction and win enough execution priority for the liquidation bonus to exceed gas, failed attempts and inventory risk. The opportunity is therefore an execution problem as much as a model problem.
Does this framework prove a strategy will be profitable?
No. The framework is an operational and evidence check. It does not predict returns or guarantee safe execution.
Should unknown inputs be treated as zero?
No. Unknown evidence should remain unknown until it is verified; zero is a measured value.
Is this financial advice?
No. It is operational research and educational material. Trading and automated finance can lose money.
Primary sources
Disclosure: Decentralised News may use commercial links in related coverage. Editorial conclusions here are independent of commercial placement. Cryptocurrency trading and automated finance involve risk and can result in loss.
