The Real Economics of DeFi Liquidations (Measured, Not Hyped)
I built a production, multi-chain liquidation bot — Rust + Solidity, 570+ tests, live 24/7 across Ethereum, Base and Arbitrum. Then I measured what actually makes money. Here's the honest answer most "passive crypto income" threads won't give you, with the data.
The myth
"Run a liquidation bot, earn passive yield." It's repeated everywhere. After weeks of live measurement across five protocols (Aave V3/V4, Euler V2, Ajna, Compound, Fluid) and three chains, the honest verdict for a solo is: the steady-state income is small, lumpy, and structurally capped — not because the engineering is hard, but because of how the market is shaped.
Reality 1 — Flash loans make it capital-light, which is the trap
Liquidations are funded by atomic flash loans: borrow, liquidate, swap, repay — in one transaction. Great, you don't need capital. But that means earnings are limited by opportunity flow and competition, not by capital. Adding money doesn't scale returns (the flash loan already covers size), so the classic "reinvest and compound" model does not apply. The bot earns what the flow allows, regardless of bankroll.
Reality 2 — On established venues you race funded firms, and "fast" isn't the lever
Liquidations are won in two steps: detection and inclusion. I measured Aave V4 on mainnet: entrenched bots took the short 1–3 block windows with a leader share around 0.67; smaller players only won the long-tail windows the leaders ignored.
The non-obvious part: submitting faster doesn't win. Ethereum block builders order by priority fee, not arrival time — you can send 500 ms earlier and still lose to someone who bid 1 gwei more. It's a fee/orderflow auction, not a latency race. A quick VPS changes nothing.
Reality 3 — The "uncontested" niches are real but thin
- Ajna runs a 72-hour time-Dutch auction — no inclusion race; the winner prices the strike best. The catch: thin flow. My observer saw a profitable take sit open ~5 hours with no competition — then days with almost no auctions at all.
- The "first-responder" thesis — be first on new markets using a pull oracle (Pyth/RedStone, self-triggerable) with no OEV protection. I built a live scanner across Ethereum, Base and Arbitrum. Result: pull-oracle share ≈ 0% — Chainlink (push) dominates the fresh markets. The uncontested-by-design niche mostly wasn't there.
Reality 4 — OEV is hyped harder than it pays (for a solo)
- Pyth Express Relay — live, but its EVM side is effectively dormant: the mainnet feed served only Solana opportunities. Zero EVM flow.
- Chainlink SVR / Oval — real and sizeable, but ~65–90% is recaptured by the protocol, and you compete via Flashbots against the same firms. Thin residual.
- API3's public OEV network was being wound down to partnered searchers only.
OEV genuinely removes the firm advantages (orderflow + the inclusion race) — but the accessible, EVM, solo-friendly version mostly doesn't exist yet.
So where's the money?
Honestly: a well-built solo liquidation stack earns a few hundred to low-thousands a month at best, lumpy — plus the rare fat window on a brand-new market before the firms arrive. A supplement, not a salary, and you can't compound out of it with a flash-loan bot.
The durable value isn't the trade income. It's the infrastructure, the live data, and the knowledge of where the edges actually are and aren't — which is what lets you build the right bot, fast, instead of chasing a myth.