Methodology

Indexed contracts

• Aave V3 Pool: 0x87870Bca3F3fD6335C3F4ce8392D69350B4fA4E2
• SparkLend Pool: 0xC13e21B648A5Ee794902342038FF3aDAB66BE987
• Morpho: 0xBBBBBbbBBb9cC5e90e3b3Af64bdAF62C37EEFFCb
• Fluid Vault Factory: 0x324c5Dc1fC42c7a4D43d92df1eBA58a54d13Bf2d

Indexed events

Aave V3 and SparkLend share an identical LiquidationCall event (Spark is an Aave V3 fork):

event LiquidationCall(
  address indexed collateralAsset,
  address indexed debtAsset,
  address indexed user,        // borrower
  uint256 debtToCover,
  uint256 liquidatedCollateralAmount,
  address liquidator,
  bool receiveAToken
)

Morpho uses a different shape. Each liquidation carries a bytes32 market ID rather than token addresses. We call idToMarketParams(id) on the Morpho contract once per market to resolve the loan/collateral tokens, oracle, and LLTV, then cache the result:

event Liquidate(
  bytes32 indexed id,          // market id, resolves to tokens
  address indexed caller,       // liquidator
  address indexed borrower,
  uint256 repaidAssets,
  uint256 repaidShares,
  uint256 seizedAssets,
  uint256 badDebtAssets,        // unique to Morpho: socialised bad debt
  uint256 badDebtShares
)

Morpho also uniquely tracks bad debt, which is the portion of a liquidation the liquidator couldn't cover and which gets socialised among lenders. Aave and Spark can't have bad debt by design (they always require excess collateral).

Fluid has a different architecture still. Rather than a single pool contract, Fluid deploys a separate vault contract for every collateral/debt pair. We discover vault addresses by scanning the factory's NewPositionMinted event (the minter field is the vault address), then call constantsView() on each vault to resolve its supply and borrow tokens. Each vault emits its own LogLiquidate event:

event LogLiquidate(
  address liquidator_,
  uint256 actualColAmt_,   // collateral seized from the vault
  uint256 actualDebtAmt_,  // debt repaid on the vault's behalf
  address to_              // recipient of the seized collateral
)

Block range

Aave V3 from block 16,291,127 (Jan 27, 2023). SparkLend from block 17,185,580 (May 2023). Morpho from block 18,883,124 (Dec 27, 2023). Fluid from block ~19,500,000 (April 2024). All protocols are continuously indexed up to the current block.

Oracle contracts

• Aave V3 Oracle: 0x54586bE62E3c3580375aE3723C145253060Ca0C2
• SparkLend Oracle: 0x8105f69D9C41644c6A0803fDA7D03Aa70996cFD9

Gross profit

collateral_usd = (liquidatedCollateralAmount / 10^collateral_decimals) * oracle_price
debt_usd       = (debtToCover / 10^debt_decimals) * oracle_price
gross_profit   = collateral_usd - debt_usd

Token decimals are read on-chain from each ERC20 contract's decimals() function and cached in our database. We do not hardcode 1e18. USDC uses 6, WBTC uses 8, etc.

Net profit

gas_cost_eth = gasUsed * effectiveGasPrice / 1e18
gas_cost_usd = gas_cost_eth * eth_price_at_block
net_profit   = gross_profit - gas_cost_usd

Gas data is fetched from the transaction receipt (eth_getTransactionReceipt) for each liquidation. ETH price for converting gas to USD is read from the Aave oracle at the same block.

Liquidation bonus efficiency (per asset)

avg_bonus_pct = AVG((collateral_usd - debt_usd) / debt_usd) * 100

The bonus efficiency table on the Insights page shows the average effective liquidation bonus extracted per collateral asset. This will usually be slightly below the protocol's configured bonus (typically 5–10%) because liquidators sometimes liquidate less than the maximum allowed amount.

Daily aggregation

-- Each day: event count, total volume, profit, biggest single liq,
-- unique liquidators, unique borrowers, top protocol by volume
GROUP BY DATE(TO_TIMESTAMP(block_timestamp))

ETH price source

ETH prices come from our price_cache table, which stores hourly WETH prices fetched from DeFiLlama's historical prices API. We average these hourly prices per day and LEFT JOIN them onto the daily liquidation data. Days with no cached price show as gaps in the ETH price line.

ETH day-over-day change

For each Rekt Day in the "Hall of Rekt" table, we compute the ETH price change as a percentage relative to the previous day's average ETH price. This is calculated server-side by iterating the sorted daily array.

Top 10 ranking

ROW_NUMBER() OVER (ORDER BY total_volume DESC) as rekt_rank
-- Days with rekt_rank <= 10 are marked as is_top_rekt = true

Pairwise overlap detection

-- Self-join: find liquidators active on BOTH protocol A and B
-- a.protocol < b.protocol prevents double-counting pairs
FROM liquidator_proto a
JOIN liquidator_proto b
  ON a.liquidator = b.liquidator
  AND a.protocol < b.protocol

The diagonal of the matrix shows total unique liquidators per protocol. Off-diagonal cells show the count of addresses active on both protocols. Hover tooltips show combined volume and profit for shared liquidators.

Multi-protocol activity breakdown

We also count how many liquidators operate on exactly 1, 2, 3, or all 4 protocols. This uses array_length(array_agg(DISTINCT protocol), 1) per liquidator address.

Liquidation size distribution

Individual liquidation events are bucketed by their collateral seized amount (USD) into 8 ranges: $0–$100, $100–$1K, $1K–$10K, $10K–$50K, $50K–$100K, $100K–$500K, $500K–$1M, and $1M+.

Monthly profit (gross profit flooring)

The monthly profit chart sums GREATEST(gross_profit_usd, 0) per event. Negative gross profit values (which occur when collateral_usd < debt_usd due to oracle timing mismatches) are floored to zero so they don't distort the monthly aggregate. This is a conservative choice — the alternative would be to subtract these losses, which would understate liquidator earnings.

Profit concentration snapshot

Liquidators are ranked by total gross profit and grouped into tiers: Top 5, Top 6–10, Top 11–20, Top 21–50, and Everyone Else. This shows what share of total liquidation profit each tier captures.

Concentration over time

The Top 5 profit share is computed monthly. For each month, we rank all active liquidators by gross profit and compute what percentage the top 5 capture. Months with fewer than 5 active liquidators are excluded. Months where total profit is zero or negative show 100% concentration (trivially true).

Gas strategy comparison

For the top 20 liquidators by gross profit (with at least 5 gas-traced events), we compute average gas price (gwei), average gas cost (USD), total gas spent, and profit margin after gas. This reveals whether top bots outperform by paying higher gas for faster inclusion or by optimising gas efficiency.

Collateral-debt pair treemap

The top 30 collateral/debt pairs (by total volume, minimum 2 events) are displayed as a treemap where each cell's area is proportional to total liquidation volume for that pair.