The ERC-721 leaderboard had a mirror-token problem

The top ERC-721 transfer source last week was not a mint everyone suddenly cared about. It was a mirror contract.

From June 17 through June 23 UTC, ORE404Mirror emitted 273,448 unique ERC-721 Transfer logs from 740 transactions. That made it the busiest ERC-721 event source in Xatu's new execution tables, more than 7x the next contract by deduped event count. The awkward part is that the verified source says exactly what it is: contract ORE404Mirror is DN404Mirror.

This is the query I used for the leaderboard. The important bit is the dedupe key. A plain count() on the raw table was too high, so I counted unique (block_number, transaction_hash, log_index, erc721, token) tuples and cross-checked the result against raw logs.

SELECT
  erc721,
  uniqExact(tuple(
    block_number,
    transaction_hash,
    log_index,
    erc721,
    token
  )) AS transfer_events,
  count() AS raw_rows,
  uniqExact(transaction_hash) AS txs,
  round(transfer_events / txs, 1) AS events_per_tx,
  uniqExact(to_address) AS recipients,
  uniqExact(from_address) AS senders,
  uniqExactIf(
    tuple(block_number, transaction_hash, log_index, erc721, token),
    from_address = '0x0000000000000000000000000000000000000000'
  ) AS mint_events,
  uniqExact(token) AS tokens
FROM canonical_execution_erc721_transfers
WHERE meta_network_name = 'mainnet'
  AND block_number BETWEEN 25333537 AND 25383756
GROUP BY erc721
ORDER BY transfer_events DESC
LIMIT 25

That returned 1,778,431 unique ERC-721-shaped transfer events across 603,491 transactions and 5,078 contracts for the seven complete UTC days. ORE404Mirror was first with 273,448 events. The second-place contract, a Uniswap Diamonds / sEReC20721_diamond_test contract, had 38,254 events from only 44 transactions. After that the leaderboard looked more like normal mint traffic: SeaDrop clones and small collections in the 18k-23k event range.

The ORE shape was the strange one. Those 273,448 events touched only 3 recipient addresses and 4 sender addresses in the deduped table. The median ORE transaction emitted 360 unique ERC-721 Transfer logs; the p95 was 384. It was not thousands of collectors minting one item each. It was a small number of transactions making a mirror contract churn out hundreds of ERC-721-shaped logs at a time.

The raw log check landed on the same number:

SELECT
  count() AS log_rows,
  uniqExact(tuple(
    block_number,
    transaction_hash,
    log_index,
    address,
    topic0,
    topic1,
    topic2,
    topic3
  )) AS log_events,
  countIf(
    topic0 = '0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef'
  ) AS transfer_topic_rows,
  uniqExactIf(
    tuple(block_number, transaction_hash, log_index, address, topic0, topic1, topic2, topic3),
    topic0 = '0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef'
  ) AS transfer_topic_events,
  uniqExactIf(
    transaction_hash,
    topic0 = '0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef'
  ) AS transfer_topic_txs
FROM canonical_execution_logs
WHERE meta_network_name = 'mainnet'
  AND block_number BETWEEN 25333537 AND 25383756
  AND address = '0x84161ecf72829d5b55a9f9c344fb0d34422d4b7a'

For ORE404Mirror, that produced 273,448 unique Transfer-topic log events across 740 transactions, matching the ERC-721 transfer table exactly after dedupe. The raw row counts were higher: 428,104 rows in canonical_execution_erc721_transfers and 446,328 raw Transfer-topic log rows before dedupe. That is why this post uses unique log identity, not raw row count.

The mechanism matters more than the label. DN404-style tokens intentionally split the world into a base token and an ERC-721 mirror. The mirror emits ERC-721 events, so an event parser is not wrong to pick them up. But if you read an ERC-721 transfer leaderboard as "NFT market activity," this kind of contract will make you lie to yourself. On June 22, ORE404Mirror was 29.6% of all unique ERC-721 Transfer logs in the table. On June 23 it was 0%. That is bursty mirror mechanics, not a durable shift in NFT demand.

The second outlier makes the same point from a different angle. The Uniswap Diamonds contract emitted 38,254 unique Transfer logs from 44 transactions, about 869 events per transaction, and involved only 62 token IDs. So even below ORE, the top of the ERC-721 leaderboard was not a clean list of popular collections. It was partly a list of contracts that batch or mirror state loudly.

The safe query pattern is boring but necessary: dedupe raw rows, count transactions separately, and look at recipient/sender spread before giving the leaderboard a market story. Transfer(address,address,uint256) tells you an event shape. It does not tell you whether humans were trading NFTs, a mint contract was distributing supply, or a dual-nature token mirror was screaming into the logs.