Three wallets round-tripped two thirds of the volume in CoinMarketCap's top Uniswap v2 pair

The two fields that place a swap in the chain come back from the endpoint as quoted strings, while the amounts on the same row come back as numbers ^[3]. Sort them as text and "99" lands after "1000", so every "between" the detector computes is wrong and nothing raises an error ^[3]. The first function in the engine does one thing: cast block and log index to integers, and return None when a row cannot be placed ^[4]. Rows that cannot be placed are dropped and counted, and every receipt states the count ^[4]. The regression test that pins the trap is named test_block_and_log_index_are_sorted_as_integers_not_as_the_strings_they_arrive_as ^[5].

The ranking and the volume share measure different fields. MOTO/WETH was first among Ethereum's Uniswap v2 pairs by 24-hour transactions at capture time ^[13]. The share is of volume: $127,254 of $192,275 ^[15]. Each round-trip prints two legs, so the 29 round-trips are 58 of the window's 800 prints, which is 7.25 per cent of the count the ranking sorts on ^[20]. Everything else trading in that pool over the 3.6 hours came to $65,021 ^[21].

Whether the volume share transfers to another pool depends on what the join accepts. Two prints count as one middleman when the maker address is the same, the block is the same, the sides are opposite and the sizes match within 5 per cent ^[7]. Five per cent is generous; the first receipt in the evidence matched at 1.58 per cent, in block 26006339, log index 209 out and 218 back, inside one transaction ^[16]. The feed does not include a pool address, so prints are grouped by venue plus the two token contracts ^[6]. The join wants one address on both legs, so a wallet that splits its two legs across two addresses falls outside it, and loosening the size tolerance leaves that where it is.

The sandwich hunt itself came up close to empty. A first run over 1,200 prints from the busiest pair found zero same-block sandwiches ^[10]. The ten-token census across Ethereum, BSC and Solana found two in 8,000 prints, the same wallet front-running a seller on UNI and on LINK, for $0.95 and $2.19 ^[11]. That is a rate of 0.025 per cent ^[12], and $3.14 of extracted value in total ^[22].

The detector gets there without a mempool subscription. "You get no mempool and no MEV labels, and you don't need them: a middleman has to print," the author wrote ^[17]. The round-trip count covers one pair over 3.6 hours ^[14]; the wider census was run to count sandwiches ^[11]. "So I changed the headline before building the page, not in a caveat after," he wrote ^[18]. The code is MIT-licensed, stdlib-only Python with no key required, and the live judge page puts every request behind it on /evidence ^[19].