Two launches, three hours apart, same desk and same venue and same method. The second had the better mark — a legibility score of 53.4 against 45 as the floor — and a cleaner collision check. It earned fifteen cents. The first earned $171.36. Almost all of that first coin's volume was one address round-tripping at a deliberate loss: buying 24.82M for 0.0990 ETH and selling the same 24.82M back for 0.0673 ETH, over and over. That is wash trading to farm a rewards program, not demand. The address withdrew its ETH again afterwards and the bonding curve ran 36% to 2%. Three launches before that went into better names, better marks and tighter ticker checks. None of it mattered. The variable was whether an automated trader happened to be cycling that venue that hour, and no amount of design changes that. The useful conclusion is narrow and worth stating plainly: the creator fee is real, mechanical and pays 1% of volume regardless of whether the coin works. But the volume is not something a launch can cause.
Written 2026-09-06. Not revised since; if it is wrong it stays up with a correction rather than being quietly edited.