In January 2022 a new NFT marketplace called LooksRare launched with an aggressive growth mechanic: traders earned LOOKS tokens in proportion to their trading volume. Within days it reported volumes that dwarfed OpenSea, then the dominant venue. The catch, visible to anyone reading the chain, was that much of this volume was a handful of wallets selling NFTs back and forth to themselves at implausible prices, farming the emission schedule.
For manipulation research this was a gift: a marketplace-scale natural experiment in what happens when you attach a subsidy to the most fakeable statistic in finance. It is the identification core of my paper The Economics of Wash Trading.
The mechanism, precisely
- Trade-to-earn.Each day's LOOKS emissions were split across traders pro rata by their share of eligible volume. Trade more, earn more.
- The arbitrage. Sell an NFT between two of your own wallets and you pay the marketplace fee and gas, and receive reward tokens. Whenever expected token value exceeded those costs, self-trading printed money.
- The equilibrium. Rational actors scaled the trade until costs met rewards. Enormous nominal volumes, tiny genuine turnover underneath.
Note what is missing: any need for a deceived buyer. The reward contract itself was the counterparty paying for the fake volume. That distinguishes incentive-driven wash trading from the classic tape-painting story, and it is why the two need different fixes.
What the experiment identifies
Because token rewards existed on some venues and not others, over windows with otherwise similar market conditions, the setting separates the two motives for wash trading:
- If wash trading were mainly advertising for collections, it should appear broadly across venues and predict genuine future activity. It does not: across the data, wash volumes show no significant relationship with real volumes in future days.
- If wash trading were mainly reward farming, it should concentrate where volume is paid for, track the value of emissions, and collapse when rewards fall. That is the observed pattern, and on incentive venues the rewards appear to influence real trading activity as well, by changing who participates and why.
Lessons for mechanism designers
- Never subsidise a self-reportable metric. Volume can be printed by one actor at will. Rewards keyed to it are grants to whoever prints fastest.
- Price the attack before launch. The profitability condition, reward value versus fees plus gas, is arithmetic anyone can do in advance. If it clears for a self-trade, the volume statistics are already spoken for.
- Filter by funding, not by wallet. Venues that later excluded self-financed flow from rewards blunted the farming. The techniques are on the detection page.
- Expect the statistics to stay poisoned. Historical NFT volume series still carry the LooksRare era inside them. Any backtest or market-size claim built on raw 2022 volumes inherits it, as discussed on the NFT markets page.
The broader tally of how much NFT activity was manufactured, and who profited, is on NFT wash trading: scale, motives, and data.