The methodology makes no sense. Our quoter currently quotes approximately the same dollar amount at all the time. If a taker takes all of our size at a level, a

The methodology makes no sense. Our quoter currently quotes approximately the same dollar amount at all the time. If a taker takes all of our size at a level, as they do tens of thousands of times per day, the resulting trade sizes will not follow Benford's law. Another dynamic is that many trades will be a bot trading against a stale quote. For whatever reason, some bots are configured to do this gradually in fixed-size chunks, which will also produce trade data that does not follow Benford's law. The authors create a benchmark rate of "automated trading" (i.e. deviation from Benford's law) at regulated exchanges and attempt to calculate the rate of wash trading at other exchanges by comparing the deviation from Benford's law to the deviation at regulated exchanges. Even if all of these exchanges would onboard every automated trading firm from anywhere in the world with directors and UBOs with any combination of citizenships, this methodology *still* would not work at all, because automated trading systems care about prices after fees, and Kraken's lowest fees are 0bps maker/10bps taker, while Gate's are -1.5bps maker/2bps taker. Another way of saying this is that Kraken has a "net fee rate" of 10bps and a "minimum spread" of 20bps (wow!) plus the price increment, while Gate has a "net fee rate" of 0.5bps and a "minimum spread" of 4bps plus the price increment. Expecting similar distributions of trades in these two conditions is extremely unreasonable, and one wonders if the fine authors consulted anyone in industry before publishing.