Conversation

This Tweet was deleted by the Tweet author. Learn more
Replying to and
One thing you can do is the BBO-trades test: see what % of an exchange's fills happen on the BBO at the time of the trade, and throw out any exchange with < 40% of volume trading on the BBO. This catches ~90% of fake exchanges.
2
5
(1/3) We think this form of testing is heavily dependent on the threshold of time defined for "BBO". More sophisticated wash trading techniques focus on RNGing this element (shouldn't be too difficult to do), waiting for 1 sec or so, before trading themselves out.
1
1
Replying to and
Totally hear you--but as it turns out right now most exchanges are lazy enough at faking volume that just doing something approximate is close enough. You definitely do get false positives, but if you set the bar at 40% then any exchange that's legit will pass.
1
3
Fully agree. We have decided to take an alternate path of using the data that is on hand (web traffic, liquidity, time and sales) to predict volumes via a trained model and track reported vs predicted, and flag out errant exchanges instead. Will be launching this soon! 😀
2
5
Show replies