The best way to spoof is not to spoof at all

The Chicago commodity trader Michael Coscia was convicted of ‘spoofing’ 2 days ago. Spoofing is where a trader submits an order with the intention of canceling it before transacting. The idea is to send a false signal of market demand or supply to trick other traders into believing an imbalance is present. Are traders that dumb?

The algorithmic revolution in market making sped things up. Humans are smart but slow. Computers have no brains but are fast when told what to do. Put them together and you get a smart fast solution to market making. The algos use human developed software to spot patterns that indicate the likelihood of the next tick being up or down – an important piece of information for profitable market making. Surely, then, a human software developer would like to ‘spoof proof’ his algo?

Algorithms work like this. A human specifies a model which relates the next price tick to a bunch of variables. The computer calculates whether or not these variables correlate with the next price tick. This ‘algorithmic training’ exercise should be able to spot and eliminate the false spoofing orders if the human who writes the algo is any good. In fact, the algo should recognise the false market and infer the opposite intention – a false sell order is in fact a future buy order that is hidden! Now THAT is valuable information…

Michael Coscia didn’t make much money spoofing largely because the algo traders ignored his false orders, and the smartest algos jumped ahead of his real intentions by buying when his false sell orders were posted. Coscia’s spoofing actually communicated his future trades to the market! How dumb is that? He would have been better off just hitting the offer – no algorithm can see that coming!