In the coming months, I’m going to be leaving MIT. I’m not sure how long my access to my computer account will last, so I’m starting the process of moving everything over. For those of you that access this blog via an RSS feed, nothing will change. However, if you have this blog bookmarked, please change the bookmark to point to www.jonbirge.net. For now, that will still put you here. In the future, who knows?
The Atlantic recently carried a fascinating article on the odd patterns in stock market data caused by mysterious high frequency trading algorithms which post and remove bids and asks on stocks with no intention of ever buying the stock in question. My best personal theory is that these mysterious bids and asks are simply ways of “pinging” the market electronic infrastructure for information about the speed and timing of the hardware running it. High frequency trading has become so fast that firms do their best to but their computers as close as possible to the exchange’s servers.
My guess is these bid patterns are just acting as clocks. To quote one of the experts in the article, the speed of light actually starts to matter. Given that the timing of trades can mean millions in profits for some of these firms, it stands to reason they would try to keep close tabs on the speed of their connection to the exchanges. By sending in a sequence of false bids (bids so far away from the current price and retracted so quickly they will never result in a buy) they can safely figure out how long it takes an order to go from their computer to the exchange’s computer. They use a sequence (ramp) of increases prices simply to keep track of all the different orders and their timing. (If they used the same price repeatedly, they’d have no way of knowing which order was which.)
My next best theory is that they are probing the HFT algorithms of other firms, trying to either reverse engineer other trading programs (which would give them a decided advantage) or search for some way to induce other programs to act in a way that can be exploited, like a big robot battle played with real money. They are trying to see if they can predict what others’ programs will do in a way that doesn’t require them risking any money. Frankly, I don’t see this as too likely, given that it would be very poor programming for anybody to let their HFT algorithm respond to ludicrous bids.
A final, rather crazy possibility, is that these false bids are a way to communicate between market participants over public, non-traceable channels. If encrypted information were embedded in the timing or changes of the bids, it would be a good way for firms to collude in a way that would be nearly impossible to trace: everybody is allowed to post bids, and every is allowed to read them. Unless you knew how to decode the information encoded, you’d have no way of proving information was being exchanged between parties. And what better avenue of communication between financial firms than the stock market’s public quote system?
One thing that is certain is that this frenetic activity, with hundreds of quotes a second being generated and cancelled, is clearly meant for the “benefit” of other computers, either at the same firm or another. It’s a fascinating phenomenon. I’m not sure what it portends for the markets, however. As people leave the market in droves (as evidenced by 13 straight months of mutual fund outflows) I wonder what will happen when the only people left aren’t people.