Latency Arbitrage: The Real Power Behind Predatory High Frequency Trading
themistrading.com
themistrading.com
... and as we all know its far easier to attack geeks than other more politically astute sections of the industry
In a case such as the 1987 crash, it's possible the such a delay in trading would've lowered stock values further as the volume of buyers versus sellers would've been less clear and selling pressure could've pushed down securities further before buyers could evaluate proper values.
I'm genuinely asking, not stating. The problem of being able to front-run block trades is a really old one.
(Wow, googling "front-running" is all HFT now.)
The article, while well-researched and well-written, seems to neglect the fact that HFT has severely narrowed the spread (now almost always one penny) whereas before it was large, and the difference went to someone else's pocket.
The other thing that seems to be missing in the thinking of the article is that there is nothing preventing the institutional traders from colocating their own box in Mawah, grabbing a piece of that cable, and joining the game.
And everything is HFT now, it seems--just look at dice for c++ jobs in chicago for one example.
I agree with the rest of your comment, at least in principle; this whole thing seems less about defending "fairness" than about defending "the way things have always been done".
The first exploits the difficulty the market has in determining the "NBBO" - the National Best Bid/Best Offer. This is the best price for buying and selling the stock on all markets. This price is important because the market must route an order to the market with the NBBO if it does not have the best price (due to RegNMS). To find this price, the market's themselves must process the quotes from all the markets. This same process can be done by private companies, and sometimes they can calculate the NBBO faster, and leverage the fact that the market has not yet updated it's value.
The second exploit they describe is much simpler. It leverages the fact that if an algorithm can determine that an institution wants to buy a large quantity of a given stock, the price movement of that stock becomes more predictable. Therefore, VWAP and associated algorithms are falling out of favor by institutions since they are most easily predated. Now, many institutions buy access to custom boutique algorithms tailored to hide the large purchases and avoid predation.
The second graf is a problem I've spent some time with for clients, and figuring out that someone is shopping a block and jumping in a threat that everyone I've talked to has called "frontrunning", but I don't know what the real name is.
I'm working at a lower level than figuring out and trading towards optimal pricing targets, but that aside, I think there's a real future to research into the programming flaws of automated/algorithmic trading systems.