How To Fix High-Frequency Trading

Tyler Durden's picture

Submitted by Anthony Tassone of Green Key Technologies,

Solution to HFT Debate: Match Based On Price & Quality, Not Time.

The recent public outcry over high frequency trading is pointless. Solutions exist. Virtually every comparable market in the world uses them already.

But, some electronic exchanges may not willingly adopt them. Doing so may disrupt their current business model. The incentives are misaligned, and competitors or regulators may need to force the issue to see change. Luckily, the issue to be forced is far simpler than most think.

It’s time to add quality to the matching process. Over thousands of years, every naturally evolved market has headed this direction – from the ancient Greeks to Alibaba.com. It’s time for Wall Street to realize what they lost along the way, and how it can fix far more than just HFT.

In 500 BC, the Greeks were auctioning off war plunder, grain, livestock, even ‘wives’. Sellers set the price, tossing out descending ask prices until a bidder emerged or the price got too low.

2500 years later, little’s changed. The world’s financial exchanges auction livestock and grain, just the same. They just added stocks and bonds instead of wives. But, the biggest real change came when the order books went electronic. The programmers who created these systems copied everything they understood from the auctions of yore, in Greece, on the streets of lower Manhattan, and in the massive exchanges that emerged from them. But, the auctioneer cannot scale to serve millions of customer around the world, so they replaced him with a “matching algorithm”, the piece of software that selects buyers and sellers.

It is that bit of logic at the heart of a national debate on the merits of high frequency trading. It is that bit of logic that lost something critically important. HFT traders are not sidestepping rules to gain advantage. Finding alpha in information advantages is exactly what they’re paid to do. And most, even among HFT firms, do so within the rules set by the exchanges. Rules set by the design of the matching engine. I know because I was once one of these traders.

The big problem in the electronic exchanges is not agents searching for inefficiencies. The problem is structural. It’s that the matching engine rewards speed and encourages the proliferation of order types that act as trap doors. That’s because when those programmers copied the auctions they saw, they lost the implicit decisions – the ones made by auctioneers deciding who to do business with or floor traders deciding which yelled bid to acknowledge.

Those traders weren’t just concerned with speed; they were making a judgment on quality too. Low latency trading is complex. Network, hardware and software optimizations -- all designed to reduce the time to send messages. Time is absolutely money. But only because the exchanges timestamp orders and rank them in the order book according to price AND time. They incentivize being at the front of the line.

Understanding where your orders are queued in relation to all the others at the same price is an advantage. It’s like being first in line to get a box that might have gold or might have a grenade. Better if there’s ten other guys behind you willing to take the box after you peek inside. The guy at the end of the line is at a significant disadvantage, he has no one left to transfer the risk to. Low latency trading is all about being first so you have the option to “scratch”, or exit for little or no loss in an adverse movement, and to profit from favorable price action.

Exchanges operating time based matching engines have inspired the low latency arms race. And they’ve been profiting from it for almost 20 years. Over time it’s grown into a tangled web of liquidity pools, data center ‘cross connects’, and low latency microwave towers. Transaction fees collected from low latency trading firms are substantial and have the exchanges bending over backwards to support the industry.

What began as a simple and logical idea – the only conceivable ‘fair’ way to do things when computers had 4K of memory and couldn’t do all that much actual computation -- ranking participants as first-in-first- out, has devolved into a technological war that dilutes the concept of providing liquidity. Instead it creates a low concentration liquidity vacuum where volume results not from the exchange of risk but from mere tinkering with the rules to see what loopholes can be found.

What if one more measure could be added that would specifically discourage that kind of tinkering? It’s been done before, many times over.

eBay is the most successful auction in the history of mankind, with over 300 million users. But, does time play much role? Not really. In fact, time extends as late players come to the bidding. Instead, eBay operates on price and quality.

Quality is determined by user feedback. Participants rate each other’s behavior, like the ability to honor terms and deliver the goods promised, failure to pay for an item won, etc. It was a simple upgrade, meant to assuage the fear of doing business with strangers a world away. But, it put all participants on equal footing, punishing those who would game the system by excluding them the next time.

The world’s electronic exchanges need to take a page from eBay’s playbook and start measuring the quality of their participant’s liquidity and factoring that into the matching process, rather than how few microseconds it took them to submit it. And, it can be completely automated, using data like the a user’s historical accept/decline ratio, or quotes/fill ratio, to determine where in the order book queue a participant is placed. Orders can then be ranked and sorted by price and quality, i.e… the likelihood that a participant will actually stand firm on the price submitted thereby creating an order book with very ‘sticky’ liquidity gravitating toward the top of the book.

Floor brokers and floor traders operated just like this, with a price+quality matching algorithm, albeit in their heads. The best price was always honored. But, given a choice of bidders, if the broker didn’t believe you would honor your quote someone else would get the trade and you’d be sent to the back of the proverbial queue. Bust a trade enough times, and even your highest prices wouldn’t reliably win.

Voice brokered OTC markets implicitly work like this today too. Traders who frequently ‘flake’, or back away from quotes, don’t get a phone call the next time a deal is available to be done. This encourages integrity among participants, something that was not transferred to the electronic order books.

Quality-based matching is the next logical step for the global electronic marketplaces looking to restore the depth and concentration of their liquidity in the wake of the national HFT debate. In doing so the regulators and exchanges will reward those for providing a real service, punish those who attempt to
manipulate the order book, and help restore market confidence.