Featured Commentary - High Frequency Trading - Ben Van Vliet, August 2012
"Earnhardt Moment:" IIT's Ben Van Vliet Discusses AT 9000, A Proposed Quality Management System for Automated Trading
|Ben Van Vliet|
|Occupation||Assistant Professor and Associate Director|
|Employer||Illinois Institute of Technology|
Ben Van Vliet is an assistant professor of finance at Illinois Institute of Technology Stuart School of Business, specializing in high frequency trading (HFT) systems. He is an industry consultant and frequent guest lecturer at industry events, including the MarketsWiki Questions: Exploring Financial Technology series.
Van Vliet has co-authored several books on trading systems, including 2008’s “Quality Money Management” in which he and co-author Andrew Kumiega laid out a set of best practices for systematic trading and investment. He sat down with John Lothian News Editor-at-Large Doug Ashburn to introduce his latest work, a framework for the development of a quality management system for HFT based on ISO 9000 engineering standards, which they are calling “AT 9000.” Though the paper was scheduled to be released in the fall of 2012, with the August 1, 2012 algorithmic debacle at Knight Capital, Van Vliet and co-authors Andrew Kumiega, Rick Cooper and Jim Northey have opted for an earlier release. To view the paper, click HERE.
Q: In the MarketsWiki Questions Series, you referred to the May 6, 2010 “Flash Crash” as the industry’s “Earnhardt Moment,” where the industry first accepted the need for change, just as NASCAR did after the death of Dale Earnhardt in 2001. Is this still the case?
A: In a way, I was stretching a bit with the analogy then, suggesting that tragedy leads to soul-searching but, with the flash-crash, it did not quite do that. But certainly the Knight collapse of last week is, without a doubt, a “Dale Earnhardt Moment.” Interestingly, it did not have a systemic effect, in that the company only hurt itself (by giving away nearly half a billion dollars to other participants in the market). With Knight being a larger and more well-established firm, and with the collapse being so significant, I think it has led to more soul-searching this time around.
Even though the death of Earnhardt was not “systemic” in that he was the sole victim in the crash, it was the perception of safety that was tremendously damaging. To the fans, to the other racers on the track, this was a very bad thing. It led to a lot of soul-searching in terms of how we run our races, what are our ethical obligations to each other and to the fans. In this sense I think it is similar to Knight. And nobody wants to see things like this happen, as it will only lead to more scrutiny and potentially unwanted regulation.
Q: Right. An industry lobbyist recently suggested to me that HFT firms are thanking their lucky stars that Dodd-Frank has been keeping the legislators and regulators busy, otherwise HFT would have been “squarely in the crosshairs.” What do you think?
A: They are in the crosshairs now, for sure. The industry is now waking up to the fact that this is not going to be allowed to continue. They will not get away with saying “Well, this is the free market; let us do whatever we want.” It is what I have been saying for quite some time - we have an ethical obligation to external market participants, and to society, to create confidence in the sustainability and fairness of the market.
Q: I was a market maker for 20 years, and I always looked at my ethical responsibilities as this: like all market participants, I enter each trade with a profit motive, and so long as I stay within the rules laid out by regulators and the exchange, I am an ethical trader. Are you saying there is more to it?
A: Absolutely. And it all changed because of the age of automation. As a floor trader, if you wanted to trade yourself into oblivion - by averaging down, or whatever - nobody cared because you did not have a systemic effect. Furthermore, if you made a mistake, you were able to amend your ways and make it right in real time. Or, in the worst case scenario, the risk manager would come down and physically remove you from the pit. Here, we do not have that possibility. Things happen in microseconds, before a human can intervene. In this system, where an individual can actually “poison the waters” beyond the market’s ability to soak up the liquidity, there is an ethical responsibility to ensure that your system is in control. Plus, because automated systems are part of the global network, people are trading in other markets based upon your trades. So, now, if one market is poisoned, the effects can ripple throughout the world in real time.
Q: So that is why you have come up with the concept of applying engineering standards to HFT systems?
A: Quality standards really came out of World War II, under the Department of Defense guidelines, and subsequently became international standards, and now have spread to other disciplines beyond mechanical engineering, such as health care and chemical industries. We are applying quality management systems, from design and development, to operation and control, of high frequency systems. The argument we are making is that it is your ethical responsibility as an HFT operator that your system run in control, or to specifications, which presupposes that you have some knowledge of the behavior of your system that, when you turn it on, it will work properly.
This is done by testing your software, proving your exchange connection, vetting your strategy, trading your system in a simulated environment, and trading with what we call a “probationary trading,” period, in which orders scale up from smaller lot sizes. This will give you an expectation of what trading should be. And there are many kinds of performance measures, such as profit per trade, trades per second, messages per second, or technological latency. All of these things are distributions, whereas you will not get the same value each time, but rather an average, or an expected distribution of values. We then prove that the distribution is stable, and we have statistical methods that prove the stability of systems. So your ethical obligation is that your output distribution is stable, and you are justified in believing that, because you have gone through the tests.
Once you turn the system on, then, your performance in real live trading can be assessed against that expected distribution in real time. Let’s say, for example, you are running an index arbitrage system. Some seconds, you are sending out 600 order messages, some seconds it is 300; sometimes it is 700; sometimes it is 450. It is a distribution. But if you suddenly get 17,000 messages in one second, your system has to recognize that that does not fit within the expected distribution, and will immediately shut down. So, you are monitoring your system in real time relative to your expected distribution, and the system responds in real time. The response could be a shut-down, or a widening of bid-ask spreads, or some risk “fail-safe” mode that it can slip into.
When you see a company like Knight say that it let the system run for 45 minutes - the second that it sensed that it was running outside its expected distribution, it should have shut itself off.
Q: So, each firm has what it would call its own “best practices” that it learned through trial and error, or through experience, but there is no industry standard?
A: Exactly. To be sure, Knight has a software quality department, and I am sure there are very ethical people there, but there are no industry standards for what qualifies as fulfilling your ethical obligation. In other words, there is no industry-wide set of tests that a firm must go through. Each firm is left to make up its own mind, and there are ethical conflicts that may come into play. such as someone being under a time constraint.
Q: Why is the economic argument not sufficient, meaning that the disincentives are there to keep firms from cutting corners? So, when a firm like Knight cuts a corner or makes a mistake, it is punished to the tune of $440 million.
A: You can certainly make the argument that there will be a natural evolution, that high quality firms will survive and bad ones will lose a ton of money and go out of business. But the ones that survive are the ones who adopt what become industry standards.
The same can be said of the aerospace industry, where there used to be a lot more manufacturers. Some could not build planes safely, cheaply enough. Eventually, firms that could not fly planes safely could not fly planes. But this evolution took between 50 and 100 years to get to where we are today, which is that airplanes don’t crash, or that they experience mechanical problems with such infrequent probability that virtually nobody worries about it. Because of industry standards, mechanical problems show up in diagnostics long before they lead to engine failure.
What I am saying is that now is the time to develop those standards for HFT. We do not have 100 years to get this right.
Q: How long would it take to develop and implement these standards?
A: There are a few thorny issues. First, we need to establish who would be required to adopt such standards. Certainly those with direct market access, but what about firms who hide behind the technology of their clearing member? Certainly the clearing member would need to be ISO 9000 certified, but what about the firm? What are the tests that we will need to abide by? And, will the regulators give in to HFT 9000 certification as a proxy for oversight? Neither we nor the regulators want them to come in and monitor our backtesting. So what we need to do is grab the ethical high ground and say that we are all going to abide by the same standard. Then we ask the regulators if this would satisfy our ethical obligation.
The precedent is there. Other industries abide by ISO 9000 standards. So, we can tell the regulators, “We expect you to treat us the same way you treat aerospace, which is that ISO is an implicitly approved method of self-regulation.” The Department of Defense used to require aerospace firms to adhere to its standards. Then ISO 9000 came along and the DoD said, “OK; you have your own standards now. That’s good enough for us. As long as you are ISO 9000 certified, you can be a defense contractor.” We can and should do that for HFT.