Although we are more than a year and a half from the 2011 Knight Capital Meltdown, we are still dealing with the aftermath. Not only are former Knight Capital employees moving to other firms, regulators are working on ways to avoid another software glitch that brought down the Jersey City broker. On August 1, 2012, Knight Capital experienced a technical mishap that caused errant orders to be sent out into the market and at the end of the day, the firm owed more than $461 million to counterparties. The next day, that figure jumped to north of $600 million and it eventually forced the company to split up and sell itself to competitors. The SEC also levied a $12 million fine to the owners of Knight for its IT glitch.
Since then the industry has talked about the need for a kill switch that could stop errant trades and potential Flash Crashes before they grow to Knight Capital proportions. This idea has its share of controversies and critics with questions: Who pulls the kill switch? What impact could it happen in the market? Would it even work?
[Traders Exclusive: Janney Montogomery builds up trading, research desks in down time.]
Traders Magazine spoke with David Lauer, a former high-frequency trader from Citadel who now serves as architecture technology consultant for Verdande Technology. This Norwegian firm offers risk tools in the oil and gas space that it hopes to apply to financial services firms. By using the same predictive data that foresees blowouts of oil pipes, they think they can help trading firms predict and react to the next flash crash.
Traders Magazine: Tell us about your firm.
David Lauer: Verdande Technology works in oil and gas where we built a predictive monitoring system using technology called case-based reasoning. It’s a machine-learning technique that is based on the notion that the past is a reasonable predictor of the future and you need expertise, domain knowledge and context to have the most useful predictive abilities. When you get to a state you have seen before, you have context around that state.
In oil and gas if there is about to be a blowout in the pipe you know this, for example, because you have seen that the torque is at a certain level and the pressure is dropping, etc.
Traders Magazine:What does this have to do with trading?
Lauer: It’s a paradigm shift from responding to failures to preventing them. We are moving into financial services with these complex risk systems that are static-based. They are prone to failure because these are complex systems that are too static and rigid. If they can learn from the past they can become adaptive. We can prevent technology failures instead of just responding to them.
Traders Magazine:Could this have helped at Knight Capital?
Lauer: At Knight Capital, there was a single server with older software that was not tracking its own positions. It was just flooding or spamming the market with huge numbers of orders. It was not new code that wasn’t tested; it was old code that was mistakenly not cleaned up.
That said, it also points to the fact that Knight’s procedure was suspect, which they have readily admitted themselves. Their monitoring systems were not able to detect this in a quick period of time, pinpoint the problem and how to respond to it.
Traders Magazine:Why are kill switches such a controversial idea?
Lauer: I think that static kill switches are a terrible idea. If everyone were to implement them they would all go off at the same time and in the worst possible conditions. As the market becomes stressed or experiences Flash Crash-type of conditions, dynamic kill switch are able to recognize anomalies in behavior that is under or over certain thresholds. They can determine if things are normal because you’re in a stressful market situation, for example.
Traders Magazine: IfKnight Capital had a manual kill switch it could have stopped all of those spam orders?
Lauer: Yes, a manual kill switch would have been able to mitigate some of the problems if they had been able to recognize what it is they wanted to kill. The big problem was they had no context around what the problem was. And it took them a long time to identify it.
In IT and trading, problems are easy to recognize but why they are happening, where they are happening and how to fix them is the major problem when you are trying to quickly react to them. That said, a manual kill switch still would have been a huge cost to them — just like Goldman Sachs’ problem when they had their options market making spin out of control in a similar way [last August.] Knight’s scenario lasted 45 minutes and Goldman was much quicker but it still cost them a substantial amount of money although nobody knows exactly how much.
These manual processes at this speed of trading might prevent another firm getting blown out but it won’t help you from losing another $100 million, let’s say.
Traders Magazine:What does this have to do with traders?
Lauer: I would say that an automated trader who accidentally pushed out bad software and ruins their career would be interested in a dynamic kill switch that would be able to recognize that and prevent it from happening.
Knight’s problem came down to a single server. When a single server can wreak havoc on the market and cause 5 or 10 percent price dislocation, take up 20 percent of the volume of some instrument – that is not good.
Traders Magazine: Speaking of automated kill switches, will traders ever trust a piece of software overseeing another piece of software?
Lauer: Nobody is going to push out a solution that is fully automated right now. That is just not on the table but everybody recognized that as systems become more complex, you need software to monitor software. Initial rollouts would have humans in the loop and I wouldn’t recognize otherwise. The trend is inevitable towards automating this. Once the solution is out there, we will see fully automated kill switches.
Traders Magazine:Is this solution available now? Who is using it?
Lauer: We can’t release names but we do have active partners. We have active proofs of concept in a few production deployments right now. We should be ready by Q1 of 2014.
Traders Magazine:Where did you trade?
Lauer: I was a high-frequency trader of equities and I worked at Citadel and then at Allston Trading. I left in 2011 and subsequently spent time working on these out of control practices. I did sit on the SEC roundtable that looked at Knight Capital and that is where kill switches were revealed as a [possible] solution to the industry’s woes.
