Given the Knight Capital Group debacle of August 1 and the subsequent discussions in Washington and within the industry regarding safeguards, Traders decided to probe the head of one market-making firm to learn the steps his firm takes to prevent algorithms from going wild. Matt Andresen is co-chief executive of Headlands Technologies, a registered market maker, headquartered in Chicago and San Francisco. Andresen previously ran Citadel’s wholesaling and market making business, a competitor to Knight.
TM: Who at Headlands is responsible for risk monitoring?
Andresen: You need overlapping monitoring. It’s a combination of traders, operations, compliance, and management.
TM: Do you have an overarching risk management philosophy at Headlands?
Andresen: The most important thing you have to have is “disciplined paranoia.” You must have respect for the problem.
That something could go wrong. That you don’t know everything. You must question your assumptions. You have to triple check. Test that assumption. It’s part of the culture.
TM: Do you monitor the exchanges themselves as well as your own orders?
Andresen: You have to monitor exchange health. Am I connected to the exchange? Is my connection robust? You have to monitor the health of the market data coming from that exchange. Am I seeing normal amounts of information? Are the prices updating within normal tolerances?
TM: When it comes to your own orders, what do you focus on?
Andresen: Are my orders sent within my risk checks? Am I buying more than I can afford to buy? Am I selling more than I can afford to sell? Am I sending too many orders? Am I sending too large an order? You have to check all of this. And, even if I’m within those risk checks, am I doing it too often versus the historical?
TM: Is all this monitoring automated?
Andresen: Some of this must be automated. So when some of that stuff happens, the system just shuts itself off. Some of the things must be exception monitored. That’s where the system throws off an error message with a sound like “HEY! PAY ATTENTION!” Also, you must have pro-active monitoring where your traders, your ops people, and your compliance people have access to the right information and can plow through that.
TM: What is exception reporting?
Andresen: It’s an automated report that is reviewed by human beings. That compares to an automated check that would say ‘hey I just got a disconnect from the exchange, stop trading.’ A
machine would do that automatically.
TM: How much of the monitoring is automated and how much relies on humans checking?
Andresen: It’s a combination of automated checks, exception reporting, and also human monitoring. I’d say that’s true for almost any firm.
TM: You used an airplane analogy.
Andresen: Right. So if my left engine is on fire, the plane knows to automatically shut that off. Compare that to the system saying ‘warning, warning, fire.’ Giving an exception warning to the pilot who then must take action. And, the third area, is the pilot looking out the window at the wings on a regular basis. Those are the three different ways.
TM: I see.
Andresen: With any complex system, you must have those things overlapping at all times. So, the computers are monitoring. The computers are warning the pilots. The pilots are not trusting the computer and doing their own checks.
