The Proving Grounds: Hyannis Port Research Enters Next Stage with Riskbot

If you're going to take a job writing about technology, it's imperative that you find techie stuff interesting, if not downright fascinating. I wouldn't have been here for six years and counting if I found the topic of fintech to be leaden.

I enjoy talking about new platforms and watching those products evolve. With that in mind, I took some interest in hearing about Hyannis Port Research's rollout of version 3.0 of its flagship product, Riskbot.

Riskbot was born out of SEC market access rule 15c3-5, with the tool being a literal box that sits between the trading firm and the exchange and inspects orders traveling between in order to prevent against erroneous trades going through to market.

It has big admirers. UBS has already publically given its seal of approval to Riskbot, which is split into two sections: Riskbot G4 Pre-Trade Risk Control and Market Gateway Appliance, which is the gateway sitting at the exchange, and Riskbot M5 Real-Time Risk Management and Control System, which collects data from all of those devices and provides services on top of that information.

While the solution was created in response to 15c3-5, Tony Amicangioli, Hyannis' CEO, told me that Riskbot has evolved to cover cyber security and surveillance needs for banks, exchanges and quantitative hedge funds.

Other types of platforms allow you to fake it a bit more easily; where latency isn't as vital and the cost of an error isn't as terminal. But this isn't exactly one of those; actually it's quite the opposite.

"The [industry] coverage was very important for us in the beginning. The first year, getting into new markets and connecting to new markets was taking us a very long time," Amicangioli says, looking back on when the company launched in 2011. It was in July 2012 that the company went live with its first customer. With 3.0, in addition to fail-over and performance improvements, Amicangioli says that the platform's ability to broaden its range of services was also a key focal point.

"We looked back at the hardware that we built and realized that we could look for denial-of-service (DoS) attacks and other things that have nothing to do with trading," he says. "Similarly, surveillance was almost free because the platform is collecting all of the trades at this one central point."

Promises & Proof

When Amicangioli was describing Riskbot to me he said that it "cuts the wire between the trader and the exchange and you just put us in the middle." They do this while only adding latencies of 453 nanoseconds client-to-market and 150 nanoseconds market-to-client, according to the company.

I said to Amicangioli that it's interesting just how sure and exact they are with their latencies; if those numbers don't play out once installed, there will be a wave of upset customers and word spreads quickly in this business.

He acknowledged that when they first launched, proving it to their claims was the hardest part. But as Hyannis has built up speed and brought on new customers, they've been able to move past those conversations more easily. Additionally, on the company's website, Hyannis goes into great detail backing up its claims of minimal latency.

It's also interesting that Riskbot was built to stop fat-finger trades that have led to massive losses both at individual firms and, in worse-case scenarios, sent out a shockwave that lead to a flash crash. If one fat-finger trade slips through and makes headlines at mainstream media outlets, it's probably "game over" for HPR if it's their client in the news.

"Convincing people that we would stop the Doomsday trade ─ that was a much more challenging process," Amicangioli says. "It required us to do what every startup does: Proving it out one customer at a time."

High Stakes

I find a product like Riskbot interesting because it is designed to stop arguably the scariest thing on Wall Street ─ a devastating erroneous trade being made ─ all while adding only a tiny fraction of latency.

Other types of platforms allow you to fake it a bit more easily; where latency isn't as vital and the cost of an error isn't as terminal. But this isn't exactly one of those; actually it's quite the opposite.

If Riskbot can continue to live up to its promises, in today's increasingly-electronic trading world, it will find a home. If it can't, it will implode.

Who says that this stuff isn't interesting?