By @Newton, August 25, 2016
It was May, 2011 and Zhang Fei was wearing the full gown and cap for his convocation ceremony at UBC. He was graduating with an MSc in Applied Math and his parents had come to Canada for the first time to celebrate his achievement. His father had met with one of his professors at UBC and helped arrange a job interview with a French bank. By June, he had relocated to Paris -- another country where he barely spoke the language.
In his first day working at the bank in La Défense, they showed him a bunch of software that implemented the algorithms he had studied and even a few that he helped create as a student. At lunch he met some of the people who made risk decisions based on the software. Very French people. Fei ordered the beef tartare to try and fit in better, but seemed to get the sense that they thought it was strange to see a young Chinese man eating raw beef -- even though the director of the risk department was eating the same thing. He didn’t want to outshine the master, but also valued the connection with the director above all else.
After lunch, he and a few people had a conversation over coffee and cigarettes. They were all very passionate about their work, keen to chat, but somehow disgusted by activities that kept them away from their work. One of them said "Zhang, what do you think of replacing the tails of a normal distribution with a power law for the VaR calculations?" Fei barely understood the question through the thick accent, just as the man barely understood Fei's answer.
Fei said "Yes, I think it is a good way to incorporate fat tails but, really, I don’t think we should be trying to do predictive analysis." Apparently he said something wrong as everyone stopped their side conversations and turned to look at him. One man even got up and walked away away.
"What?" asked another young man, who looked Vietnamese but sounded French. Fei thought it was strange to see an Asian person who didn’t sound Asian. He had never met anyone like that growing up in China and it gave him hope that he could fit in someday.
"Well, I think prediction is hard," he paused and waited for laughter at the reference, but none came. "We create a false confidence for decision makers when we aggregate and reduce information. Information is power in markets and we should try to do work that helps decision makers access and consider the subtle features of all the information we have access to." Someone else immediately said "Right, well that is stupid. Have you all seen my paper on calculating the median path from a stochastic process? It really is quite novel and…"
Before he knew it, it was his last day at the bank. Fei had worked there for just under one year. It was a quite poor experience, all around. He felt shame on behalf of his family and teachers who had believed in him.
His last day was very different from the first one. The meeting was held at the downtown office by Montparnasse Tower, rather than La Défense. From the windows in the boardroom you could see the famous landmarks of Paris. It was beautiful.
The reason he was being fired was that he had released software and research in his own name online as part of a coursera course on data visualization. His work was popular, even getting some citations for the papers on arXiv and several hundred downloads for the software on the Mathworks file sharing site. He asked the HR person if the bank saw any value in the work he had made public: "No," was the immediate and complete reply.
"Thank god I am wealthy!" Fei thought to himself as he left the bank offices for the last time. He took his few possessions and relocated to a small town in Switzerland. He found that he fit in better and people respected him more in the peaceful town of Zug. He was free to do as he wished. He even visited a brothel outside of town a few times, but enjoyed the wholesome lifestyle of the Swiss more than anything. He threw himself into learning data visualization and software development like never before. He continued to develop a strong reputation online as an original thinker.
The first visualization software he created was a library of videos of "market episodes". They showed all the transactions and standing orders that occurred in a period of time for a particular financial instrument. He shared some of the software behind the videos and even shared some gifs of a few interesting episodes, but generally kept things to himself to ensure that he could get by on a personal data feed from the exchanges rather than a corporate one.
Fei built a UI to watch the market episodes that kind of looked like youtube. It let him watch videos of market episodes and then edit them in various ways: slow down, speed up, even reverse or splice different segments. He spent months tweaking the software, collecting data, and watching the market up close and personal. He mainly watched instruments that traded on the CME and CBOE. He particularly liked the illiquid derivatives that traded on liquid underlyings. Like six month, out-of-the-money options on natural gas. Or, even better, OTM options on the VIX index.
Eventually he made it so that he could show several different contracts at the same time. For example, he would watch two charts of the front month futures contract at different time scales and another four charts of options with different strike prices. Six charts showing all of the transactions and orders that pertained to one futures contract and four options contracts. These graphics started to let him see how trades spread across different instruments in real time. They seemed to echo and reverberate and a strange, but intuitive way. Nothing would be the same for Fei again after this.
Fei was absolutely blown away on January 1, 2013, when the VIX index fell twenty percent in one day. By the end of the month, it had touched a record low of 11.05. He had never seen anything like it and it inspired him to start speculating with his own money for the first time in his life. He had lived cheaply in Switzerland since leaving Paris in mid-2012, but the lack of cash flow had put him in a tough position. His sister helped him by giving him some US dollars, helping him open an account with Interactive Brokers, and even holding his hand while he made his first trades.
He came up with a market-making strategy for VIX options that involved holding an inventory of both calls and puts, and posting orders on both sides of the markets at all times for all the contracts that he held. It was a simple strategy, but he was confident to implement it based on the market depth that he could see in these illiquid options. He learned how to position his orders in different ways that would help them get filled, or not, so that he could keep some balance between his options positions and cash. He updated most of his orders weekly, which was very different from the HFT-dominated market that seemed to jump at it's own shadow. Even with a fifty percent spread, the quotes on these options moved as if the traders behind the orders were afraid of actually getting filled.
Rather than obsess over partial fills, near misses, and lost opportunities, Fei focused on his analytic tools while implementing his trading strategy. He created all kinds of new functions for comparing and combining data from different episodes. Non-parametric stuff that helped give a sense for the context of different features of market movements. He believed that most of the statistics were just noise and he didn’t think much of them, until he started learning about machine learning and clustering algorithms. He started to see the potential there and quickly made something that would give a series of suggestions of what might happen next in a particular market episode, just like the auto-fill function in Google. Still, he didn’t want to try and become a predictor, so he gave this project away to some of the top researchers and developers he knew online. They worked anonymously and made all their work public, despite people who commented that they were crazy for doing so.
One of the best programs Fei made categorized different market episodes and showed how they were distributed over time. It was a hyper-descriptive calendar that showed when different types of market conditions had occurred in the past with all kinds of secondary information on what those market conditions actually looked like. He found that it was very helpful for understanding when he was likely to get filled on his orders to buy or sell VIX options. Soon after building it, he started to use it to help set and update orders more than anything else.
By the end of 2013, Fei had made hundreds of round trip trades across dozens of options and had increased the cash balance in his portfolio by four times. His total portfolio was up about six times, depending how you valued the illiquid options that he held at the time.
Fei kept his head down, trading and researching, for another two years until he heard that Eric Hunsader had won a whistle-blower award from the SEC for his work on HFT in March 2016. Apparently Eric had showed evidence of traders causing delays in the national order updating system. It didn’t surprise Fei that Eric had used cool, new data visualization techniques to show this manipulation. Fei knew that careful data visualization could be almost undeniable in the clarity of it's implications, unlike so much else in modern economics.
Fei went back through Hunsader's body of work at Nanex and wondered how he made missed all of it -- parts of it were so similar to what he himself had been doing with visualizations of the order book! Fei had a warm feeling of comfort and confidence when he realized that he wasn’t alone. He had never really fit in or enjoyed his time at the French bank. And no-one really enjoyed living or working in China. He knew he was lucky to go to Canada and Europe but, for the first time, he really started to feel lucky. It was March 2016 and spring was just around the corner. That summer became the most healthy and balanced year of his life.
In September, he got an email from a web developer in Halifax, Nova Scotia. He got lots of emails, but this one was different. Apparently the person who sent it knew something about Fei's private data visualization software. That was enough to catch his attention and earn a prompt reply.
Fei took a phone call with the web developer, who had created a popular chat site dedicated to junior mining companies in Canada. Some of the most risky and rewarding public companies on the planet, apparently. And very illiquid. The developer wanted to discuss adapting Fei's software to the Canadian equity markets. The developer already had a corporate license for the data and a small technical team implementing something similar.
The developer asked him "What do you think the Bloomberg terminal is going to look like in fifty years?" Fei paused, then replied "Oh, well you sure are asking the right questions. Let me tell you a bit about what I've built and how I've been using it. You see, I don’t really believe in predictions. Or, better said, I really don’t believe in predictive analytics. So I created something that shows all of the transactions and standing orders in the market…"
>>
