A recent episode of the Planet Money podcast profiled Thomas Peterffy, one of the first people to experiment and be successful with high-frequency trading. They told the story of how he was doing algorithmic trading before any of the stock exchanges supported electronic trading, and before NASDAQ even existed. So how did he do it? That’s the fascinating part.
He made his money building a system that was able to assign a fair market price to stock options. He then compared these values to what the options were actually trading for, and arbitraged the difference. Back in the late 1970s when he first started, he would print out the numbers and bring them to the trading floor in a huge binder. When the stock exchange banned him from bringing the binder, he stuffed the papers into every pocket his suit had.
Then Peterffy got himself a system called Quotron, a computerized service that delivered stock prices to brokers (it was a replacement for the widely-used ticker tape system). If he’d used the system the way it was intended, he would’ve read the quotes as they came in on the Quotron, manually input them into his algorithm, run the numbers, and cashed in. But that wouldn’t have been that much better than just using ticker tape, and the fact that he had a computerized system meant the data was in there somewhere, in digital form. If he could figure out how to retrieve it he could pipe it into his system and save a crucial, time-consuming step.
Nowadays if we wanted to do something similar, we might look into whether the Quotron had an API, and if it did we’d query that for the information. If it didn’t have an API, well, we might look for another system that did.
But Quotron had no such ability. So he did what any hacker worth his salt would do. He broke out his oscilloscope, cut the wires on the Quotron, reverse-engineered the data signal, and patched it into his system. And you think screen-scraping is hard?
When NASDAQ, the first all-electronic stock exchange, came online, he was faced with a similar system. Brokers could trade directly on the exchange via computer. This was no doubt a huge breakthrough, but there was still no way his system could make the trades automatically. So, again, he busted out his oscilloscope and patched his way into NASDAQ.
Eventually the folks at NASDAQ caught wind of this, visited him at his office, and reminded him that his terms of use dictated that trades must be made via keyboard input, not by splicing into the data feed. They gave him a week to comply with the terms. So what did Peterffy do? He built a robot to type the trades out on the keyboard. Of course he did. When the NASDAQ official returned a week later, all he could do was stand agape, in awe of what Peterffy had done.
We developers could learn from Peterffy. The ease of software engineering has made most of us too complacent. When Twitter’s API terms change, we complain about it for a few days, and then change our business models to suit the new rules. But the real innovation, the real interesting stuff, the way we’ll make $5.4 billion like Peterffy did, is by bending the rules and building systems that give us a leg up on the competition, or, better yet, improve people’s lives.
To be sure there are lots of hackers on the fringes of legality doing very interesting things, but the rest of us are somehow content to toe the line. We shouldn’t do anything that’s illegal, but we should get close. Innovation comes out of spurning the status quo, not complying with it. It’s time for people who know how to build things to bend the rules a little, and see what comes out the other side.
(The podcast was based on Peterffy’s story as told in the book Automate This: How Algorithms Came to Rule Our World.)