The Wolves of K Street

https://www.theringer.com/2018/3/21/17130490/predictit-politics-election...

By David Hill Mar 21, 2018, 6:20am EDT

The United States government stood at the brink of crisis. For more than three months, there had been no agreement in Congress over an appropriations bill to fund the government. The Democrats had taken a stand to protect 800,000 young people who had arrived in the United States as minors and been protected by former president Obama’s Deferred Action for Childhood Arrivals policy, which President Trump had rescinded, giving it an expiration date of March 2018. Trump had won the presidency on a campaign that blamed the nation’s problems on immigrants, promised to deport them, and pledged to build a massive wall to keep them out. The issue divided the American people, and that division was now threatening to shut down the government.

Without a bill to protect these immigrants, Democrats refused to approve the legislation. The government’s funding was set to end at midnight on Friday, January 19, 2018. As of Friday afternoon there was still no path in the Senate to break a filibuster. If they failed to reach an agreement, the government would shut down. Hundreds of thousands of federal employees would be furloughed. U.S. government operations around the world would grind to a halt, and stay that way indefinitely until the two sides could come to a compromise. The situation was tense, the stakes were high, and I, like many other Americans, was glued to my television set watching the “shutdown clock” tick down on CNN.

I was captivated as senators huddled together in small groups around the Senate chambers trying to wheel and deal a compromise to break the filibuster. But while most of America was hoping for a deal to be reached and the crisis to be averted, I was biting my fingernails, hoping more senators would defect from the compromise, that there would be no agreement to save the “Dreamers” from mass deportation, and that the government would shut down on Friday night. I have always considered myself a politically engaged person, with fairly radical left-wing politics. On this night, however, my political values and hopes for the country had nothing to do with my interest. I was sweating a “no” vote because I had $500 riding on a shutdown on PredictIt.

PredictIt is a real-money political prediction market based in D.C. and sponsored by Victoria University of Wellington, New Zealand. PredictIt operates as a kind of “stock market for politics,” and is used to study the efficacy and value of markets in predicting future outcomes. The site was launched in 2014 not long after the U.S. Commodities Futures Trading Commission shut down a different prediction market, Intrade. Intrade was popular, but accepted bets on more than just political events, including the weather and the price of gold, which the CFTC considered a commodity future. PredictIt, however, had the approval of the CFTC to operate because the site’s work was for “academic research purposes only,” and because PredictIt had agreed to offer contracts on only political events and limit the amount that could be invested in a single contract to $850.

A PredictIt contract is essentially a bet. Each contract is a typical proposition bet between two people that an event will or won’t happen, only in this case the bet is structured like a futures contract. Traders can purchase “yes” or “no” shares in any given question. For every “yes” contract, there is another trader holding “no.” Traders can offer their shares for sale on the market at any price they wish. At the resolution of the event, the winners are each paid $1 per share. The losers’ shares are worth nothing. Throughout trading, prices will fluctuate depending on demand. A number of researchers believe that this type of market-based approach provides more reliable data than things like opinion polls or even expert opinions. PredictIt shares its data with professors at more than 50 universities, including Harvard and Yale.

This data collection and the association with Victoria University is what allows PredictIt to operate in the United States. It’s the same exception that has allowed the Iowa Electronic Markets, the O.G. of all real-money political prediction markets, to operate at the University of Iowa unmolested since 1988. Unlike the IEM, however, PredictIt is not entirely not-for-profit. Although it is owned and operated as an educational project by a nonprofit university, the founders of PredictIt, the brothers John and Dean Phillips, operate PredictIt’s software through their for-profit company Aristotle Inc. Aristotle takes a 10 percent fee from winning bets for their services.

Since 2014 PredictIt has seen enormous growth. On any given day the site offers hundreds of markets, on questions from “Who will be the next justice to leave the Supreme Court?” to “Will there be a federal gas tax hike in 2018?” The most popular propositions, however, are often hotly contested elections, such as the Alabama special election held in December to fill the Senate vacancy created by Jeff Sessions’s appointment as attorney general. That race, between Roy Moore and Doug Jones, saw more than 10 million shares traded between PredictIt users. In 2017, PredictIt users traded more than 300 million shares. That volume was largely the product of a spike in interest during the 2016 presidential election.

According to spokesman Will Jennings, PredictIt’s users are mostly young men between the ages of 22 and 35, primarily from large metropolitan areas. “People in finance and politics. A lot of people with heavy data backgrounds,” he says. “Of the traders I’ve met, one was a neuroscientist, another was a college professor, another was a math teacher.” But among the around 80,000 users on PredictIt live a small and dedicated tribe, a cabal, who have transformed political prognostication into more than just an art, but a lucrative business. “There is a group of power users who keep the site open,” says Jennings. “They live on the site.”

In the fall of 2015 Tom Gill was an undergraduate math student at Rutgers University. He hadn’t been much of an observer of politics before, but the Bernie Sanders campaign caught his attention. Gill found himself agreeing with Sanders and rooting for him in the Democratic primary. As the first Democratic debate approached, Gill wanted to have a little more skin in the game. “I liked the idea of putting my money behind my convictions,” he says. He deposited $10 on PredictIt, which he had seen advertised on Facebook. He bet that Sanders would see the biggest polling boost from the debate. He lost his $10.

A few months later, Gill had graduated and was working as a quantitative analyst at Jane Street Group, a trading firm in New York. The Wall Street traders he worked with loved to gamble. Prop bets were a part of the work culture. After losing some money to his coworkers betting against the Cubs, Gill turned to political prop bets to try to even the score. It was an election year, so everyone had strong political opinions. Gill would find lines on PredictIt that he felt were fair and then offer to take whichever side of the bet his coworkers didn’t want.

Gill hadn’t paid much attention to PredictIt while he was trying to settle in at his new job, but when checked back in, he noticed that PredictIt had grown. There were new and interesting markets and there were more people using the site. Since he now had some disposable income, he deposited a few hundred dollars on the site. By the time the Connecticut primary rolled around, Gill felt confident enough to deposit a few thousand. There were a handful of primaries that night on the East Coast, and Gill felt good about a number of the contracts he could buy. He bet it all. The next morning he had lost everything. He was wiped out. “I said to heck with this. It’s fun but it’s not worth it.” Gill swore off PredictIt. He didn’t deposit any more money, but he kept logging on and watching the markets. As the election progressed, the number of traders kept growing. Gill saw lots of people making lots of money. He wondered what he was missing. “How can I crack this code?” he asked himself.

Around the same time, Jon Kimball discovered PredictIt. Kimball started using the site in April 2016, just as the primaries were heating up. He was a close follower of politics and the news, and he thought he might have an edge by being glued to his computer and ready to make trades as new information broke. He was quickly hooked, spending all of his time hanging out in the comments section and monitoring Twitter for information related to his investments. Eventually he involved his childhood friend, David Rees, a comedian, author, and host of the TV show Going Deep With David Rees. Rees and Kimball teamed up with the writer and producer Starlee Kine, who had recently produced and hosted the hit podcast Mystery Show. The trio created a podcast called Election Profit Makers in which Rees and Kimball would discuss their PredictIt trades and the week’s news. The podcast was largely tongue-in-cheek, with Rees investing small amounts of money, often only a few dollars at a time. Kimball played the straight man to Rees, and he took his trades very seriously. As the general election approached, Kimball was up about $6,000. He had discovered that the best value on the site lay in betting against Trump supporters, who were legion on PredictIt. “They got excited about anything,” he says.

Gill noticed this, too. Advertisements on sites like Breitbart were bringing the Trump faithful to the site in droves, and they were backing contracts based on what right-wing media and often Trump himself were telling them was likely to happen. Both Kimball and Gill noticed that “yes” shares in the “Will a federal criminal charge be filed against Hillary Clinton?” market were wildly overpriced. “These people bet ‘yes’ no matter what because of the lunatic right-wing pundits they listen to,” Gill says.

“They were convinced Hillary was going to jail,” Kimball says. “I made a lot of money in that market.”

As the first debate of the general election between Hillary Clinton and Donald Trump approached, Gill decided to get back in the water. He thought he had finally cracked the code. There was a market asking whether anyone in the debate would say the words “fake news.” The pricing was split 50-50, a toss-up. It was a revelation to Gill. “This is unusual. Saying ‘fake news’ isn’t a toss-up,” he says. Gill thought it was highly unlikely, which meant that “no” shares were underpriced. “The kind of people betting on ‘yes’ were like me six months before.” They had heard the words “fake news” ad nauseam in the press and out of Trump’s mouth. Why wouldn’t he say it in the debate, they wondered. Gill wondered, Why would he? It wasn’t the kind of thing Trump said in his more scripted moments. It was something he uttered at campaign events, throwing red meat to his supporters. Gill invested heavily in “no,” and he won big.

Both Kimball and Gill went on like this for months, cleaning up while betting against Trump supporters. As Election Day approached, PredictIt had Trump at around 35 cents, where he had been for much of the election. Kimball invested in more than just Clinton winning the election; he also bet in individual state markets, markets about the margin of victory — whatever he could get his money into. He was all in for several thousand dollars. Fifteen minutes after returns started to come in, Clinton “yes” shares reached 90 cents, and Trump shares bottomed at 10 cents. Kimball watched his investments grow, excited both for the country and for his bank account. Then they called Florida for Trump. The Hillary “yes” shares started to drop. By the time they called Pennsylvania, Kimball was busted.

I bet on a shutdown because I believed Democrats had no other leverage, and no other choice, but to let the government’s funding run out. I believed their commitment to the Dreamers was real and that the Republican leadership and the president were similarly committed to their position. I couldn’t see how a compromise was possible. Plus, the price was irresistible at 20 cents a share. That price had more than doubled already, and I was contemplating selling my investment and taking the profits. But as the debate continued, and the “no” votes stacked up, the chatter on the talk shows and the internet grew hysterical. A shutdown loomed. Neither side was budging. Even Republicans were bailing on the compromises being floated. I was sitting on 5-1. Did I really want to sell out now? I turned to the PredictIt comments section for some direction.

The comments sections on PredictIt markets are wild places. They’re where you can find a debate raging about whether or not DACA will lead to “chain migration” or whether a shutdown will have a greater negative impact on Republicans or Democrats in the midterms. There are conversations about the pricing in the market and whether or not it’s fair, or what a fair price might be. There are people bragging about their positions that they’ve maxed out at $850 and encouraging others to get behind them. There are people bragging about having already unloaded their positions at a profit. There are people sharing tweets and links to news articles relevant to the market. For example, in this discussion I’ve already seen a tweet about a large order of Buffalo Wild Wings being delivered to a Senate office building — perhaps evidence of senators settling in for a long night — and a tweet about cheers being heard coming out of Senate Majority Leader Mitch McConnell’s office. The “no” shutdown shares climbed a bit after that tweet was posted. There are also more than a few Pepe frogs, Facebook memes, and instances of “cuck” being tossed around as a verb. It is, after all, still an internet comments section.

“The most dynamic part of PredictIt are the comment sections,” Kimball says. “You can learn so much just by reading the comments.” The hardcore traders on PredictIt are incredibly attuned to the world of politics. They follow dozens of congressional and White House reporters on Twitter. Their favorites are the reporters who tweet the most information, however trivial or banal. A fan favorite on PredictIt is HuffPost’s Matt Fuller, who is known to tweet rumors he overhears or to describe people’s body language or facial expressions as he live-tweets roll-call votes. Fuller’s tweets are sometimes so specific to PredictIt markets, many traders wonder if he is also trading on the site.

PredictIt traders traffic in information, because they make money by having the best information first. But this contest for inside information also presents opportunities for manipulation and even deception. “It’s sort of like Twitter in a way,” Kimball says. “You have to know who to trust. Just like on Twitter, where people can get tricked and follow a guy like Eric Garland and think he’s smart, there are people on PredictIt who tend to lie, who try to throw people off.”

Each market’s comment section begins with a warning:

The Disqus comment section is for informational purposes only and should not be relied upon when making any decision to buy or sell shares. PredictIt does not monitor, evaluate, or assess the accuracy of comments. PredictIt participants should seek out independent, verifiable information before making any investment decision.

It’s useful advice. It’s hard to know whether what you’re being told in the comments section is being posted by someone who has maxed out their position and legitimately wants more people to buy in so the price goes up, or someone who wants to buy the opposite side of what they are promoting and looking to drive the price down lower by getting suckers to buy the other end. It’s also hard to know whether what you’re reading is even real.

“I was dumb money,” Kimball says of his early days on PredictIt. “I would fall for tricks.” Once he invested in a market predicting the outcome of a Fox News poll. He didn’t have cable, and he worried that the news of the poll would show up on television before it showed up on Twitter or the web. He monitored the comments section of PredictIt for news about the poll. As Kimball refreshed the site, a user who had taken the opposite side of the trade posted “BOOM! I TOLD YOU!” Kimball quickly traded his shares, worrying that he was on the wrong end. But it was a head-fake. The poll wasn’t out yet. Another trader got Kimball’s shares on the cheap.

Legends of these kinds of shenanigans abound on PredictIt. A group of traders once created a fake polling company and made a bogus poll about Kid Rock’s chances for winning a Senate seat to represent Michigan. They linked to it on PredictIt to drive up the price, but the mainstream media found it and reported on it as if it were real. Multiple sources told me that another trader once sent RealClearPolitics a Photoshopped screenshot of some poll crosstabs, claiming to be an insider at a polling company, and RCP used the information in its polling average. The trader was able to make trades on the new average in a PredictIt market. Then after RCP discovered its error and removed the fake data, the trader was able to trade again on the price’s fall. Traders have posted fake tweets or posted links to old news stories claiming they were new. Groups of traders sometimes work together to start a cacophony of posts in support of a certain trade in order to “pump and dump” shares in a low-volume market. “In the end,” Kimball warns, “you just don’t really believe anybody.”

But Kimball didn’t totally ignore the comments, either. He worked to suss out who was good, who was bad, who you could trust, and who you couldn’t. He tried to read between the lines, to figure out what was a pump-and-dump and what was legitimate. “There were times where the ‘cool kids’ would show up and they would all be sort of chatting and talking to one another and goofing off, and you never really knew what position they were taking but you knew that something was about to go down. Do I want to just watch this, do I want to try to play this? Somebody is going to get hurt and it’s not going to be these guys, because they are sharks.”

I was in need of these cool kids. I was holding shares of “yes,” betting big on a shutdown, and I didn’t want to get hurt. I needed to know what the sharks were betting on. There was only one thing that could put my mind at ease about my investment. I scrolled through the comments section looking for Rainbow Jeremy.

The general election didn’t wipe Tom Gill out. He had some side bets that kept him afloat, like betting against a Trump landslide and accurately predicting Gary Johnson’s percentage of the vote. But after Election Day, after the shock of the Trump election turned to the shock of the Trump inauguration and the Trump presidency, Gill realized that there weren’t many more PredictIt markets to bet on. The elections were all over. The action dried up. How could he make back what he lost betting on Hillary Clinton if there weren’t going to be any big elections for several months?

PredictIt had a solution. It started a new market that would roll over week to week. It would ask traders to predict how many times each week the president of the United States would tweet. Gill put together a mathematical model he thought could give him an edge in predicting how often Trump would tweet. He put it to work and invested a few grand in the market the very first week it was offered. He woke up the next morning to discover the president had gone on a tweetstorm and Gill had lost it all while he was sleeping.

While many users on PredictIt were trying to figure out how to make money in these new Trump tweet markets, one user, who goes by the handle Jane Kay, was mastering the game. The 32-year-old pediatrician was becoming a known quantity on PredictIt despite rarely paying much attention to politics before setting up her account during the 2016 election cycle.

“Like a lot of people my age, I became more politically aware during the last election,” she says. “It seemed more important as Trump became a major contender.”

She read about PredictIt in a news story online. Though she considered herself very conservative with money (“All of my retirement accounts are in index funds”), she was inexplicably drawn to PredictIt. She liked debating in the comments section, where the anonymity allowed her to come out of her shell and blow off steam. “I could call people stupid, which I would never do in real life.”

She also loved that she could blend science and current events to make money. She was drawn to the polling markets, where traders tried to predict what polls would say before they were released. The markets were very mathematical, which appealed to Kay. She realized that understanding the math and doing some research gave her an edge. Like Kimball and Gill, she made money by betting against overconfident “Trumpers” who were “betting the Trump side no matter what.” By election night, Kay was up about $1,500, which was a lot of money to her at the time. Her confidence, the math, the polls, and the reliable irrationality of the Trumpers all failed her that night, however. She was all in on Hillary Clinton winning, and she lost it all.

“It seemed like a bad idea after that. It was my first foray into gambling,” she says. Kay works in academic medicine, mainly doing research, where the money isn’t as good as it is in general practice. She worried she didn’t have the financial freedom to lose money like that. “I thought maybe I shouldn’t do it anymore. But I had started learning about polling, I had a big Excel spreadsheet, links on my computer, I had done all this background work. I felt like I was getting my feet wet, so I stayed since I had done so much work.”

Like Gill, Kay also discovered the tweet market. She noticed that there were identifiable patterns to when Trump would tweet, what issues would set him off on a rant, what events in his schedule would lead to long breaks from his phone. She discovered that his schedule was available online. She also noticed that other traders would overreact every time the president would tweet. “Once the tweets came, it would shock everyone and they’d panic up. I could sell my shares and buy into the lower bracket that was priced too low.” Kay followed the patterns and found she was able to predict tweet after tweet. After watching the tweet markets for a while, she decided to deposit more money on PredictIt and take a stab at it. She made $2,000 in a single day.

Kay has never invested in the stock market outside of her 401(k) at work. She’s never gambled. She never talks about money with her coworkers or family. Last year she made over $17,000 betting on Trump’s tweets, a secret she’s never told any of her friends. “It’s embarrassing,” she says. “It’s a silly thing to be good at.” On PredictIt, Jane Kay is a legend. She’s known as a tweet specialist, rarely betting on anything else, unless it’s something really juicy. “Once and a while, [if] it’s really obviously mispriced … I’ll jump in. It seems like the Trump supporters are more emotional. Like the French election.”

The French election last spring was the first big political showdown to garner international coverage after the U.S. general election. Marine Le Pen was a far-right, anti-immigrant, nationalist candidate running against an establishment center-left candidate named Emmanuel Macron. There were obvious parallels to Trump-Clinton. When Tom Gill looked at the prices in the PredictIt market on the French election, he thought it was ridiculously mispriced. Le Pen was down over 20 points in every poll, but she was trading at 35 cents. Gill had a hunch he knew why the price was off. The Trumpers.

“These Trump folks had just made a lot of money on the general election,” he explains. “They had to listen to a drumbeat of media telling them that they were stupid. They had Trump at 10-1 and they won. So they learned to tune it all out. Now they think they’re going to win again.”

Gill did some modeling for the outcome of the election. He believed that not only was Macron underpriced, but that shares in the margin-of-victory market predicting a Macron blowout were also underpriced. Gill decided to take a stand. He invested heavily. In the end, Macron overperformed his polls and blew Le Pen out of the water. Gill made almost $7,000 in one night.

That summer, Gill lost his job at Jane Street and was contemplating his next move. He looked around at other trading firms, but he wasn’t excited about jumping into another office job. He wanted to work from home, ideally for himself. He considered trading his own money in the market, but he knew he couldn’t compete with the bankrolls and technology of the big banks and traders. He thought about PredictIt. He had been making money there, but it wasn’t consistent, and it wasn’t enough to live on — at least not at the level of comfort he desired. Was it even possible? he wondered. Could anyone make a full-time living on PredictIt?

He decided to give it a try. He had some ideas about how to make it work. For one thing, he’d need to be involved in a lot of markets, since there was an $850 limit. But that limit was a double-edged sword. While it limited how much he could get behind a bet he felt strongly about, it also created more opportunities for mispriced markets. If traders weren’t limited to $850, big bankrolls could push mispriced markets into more accurate pricing. The limits meant there would be more mispriced markets than there ought to be. “I developed an eye for those mispriced markets, picking out weird one-off situations. I hunt for those opportunities.”

Once Gill started playing PredictIt full time, his reputation grew quickly, mainly because instead of holing up in one market, his money was spread among lots of them at once. “I was everywhere on the site.” He started to make friends with other hardcore traders, and they would share information. “They would give me advice, I would reciprocate.” His advice proved valuable, and soon other traders were coming to him asking him what he liked.

Gill also discovered that once he was maxed out on a position, which he almost always was, then he had no incentive to keep his trades secret from the market. In fact, he had every incentive to shout it from the rooftops, to get others to agree with him and invest in the same position with him, and help drive the price up. He would take screenshots of his account screen and post them in the comments, showing others what all he had bought at what price, as proof that he wasn’t bullshitting them. Traders appreciated the honesty. When his trades turned a profit, as they often did, his legend grew. They knew him not as Tom Gill, but by his avatar, the British comedian Sacha Baron Cohen dressed as the character Ali G, and by his username, “Rainbow Jeremy.”

“There might be 100 people on PredictIt that make five figures,” Gill says. “There are only 10 who make six.” Gill is one of them. Less than one year after committing to PredictIt full time, Rainbow Jeremy is on the top of the heap, making between $15,000 and $25,000 a month betting on congressional elections and Trump’s tweets.

The hours wore on, and a government shutdown looked inevitable. My shares had climbed to 75 cents. Rand Paul was a “no” vote. Lindsey Graham was a “no” vote. Forget about the Democrats — it looked like there would be enough Republicans opposing a continuing resolution that this was going to be easy money. Then someone posted a tweet that said Lindsey Graham and Chuck Schumer just fist-bumped on the Senate floor. The price of my “yes” shares started falling. Rainbow Jeremy showed up and said he was going to buy “no” shares. I couldn’t click “Sell” fast enough. I sold everything at 70 cents, and bought back in to “no” at 30. Now I was on the No shutdown train, rooting for a deal. When it came out that Schumer had offered Trump $25 billion to build a border wall, in my heart I was sickened and mortified. But I was also excited that I might cash twice in the same market on opposite sides of the bet. I shared this with my wife, who sat next to me on the sofa watching this all unfold on C-SPAN. She looked at me with pure disgust. It was complicated, this feeling.

“I think we all felt that way, me and David and Starlee,” says Jon Kimball. “The idea of bidding up on a contract for Trump to do something that’s awful for the country and then he does it and you make money on it, that doesn’t feel right. On one hand you think ‘Why shouldn’t I profit on the chaos?’ On the other hand, the idea of rooting for that chaos to occur just so you can make a few bucks seems wrong. Or maybe not wrong, but there’s some sort of a conflict there.”

On the final episode of Election Profit Makers, the day after the election, Kimball and Rees and Kine aren’t able to carry on the same tongue-in-cheek shtick that had carried them through the election. They were sincere, angry, and sad. They cried. Kimball says that the money he lost wasn’t the point. There was much more at stake. He still believes that. He hasn’t been back on PredictIt. “It has a different feel to me now,” Kimball says. “It feels like we’ve entered into something that’s more important politically, and I can’t really be focused on PredictIt and the state of the country at the same time.”

But others are able to negotiate the trade-off that betting on PredictIt in the age of Trump requires. “I’ve bet on Donald Trump having success before,” Gill says. “That’s just the way it is. I don’t support him as a president. But do I think he’s going to stick around the rest of his term? I do. Not because I want him to, but because I think that’s what will happen.”

Jane Kay says that before she started betting on PredictIt, she wasn’t politically engaged. “I’m super aware of what’s happening every day in the political world now. … Because Trump is president, that made everything seem more important and urgent than they did in the last eight years.” And there’s still money to be made betting against the scores of Trumpers using their money to back the president and his positions on PredictIt—data and fake news be damned, taunting and trolling, win or lose.

“I just tell myself I took their money,” Kays says, “and it doesn’t bother me too much.”