It’s January 9, 2021. Yesterday, two days after the President (who by the way is Donald Trump, the guy from The Apprentice) ginned his supporters up into storming the Capitol (wtf?) and coming within minutes of possibly lynching the Vice President (okay, maybe this wouldn’t have happened), Twitter finally discovered the justification to shut down Trump’s account. This is all horribly and hilariously and unimaginably absurd.

Let’s get even weirder. For the better part of the last four politically insane years, a community of gamblers wagered stupid amounts of money betting on a simple question: How many times would Donald Trump tweet this week? The game ended for us before it ended for the President, but now that it’s completely over, I feel this tiny corner of Internet weirdness deserves some remembrance. After all, there are very few other people on this planet that understood Trump’s twitter habits – and by extension, Trump himself – more than the people who bet on them.

The Genesis

It starts with PredictIt, which of course readers of this blog are familiar with. In 2016, the run-up to the election featured a suite of recurring weekly markets in which traders bet on who would be leading in the polls, Congress’ approval, the direction of the country, and Obama’s job approval. These were fairly successful for the website in that degenerate bettors loved markets where small bets can hit big (which they did) and savvy bettors enjoyed the game (and gamesmanship) that came with figuring out how polling averages worked, polls were released, and how that information eventually reached the market. In the process, a community formed. In particular, the discussion boards for the Obama approval markets (OA) became a place for happy numerically-inclined (and often left-of-center) traders to hang out. Thursdays, for example were “scotch-and-base” nights when a trader (“baselevel“) and others would drink scotch and answer questions from newer traders who had no clue how polling markets worked.

The OA crowd got tight-knit enough that after the election, an IRL get-together was planned for Chicago (in the middle of the country). The organizers were baselevel and m4ry and it was relatively well-attended, with I think around 20-30 traders showing up in what was the first or one of the first real-life PredictIt trader meet-ups (I didn’t go – at the time my net PredictIt winnings were literally $1200 and I couldn’t justify the cost). John Aristotle, the founder of PredictIt, was so jazzed there was a meetup happening among people who met on his website that he even flew out from DC and came by himself, talking a lot about his big plans for site expansion and so forth (as I understand it).

Anyway, a few weeks after this meetup in the middle of December, John reached out to some of the traders he had met. Polling markets were drying up a bit and he wondered if the polling market crew had any ideas for any new weekly recurring markets. At this point, I was involved in the twitter group chat with a handful of the OA veterans, and we began tossing around ideas. At this time, the mainstream was just getting used to the idea of paying attention to Trump’s tweets. Every tweet was a news story in November and December of 2016. He was threatening GM and Chrysler via tweets and nuking their stock prices (briefly). Quants on Wall Street started building bots that would keyword search his tweets for relevant companies so they could automatically trade in response to a twitter attack (the stock market was the first tweet market). There was this notion that Trump “used” his twitter account strategically (we’ll get there, but any tweet market trader would laugh at this now), and his “grasp” of social media was being talked up and praised. Here was a President-elect who posted his own tweets in a way that was unpredictable and – at that time – seemingly unavoidable.

So in our brainstorming session, at one point I said something like “you know what would be cool – if we had a market on how many times Trump tweeted in a week”. I’d gone and done a PhD at this point and yet this tiny little thought ended up snowballing into the most lucrative idea I’ve ever had in my life (and probably changed the course of my life – which is a deeply strange and weird thing to reflect on now). Others immediately agreed, and we began working on it. Dmp (another trader) wrote up an email proposing a more fleshed-out idea (m4ry and I did some research on possible brackets) that everyone had contributed to at that point. We sent it in to Aristotle, heard back that he loved it, and then heard nothing else. (One note here: we weren’t the only ones thinking of this as a market idea, at least one other OA person, a trader that goes by “klumein” also posted about a week after we sent the email in that a market on Trump’s tweets would be amazing. I’m sure there were others.)

There was a problem in our proposal: we didn’t have a good way to keep track of the count. We suggested just manually counting the tweets, but we knew that the PI resolution people probably wouldn’t want to do that. As December went away, we figured our market idea was dead or otherwise found unworkable by the powers that be. Imagine our delight, then, when early January brought us this:

https://www.predictit.org/markets/detail/2817

The very first tweet market. PredictIt had solved the tweet count problem: apparently you could just hover your mouse over the number of tweets on a user’s homepage and even if they had thousands of tweets it would give you the precise number. The game was on, and everyone was stoked:

The Early Days

The early markets were instant classics. That first market? It did 1.7M in volume. And he barely tweeted back then! Look at these brackets, those of you who only came to discover tweets in 2018 or 2019:

People had absolutely *no idea* what they were doing. 50+ opened at 33c then rose to 50c then fell back down then did it again (it hit 50c or higher six out of seven trading days and ultimately lost). Then there were the complex rules. Did retweets count? What did deletions do? We had to figure it out ourselves:

People didn’t know what “the count” was. A tweet would happen and it would take a minute before the market reacted and the comment section had sorted out what the current count was. Every time a tweet was posted, someone would post “BOOM” in the comments, leading to a cottage industry of shitposters spamming “BOOM” every so often just to see if people would react (hi, JonH). When a tweet killed a bracket, it would actually take longer than a second for it to die (no one was using Twitter’s API back then).

[By the time we got a few months in, people started creating websites to track the count publicly – there was cabalcount.xyz (now defunct) and then, most notably, picount.com created by PI trader Abe. This fantastic resource made it so everyone knew where the count could. Eventually, we even got to the point where traders like Dave were offering paid services of a twitter bot (pitwets, with free version pi_alerts) that would notify you of each accounts tweet instantly, the count, whether a deletion happened, what the “pace” was and so on. An entire cottage industry built within this little fairy garden of the internet!]

The first market also birthed the first Trump tweet analytics, with m4ry among the first to make a spreadsheet tracking his habits and attempting to model them:

And all the OG traders were there. The very first tweet market? RJ took an enormous bath:

Even Domer, who later came to more or less hate tweet markets (well, the non-Trump ones especially) enjoyed himself:

The early markets were pure. Degenerate? Very. But we were living in a world where Trump was about to be President. Weird stuff happens now. So, sure. Why not bet on how many times a madman will tweet?

Switch to POTUS confirmed??

As inauguration loomed, so too did a huge question for tweet traders. Would Trump keep tweeting on @realDonaldTrump, or would he switch to @POTUS? That market (I can’t dig up the link right now) was absolutely insane. The first few posts following his dark inaugural? All went to POTUS. The lowest bracket went north of 70c. Later in the afternoon? Trump tweets himself. Back down to 20c. Then back up. Over 4M was traded, with the final result landing again somewhere in the 40s for the weekly count.

With the uncertainty of an account switch, PI obliged us by creating a second tweet market for the @POTUS account, just in case. But then both markets did very well, and after a while – why not one for @VP as well? By summer 2017 we had four weekly twitter markets: @realDonaldTrump, @POTUS, and two for @VP (one that ended Tuesday and one that ended Friday, which never confused anyone ever causing them to make a trade in the wrong place). At one point in 2019 we had like twelve tweet markets (it was too many).

Tweet markets consume all

Around this point, it became clear that you could actually kind of make money trading tweets. I had started writing guides explaining how the markets work. Traders were getting more sophisticated. We understood what we had to do to win. For Trump’s account, that meant we had to know him. We had to learn when he tweeted and when he didn’t (waking up at 6am on a Wednesday was absolutely required – for years the “bong” of my computer signaling a Trump tweet was my alarm clock). We had to learn why he tweeted what he did (he saw it on Fox news – I bought a YouTube TV subscription at one point so that I could find where he was in his DVR-viewing-sesh and scroll forward to find likely topics). We had to learn how to get the tweets as fast as possible (I had begun learning to code in grad school, but it was PredictIt forcing me to do so for polling and twitter markets that actually taught me). We had to know when there were deletions. We had to learn how the markets reacted to a tweet. In effect, tweet markets taught me how to trade. (There’s honestly probably a prediction market research paper waiting to be written on tweet markets about how just the existence of a novel market type produces smarter prices over time as the participants gather experience.)

Yet for all the time-consuming nature of these markets (one of the things about Trump is that the man could get himself going at literally any time), they remained probably one of the healthiest outlets out there – for me at least – when it came to processing the manifestly insane shit that the President said. When you saw an amazing tweet from him, it wasn’t just some alarming thing that a reporter put into your timeline. Now there was an entire community of traders digesting it right along with you, marveling at how stupid or out of left field the tweet was, its typos (would he delete and repost?) and what it meant about what he cared about. Many times you’d see a Trump voter in the twitter market comment section even remarking how nuts Trump’s latest missive was. Looking back, it’s actually shocking how little political arguing there was among the tweet bettors despite it being full of folks on left and right and centering on Donald Trump. For better or worse, tweet markets took at least some of the craziness of this Presidency and boxed it up into an easy-to-understand game where everyone was more or less on the same playing field.

Trump is predictable

One of the biggest false memes in coverage of Trump and his twitter accounts over the years was how his tweets proved how unpredictable he was. In a sense – sure. He could tweet at random times, but mainly he tweeted while watching TV in the morning and the evening. Yes, sometimes things came out of left field, like when he wished Happy Birthday to Maxine Waters at 10am on a Wednesday morning, completely up-ending a market. But overall, betting on his tweets taught you how he worked. The profit incentive offered by the market forced you to learn how narcissism operates. There is a logic there. You just had to have a lot of money on the line to understand. A few lessons from over the years:

-Big news made him go quiet more often than not. When he doesn’t know what to say, he waits. He was quiet after firing Comey became a huge thing. He was quiet immediately after Charlottesville. He was quiet just now, after the Capitol was stormed. You’d get some tweets, sure. But you wouldn’t get a lot of tweets. He’s more conscious when he knows everyone is definitely paying attention. It’s when he was feeling loose that you’d really get some special tweets, or lots of retweets.

-He rarely if ever used twitter as a strategy. There was never a grand design in his tweets (other than some spam endorsements). He just posted. Mainly, if there was a goal, it was to shape the reality perceived by his supporters into the reality he wanted to live in (or thought that he lived in). For over a year leading up to the 2020 election, he would tweet stuff like “96% support in the Republican party – thank you!” Why? To remind people that he had control over the GOP but also to reaffirm the same to himself. You’d typically see these tweets during times of intra-party friction. The number, of course, was more or less completely made up, having derived from a straw poll of CPAC attendees in 2018 and metamorphosed from there (the original number was 93%, it got up to 97% by the time Trump was done with it).

-He was tweeting to his supporters. To him, if you followed him it was because you liked him. His follower count was one of his most prized possessions – narcissism craves validation above all else. He wasn’t ever really tweeting to the media or to his enemies. He’d tweet at them, sure. But the point of the message was always “hey guys, this is how we think about this, ok?”. His trademark ending of tweets with something like “Sad!” encapsulates the whole thing. He’s literally communicating how he feels and how he thinks you should feel too and he wants to make sure the message is heard loud and clear.

-Anger did not produce tweets, but fear did. Donald Trump is deeply insecure, of course. The thought of people laughing at you behind their backs or being otherwise humiliated is the biggest injury someone could ever do to you, in his mind (hence his use of that comparison in the soon-to-be impeachable phone call to the Georgia SOS). When Joe Biden received a firefighter’s union endorsement in Pennsylvania, Trump erupted with over 60 retweets of various purported firefighters that replied to some other tweet angry about it (there were several other PI traders in the replies hoping to land a retweet themselves so they could delete it later and flip a market at some point). Trump hates losing something he thinks belongs to him – his insecurities can’t stand it, hence the furious retweets signaling that in fact he hasn’t lost anything at all. (He will also have to lash out for losing his twitter account at some point. He needs attention and will find another way to get it).

-Retweets and his growing use of them in the back half of his presidency remains underappreciated. There was a time in 2017 where he barely knew how to use the platform (it seemed). He couldn’t thread his tweets, and would often end a tweet with “….” and start the next with “….” to signify they were connected (the two tweets might come hours apart). By the end of his presidency, he threaded tweets himself and engaged in hour-long retweet sprees, often going for 100+ retweets. He began to use the retweet button as a quick and easy way to amplify messages and people that he wanted his followers to internalize. His habit was to stumble on a good account, click through to the username, and scroll through their posts, retweeting as many as he could until he got bored and found another username to scroll through somewhere in the replies. No one else tweets like this. And no one other than tweet market traders really understood what was going on – by the time this habit formed, the media was more or less bored by his tweets and no one else was paying that much attention (probably not even his followers).

It’s All Gone Now

Of course, it’s over and these lessons learned don’t really matter. The markets died first in June of 2020 – victims of regulators who finally decided they crossed the line (the fact that at one point we had markets for Yang Tweets and someone trading on them threatened Yang’s life if he didn’t stop tweeting, getting the FBI involved, probably didn’t help). I’m pretty sure I legitimately grieved their loss, which is both sad (in the pathetic way) and also kind of amazing (also in the pathetic way). I mean, what a weird fucking hobby! The spreadsheets I maintained… good lord. While the money was cool (around +$120k lifetime in tweet markets), the fact that I have memories and stories about “that time when Trump retweeted 60 firefighters” kind of says it all. Normal people do not have these experiences! And all starting with one random kernel of an idea I had back in December 2016.

And now Trump’s account is poofed! Kind of have the feeling they’ll bring it back, maybe in some archival form in a few years. (Just had a thought: the Trump Presidential Library will probably have some sort of exhibit on his tweets). But what a change. No longer is the “I wonder what he’ll say about this” itch going to get scratched. Will GOP politicians try trashing him more often now that they don’t have to care about him tweeting? Are all those primary threats vacant now that Trump can’t drive headlines? Will we care as much about Trump’s 2024 run prospects without being able to rely on him keeping us informed? Or will he re-emerge elsewhere? (I’d assume so).

Whatever happens, we probably won’t be trading on how many times he Parlers (is that the term?) or whatever. So rest in peace to very weird corner of the internet that tried to make sense of a very weird presidency.