Hello! I like making predictions. I like measuring objectivity. I like self improvement and self reflection. Thats why I write down what I think and expect and then follow up with my own records years later to see where my thinking went wrong. Then I try not to be wrong again. One way I do this (a fun way of doing this) is gambling in prediction markets.
This is document is a summary of my performence in the political betting market during 2018.
A few years back I put a small amount of money into predictit.com, which is a political prediction betting market website. Each market is a question like "Will Bernie Sanders be the democratic presidential candidate in 2020?". You can buy a bet in the "yes" or "no" direction, and if the bet goes your way, you get $1, otherwise you get $0. Naturally some predictions are more likely than others. The price of very likely predictions are closer to $1, for example the prediction market that Vladimir Putin will stay in office in one year is usually close to 90 cents a bet. So if you make that bet, and Putin stays in office, you only win 10 cents ($1 - 90 cents). On the other hand, the market that Oprah will become president usually prices the bets very low. Its unlikely that Oprah will become president, and you will probably lose your money, but if you make the bet and she does, you get 10x-20x what you bet.
The individual bets are treated similar to stock shares, and you can buy many many bets, each one evaluating to $1 or $0 depending on if it turns out true. You can also buy and sell them as they change value, so you dont necessarily need to wait out until the bet finishes.
The cool thing about bets working in this way, is that the price per share represents an exact probability between zero and one. The collective market price of a particular bet is a pretty good estimate of the actual probability of these events occuring in real life.
I ended 2018 with 4% more money than where I started. For some context, the US economy as a whole grows at 2%. Index funds grow at 2%. The best performing hedge funds have around 30% return. So overall I feel good about 4%. I beat index funds and retirement accounts.
But aside from that, overall I dont feel too good. I saved 2018 at the last moment. If you looked at my year-to-year winnings just before or after January 1st 2019, I did much worse than 4%. My 2019 check in will include some big loses.
Evaluating Ruth Ginsberg's odds are a little morbid, because the liklihood of her death is an important factor. I computed a number for her liklihood of dying in the short term, and also a number for the likelihood of any of her colleagues leaving. The relative odds of her death and any of her colleagues leaving has been my primary metric for determining when to buy and sell. Basically, Ive just been buying and selling as the market rate goes above and below my pre-calculated rate. I think it worked well. The market went up and down, mostly in response to the most recent news reports about her health or her colleagues. I dont think theres anything of substance behind those short term changes, so patient people like myself can just buy and sell against the short term changes and win on average. Also I made money because the bet actually completed in my favor, because Anthony Kennedy retired.
I also bet in the "Anthony Kennedy is next to leave" market, in the direction that he would NOT be the next to leave. It was a similar story to how I treated the volitility in the Ginsberg market. Everyone has known for years and years that Kennedy had been thinking about retiring. Every few months there had been an article from an insider leak saying that Kennedy was going to do it any day now. Every time such an article came out the market jumps up a little bit, as if hes going to retire next week. I just bet in the opposite direction, treating the news as carrying no weight at all. The market would always gradually drift back to its previous position as the recent news faded into old news. Eventually he did actually retire, but at that point I had sold off all my shares that he wouldnt be the next to go, so I didnt suffer. I think the mistake the market was making, was always treating his inevitable retirement as being a few weeks away, when in reality it was months if not years away.
This one was really interesting. So Paul Manafort and his associate were charged and sentenced with crimes in 2018. Since thats the opposite of what I bet, it would seem like on the face of it I lost. I didnt lose however, because by time that came true I had already sold all my bets. Before that came true I won a lot by buying and selling at the right times.
My original thinking on these bets was:
- I have no idea if Paul Manafort is guilty of anything, and I am not particularly insightful on tht question over anyone else. So that entire category of information is worthless to me.
- I have hunch that the legal proceedings will last longer than a year. Proceedings are generally slow. I feel like theres a general "quickness" bias when people make predictions: when people are certain something will occur, they think that it will happen tomorrow, when usually the inevitable takes a long time to unfold.
- The bet conditions said "sentenced" and "charged", but what if Paul Manafort entered into a plea agreement? Thats a very likely outcome and it could technically mean he isnt charged or sentenced. People are probably distracted by the sexy hyperbolic information regarding his guilt, and not focused on the boring dry information about legal terms.
So I entered into the Paul Manafort market in the direction that he will not be charged or sentenced in 2018.
But then, the news quickly developed on this topic. It came out that Paul Manafort's associate was gong to testify against him; so Paul Manafort looked screwed. The market swung against me.
If you cant tell already, I am a big skeptic of the news, especially breaking news. If the market moves based on meaningless and unreliable information, then that means the better and more accurate bet price was the one before the news; before it was altered by useless information. So, I just doubled down on my Paul Manafort bet, and bet even more that he wouldnt be charged. I also entered the market regarding his associate, betting he would turn out innocent as well, even tho the news was reporting that he was making an implicit confession.
That worked out really well. The news completely flipped the next day saying that his associate wasnt testifying against him; treating it like a journalistic error. The markets made a huge swing in my favor. At the end of the day, I cashed out. The next day however, the news flipped again, but at that point I had nothing tied up in the market and I didnt feel that I knew what was going to happen next.
I talked about this in my previous prediction market check in. I bought into this market when the market only gave Trump 33% odds of staying in office through 2018. This was a big win for me. It felt like free money from the start. All I can add to my previous comments in my previous check in, is that almost every US president completes their term, and on a year to year basis, almost every US president finishes out the year in office. I think the amazingly low 33% odds that he would stay in office comes from a belief that the Trump presidency is exceptionally shaky and abysmal. It feels like that for me too, but then again, so has every US presidency I have ever lived through. I dont put any weight in that kind of information. Theres no objective value in what people comment on the current US president, so I disregard it entirely. That leaves me with the first approximation of US presidencies generally, which finish each year in office at a rate well over 90%.
I bet on the outcomes of two supreme court cases. Both of them had roughly 50% odds in the market. I won both of them.
For these bets, I relied on two sources.
- In college I read a ton of court opinions and academic law articles. I am not any where close to being an expert on any legal topic, but I do think reading a lot gave me a vague sense of what sounds like plausible and and credible law. So on these cases I asked myself "Does this sound like the law? Does this seem like the my best guess about the current direction the law is already pointing in?"
- Since I used to be into law, I think I know all the right blogs to go to that will give the most dry and straight forward legal analyses there are. So I just went to them and saw what they said. I dont think other people know about these blogs, so I didnt think that information was already factored into the market prices.
A bunch of news came out about some company exploiting facebook to gather user information for the purpose of getting political campaign insights in connection to Russia trying to influence the 2016 presidential election.
When following along a prediction market, one follows the topic as it unfolds, and one is always wondering what will happen next. At the beginning of a this facebook political campaign episode, it was unclear if it would explode and continue exploding into something bigger and bigger, or if it would be entirely forgotten about next week. I took the stance that it would probably fizzle out. I also took the stance that if it didnt fizzle out, having Mark Zuckerberg testify in front of congress is a bizarre and overly specific outcome.
So I bet it wouldnt happen. And I was wrong. It did happen.
I think I underestimated two things.
- I think facebook being attached to a news story is a bit of a controversey multiplier. If facebook is associated to a new story, no matter how tangentially, the story becomes super super news worthy. I think theres a vague nefarious connotation of a big company that knows about us that makes it a bigger deal. To some extent I knew this already, but I considered it silly. That lead me to neglect its relevance. But, however silly, its still a real phenomenon. Even if the controversey about facebook users and political campaigns was just sensationalist fluff, then whats fluff today will probably still be fluff tomorrow, and indeed the whole occurrence of testifying in front of congress is silly fluff for congress people to grandstand. It was fluff, but it was real fluff, and the bet itself was about fluff happening.
- I think having to testify in front of congress is a more normal consequence of big controverseys than I realized.
I went into this market with a lot of thoughts.
- Unemployment is really low right now. Almost tautologously true: it goes up after its unusually low.
- The unemployment data hasnt changed for several quarters. And for several prior quarters it had been going down. So it looked like unemployment was at a critical point in time, where it would change direction.
- Recessions and unemployment follow a pattern, and if you just eye ball graphs of economic data for the last several decades you can see patterns in it. We look like we are at the point in the recession cycle where things tip over and a recession happens.
So I thought the numbers were on the eve of changing, and that the expected direction will be up.
I was wrong. The unemployment rate didnt change at all. It stayed exactly the same between quarters.
What did I do wrong?
- I think now that its probably just hard to win in these markets. The points in my list above are really well known. Im not exceptional in my ability to spot them. You cant win in a market if you know and believe the same things as everyone else. You have to be able to beat everyone else.
- Quarter to quarter information is too short term and too chaotic. I dont think any of what I thought was wrong, but instead, it wasnt useful in predicting quarter to quarter events.
This was kind of a long shot bet that I expected to lose, but I expected to be a big win if I did win. The market had her at 86% odds she would be re-elected. I knew she was the democratic leader in the house of representatives. I knew that people usually stay in their leadership position for 9 years. I knew she was at her 9th year of leadership. I knew that her party really hadnt been doing well, and I had an expectation that poor performance usually leads to leadership changes.
I figured that if she does run, and being an incumbent, she would likely win re-election. But it seems from a biographical overview, this is where her career would end. So maybe she wont run at all, and instead she will retire. I guessed the odds she would be re-elected were closer to 68% than 86%.
But she did run, and she was re-elected so I lost.
I am noticing that my biggest losses are in new markets that just open up, on topics I havent thought about very much, that conclude within a few months. I should probably stop betting in those markets. Whenever these markets pop up, I always make a gamble, betting that my first impressions are better than the first impressions of others. It never works out.
I notice that on legal topics like court cases, I have yet to lose. I think I might have a better sense of how the legal system works. I think other people are really biased in how they perceive the legal system. I think when most people are predicting legal things, they are seeing it like its a partisan political conflict ("Party A is fighting party B, who will win?"). In reality its a bunch of non-partisan lawyers following rules ("Does rule A mean that condition B applies to circumstances C and D, and do we need to go through proceedure E to find out?"). Its fairly objective and boring, and the kinds of people who like making political bets dont seem to be aware of this.
I also notice that my wins are very long term. It seems to help to be involved in the market before it blows up. Its almost like, if theres some big exciting development, then its already too late to make good predictions about it. You have to get involved in a topic before it explodes. Find something thats currently uninteresting to the general public. Do a lot of research. Then make your bet. Then, when the topic blows up, react to the developments in an informed and prepared way. Once the topic blows up, new information is mostly worthless, but knowing how other people react to the new information can be quite useful, because then you know what the short term fluctuations are going to be.