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 what I 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 2019. Here are some previous write-ups I have done on previous years of predicting.
- heres my 2018 betting market run down
- heres my 2017 betting market run down
- heres my most recent set of calibration predictions
This is the point in these check-in write ups where I explain political betting and predictions, but I have already done that in my previous check-ins for previous years. So if you are asking something like "what is this?"or "what are you talking about?", read the beginning sections of last years check-in.
Pretty bad. This was my worse year yet. I dont have exact records, but I think I lost ~13% of the value of my predictit.com betting portfolio (no biggie, I started with onl $100 anyway). That said, 2019 was perhaps my most fun and educational year. I did lose a lot, but after doing this prediction stuff for 4 years, I have developed a new satisfaction with being wrong that I didnt start with. Losing roughly means I learned something, which is nice.
Overall I lost really hard predicting who would run for president. What happened?
Here is a lot of background. I have known a lot of people who do not trust political authority figures. At least from my perspective, it is pretty popular to believe that the worst people end up at the top of society. But despite hearing this idea so many times, I never heard anyone provide any deep reasoning behind it. I figured that this popular distrust of political authority comes from deep innate prejudice rather than rational judgement based off observed evidence. Indeed, I started to believe the opposite; that the best rise to the top.
In 2016, I wondered, what if I assume the opposite of all my highly prejudiced peers? Would the world make more sense? Could I anticipate the future better? After the 2016 presidential election I consumed a lot of political media, like interviews and podcasts with democratic politicians, and inevitably I heard many candidates get pitched the question "Are you going to run for president?". My betting model was this: they are elites and elites are better people, so I presume their answers to the question "Are you running for president?" will be truthful, and that I could just go off their answers directly to predict the future.
Is this a good model? Does it beat the market? No. It doesnt. It fails incredibly. Every single democratic candidate that said they werent running for president did in fact run for president. I think they all lied.
Why did they do that? Well, I am speculating here, but I think its marketing. Marketing campaigns- whether they be for movies, video games, or election candidates- need to go along a planned timeline designed to drum up excitement. If you are a democratic politician running for president, you need to schedule your popularity to peak around 2019 when the democratic party is close to choosing its candidate. If you start your campaign too early (like in 2017), your campaign will run out of steam. People will hear that you are running for president, get excited, but then gradually grow bored of you. The day you say you are running is day 1 of your campaign and thats simply how it is. So if today is not the day you should begin your marketing campaign then the correct answer is no. "No" is how you retain control of your campaign. Furthermore, elections are competitive. If you havent announced yet, your competition has to play defensively, because they know you have the power to announce your candidacy and steal their thunder.
What about all those peers of mine who distrust political authority? Did they know better than me? Well, I guess so, but I still think they are deeply prejudiced against authority. The lesson I am choosing to learn, is that being biased doesnt mean you are wrong. They were right and I wasnt. Being biased even has certain advantages over rational thought: thinking is sometimes expensive in terms of time and effort, and biased gut-feelings are basically automatic and free. Imagine if your gut feeling is correct 60% of the time, but spending an hour researching a question leads to correct answers 70% of the time. Well, is that hour worth 10% more accuracy? Possibly not; its depends on the question. Isnt it also possible that for some questions, where gut feelings are pretty well calibrated to start with, and where more rational thought would be directed to all kinds of tricky intellectual traps, more thinking actually leads to less correctness?
What does it mean to run for president? I didnt really think about that question too hard when I made my bets. Predictit.com considered someone "running for president" when they filed with the election commission. That is a very low bar and tons and tons of candidates do this regardless as to how much campaigning they intend on doing. It seems like if politicians are even thinking about running a little bit they go ahead and file their candidacy.
Also, this was after the 2016 election, where for most of the democratic primary the possible candidates were Hilary Clinton and Bernie Sanders. Two candidates is a small group, which I inferred would indicate a small pool of candidates for the democratic 2020 election primary. This time around the democratic party changed their primary rules which enabled a large number of people run for the democratic presidential candidate. How things went for the 2016 democratic primary did not repeat for the 2020 primary.
In 2017, I eye-balled a lot of the contenders for the democratic ticket at the time, and Kamala Harris was my pick for most likely to win it. She seemed tough and I figured that represented a kind of effectiveness that people would appreciate. I considered at the time that people would be put off or creeped out by someone who seemed authoritarian like she did, but I dismissed that thinking that people have much baser and simpler political interests. As if, people just want to win and toughies win and higher minded ideas about fairness or good character are less popular. Also, she met the demographic checkboxes of being female and non-white, and I figured that was something people care about.
But she didnt ultimately win the primary. Where did I go wrong?
People did not like her creepy toughness. I guess I neglected the extent to which ordinary moral feelings like "truth is good" and "its bad to put innocent people in jail" matter to how people feel about candidates.
Ignoring all information roughly approximates letting new information influence you to a small degree
The best forecasters follow a general strategy of updating their beliefs frequently, but (and this is key) also not letting their judgement be swayed very much by each new bit of information. If bits of information were like pebbles, one new bit of information should not change your judgement very much in the same way that adding a single pebble to a huge pile of pebbles wont change the weight of a huge pile very much. So, your technique for predicting the future should be to update your beliefs often but in small amounts. For a lot of bets, I have been able to do a lot of information ignoring quite effectively. This ends up being effective, I think, because "ignore information entirely" closely approximates "update your opinion by small amounts". Zero is a small amount. For Kamala Harris, I wanted to do a lot of information ignoring, not for any particular strategic reason, just because it works in theory and in practice and I wanted to spend my free time doing things other than checking in on Kamala Harris every day.
But I think I went too far with the information ignoring. Kamala Harris had a whole arch of popularity in the primaries, wherein many new bits of information became available, and I did not pay attention to any of it. For example, I should have watched the democratic primary debates where she apparently flopped pretty hard. Considering I made my bets early when the price was low, if I checked in more on her campaign on a daily or semi-daily basis, I might have recognized her popularity peak and made money selling my bets. But I didnt, so I lost.
However, Kamala Harris was picked as Biden's VP; so maybe I did have some vague insight in 2017 that Kamala Harris has some real significance? Either way, if I did it was not reflected in the bets I made.
So, none of these people were selected as the democratic candidate, which looks like a lot of losing, but I actually broke even on these bets. Pete, Mike, and Amy all had their moments of popularity in the primary election, and I did a good job anticipating their popularity before it happened, immediately before it happened. I am a little proud of that, but I failed however, by not selling my bets before their popularity died. So, I just broke even.
I didnt think Biden would run at all, because he said he wouldnt. I judged Elizabeth Warren as a unlikeable person and therefore undesirable as a presidential candidate; and I figured that the democratic party knows shes not a good candidate and would not select her, and (bear with me) I figured she knows that the democratic party knows shes not a good candidate, so she would know its pointless and not run in the first place. She did run, so I was wrong (I guess it doesnt hurt to try running even if you expect to lose), but she did lose the primary, so I feel like I was right regarding her merits. I thought Bernie Sanders would in fact run because he always runs, but I also thought he would lose roughly because he always loses because he is an extreme candidate (altho I guess extreme candidates are getting more popular?).
An idea that guided me in a lot of these bets about who would win the democratic candidate, is what I call the "rational party hypothesis". Its the hypothesis that political parties want to win and will operate in such a way that they will select candidates who are most likely to win.
I believed in this hypothesis in 2016, and now I dont really believe it. If the democratic party chose their candidate by letting the biggest boss of the democratic party choose whoever he or she wants, then I think the choice would be pretty aligned with winning. But, thats just not how organizations work. There are many people involved in the process, and most of the decision making is done indirectly via election rules and delegates. So, no one really picks, theres just a chaotic process and someone ends up as the candidate at the end of it.
I kind of suspect also, that the democratic party has its own internal politics at work, and that this internal politics distorts the result away from "most likely to win".
All this together led me to overestimate the odds that young, moderate, and level-headed newcomers like Pete Buttigeug; and to underestimate old establishment people like Joe Biden. I still think Pete would have been a great candidate; much better than Biden.
I notice I did a really good job anticipating the momentary popularity of many democratic candidates. I knew who was going to become popular, and when, and I knew who was not going to become popular. This didnt help me much in the end however, because many candidate were popular only for a limited period of time, and I did a bad job anticipating when each of these waves of popularity would end. I was never able to "cash in" because I would be selling the bets for roughly the same amount I bought them at.
I should probably reflect more one why I could forsee popularity, but not forsee unpopularity.
These bets were my big wins on predictit.com for 2019. Among these bets the average market odds was roughly 50%, so people thought it was very likely Trump's associates were going to get busted for doing something illegal.
Why didnt I? Well, I dont know if I can add much to what I have said in previous check-ins. I took the apriori position that the Trump presidency- as a presidency and not a media spectacle- is fairly ordinary and corrupt to a historically average degree. The associates of the president getting busted for crimes is historically rare, so, apriori we should expect it to be unlikely for the current presidency as well.
I would like to note that I didnt bet that Trump himself wouldnt get impeached. That seemed very likely to me and I felt the market was already close to the accurate odds. Firstly, because presidential impeachment is more historically normal than it sounds on paper (almost ~7% of all US presidents excluding Trump). My feeling was that Trump getting impeached fell more into the category of politics than criminal justice. Impeachment sounds like "busted for a crime" but its just something put to a vote by congress people. The democratic party wanted to do it, so they did. Compare that to minor players like family and administration officials getting charged for a crime; the political spotlight shines on them less, and the charges brought against them go through a real legal process.
Also, it seems like a lot of people really really believe in a strongly political way that the Trump presidency is "not normal", so I am a little worried that my asserting that it has in fact been normal is like an offensive comment to some people. I followed along these Trump related betting markets since 2016, and I felt it was quite amazing what people were betting. In 2017, it was a very popular bet that Trump would not even finish his term. It was a popular belief that the Meuller investigation would expose something horrible about Trump, and destroy him. But neither of these things happened, so the "Trump is not a normal president" model of the universe did not predict the present. So, I think its just not correct.