There’s a failure mode where pundits make lots of predictions (“Iraq will be full of WMDs!”, “Clinton will coast to victory easily!”), get proven wrong again and again, but keep their editorial columns and TV shows and reputations. If you’re wrong often enough, at some point people should stop treating you as a sage and start looking for wisdom somewhere else.

There’s an opposite failure mode where somebody is wrong once about some freak event and their opponents demand they never express an opinion again.

The solution is to predict probabilistically and keep careful track of your results. For example, I can say “I think there’s a 75% chance that Clinton will coast to victory easily”. If I’m right about most of my probabilistic predictions – for example, my 75% chance predictions come true about 75% of the time – then you can trust that I know what I’m talking about. If I’m wrong – for example, my 75% chance predictions only come true 25% of the time – then I don’t know what I’m talking about and you might want to ignore me.

Probabilistic predictions naturally lend themselves to bets. I try to take advantage of this to convert vague philosophical disagreements into testable arguments on real points. For example, in a thread about whether Donald Trump was racist, myself and some commenters were able to clarify our disagreements into a few points about what Trump would do on specific racial issues, then make bets about whether or not he would do them. If I’m right, I profit off of my opponents’ ignorance; if I’m wrong, I at least prove that my convictions were honestly held. In either case, both me and the people I disagree with get some real and hard-to-avoid feedback.

Every year I make many predictions about upcoming events. See 2014, 2015, and 2016. Those are recorded separately on their own pages. This page is for scattered predictions I make in the course of blogging about other stuff.

I am happy to make bets on these predictions (or my collected yearly predictions) with anyone who disagrees. Because of transaction costs, I will bet you only if our estimates differ substantially – I won’t bet if I think the odds are 50% and you think they’re 49%. In general, the “steps” I use for predictions are 50%, 60%, 70%, 80%, 90%, 95%, 99%. I will make a bet with you if you are willing to give me odds at least two “steps” away from me. If I predict something is 50% likely and you think it’s 70% likely, then bet me at 7:3 odds. If I think something is 99% likely and you think it’s only 90% likely, then bet me at 9:1 odds. If I think something is 99% likely and you think it’s 95% likely, then it’s not worth my time and I probably won’t bet you.

On some predictions/bets, the outcomes will be obvious – for example, how much the Dow Jones increases during the space of a year. For others, it might be controversial – for example, whether or not Trump does a bad job as President. If I’m making a prediction without a bet, I will trust myself to score subjective predictions. If I’m making a bet, then myself and the person I bet will agree on a neutral arbitrator to score the prediction.

Here I’m recording all the predictions and bets that I’ve made since 1/1/17

As of 11/15/16:

1. Total hate crimes incidents as measured here will be not more than 125% of their 2015 value at any year during a Trump presidency, conditional on similar reporting methodology [confidence: 80%]
2. Total minority population of US citizens will increase throughout Trump’s presidency [confidence: 99%]
3. US Muslim population increases throughout Trump’s presidency [confidence: 95%]
4. Trump cabinet will be at least 10% minority [confidence: 90%], at least 20% minority [confidence: 70%], at least 30% minority [30%]. Here I’m defining “minority” to include nonwhites, Latinos, and LGBT people, though not women. Note that by this definition America as a whole is about 35% minority and Congress is about 15% minority.
5. Gay marriage will remain legal throughout a Trump presidency [confidence: 95%]
6. Race relations as perceived by blacks, as measured by this Gallup poll, will do better under Trump than they did under Obama (ie the change in race relations 2017-2021 will be less negative/more positive than the change 2009-2016) [confidence: 70%].
7. Neither Trump nor any of his officials (Cabinet, etc) will endorse the KKK, Stormfront, or explicit neo-Nazis publicly, refuse to back down, etc, and keep their job [confidence: 99%].
8. No large demographic group (> 1 million people) get forced to sign up for a “registry” [confidence: 95%]
9. No large demographic group gets sent to internment camps [confidence: 99%]
10. Number of deportations during Trump’s four years will not be greater than Obama’s 8 [confidence: 90%]

As of 01/06/17

11. Donald Trump remains President on 12/31/17: 90%
12. No serious impeachment proceedings are active against Trump by 12/31/17: 80%
13. Construction on Mexican border wall (beyond existing barriers) begins by 12/31/17: 80%
14. Trump administration does not initiate extra prosecution of Hillary Clinton by 12/31/17: 90%
15. 2017 US GDP growth lower than in 2016: 60%
16. US unemployment to be higher at end of 2017 than beginning: 60%
17. US does not withdraw from large trade org like WTO or NAFTA by 12/31/17: 90%
18. US does not publicly and explicitly disavow One China policy by 12/31/17: 95%
19. No race riot killing > 5 people by 12/31/17: 95%
20. US lifts at least half of existing sanctions on Russia by 12/31/17: 70%
22. Donald Trump’s approval rating at the end of 2017 is lower than fifty percent by 12/31/17: 80%
23. …lower than forty percent by 12/31/17: 60%
24. Highly-publicized stories about Trump successfully keeping businesses in the US on a case-by-case basis, which never add up to a significant number of jobs saved, will keep coming, and be a central point of how his administration relates to the public over 2017: 50%

As of 4/6/2017

25. Significant increase in the number of US combat troops on the ground in Syria this year: 70%
26. Fewer than ten US soldier casualties in Syria this year: 60%
27. No very large (eg Iraq-scale) US intervention in Syria this year: 95%
28. Islamic State loses control of Raqqa this year: 95%

As of 4/9/2017

29. If there’s ever clear evidence about who was responsible for the chemical attack on Khan Sheikhoun, it turns out to be someone other than Assad. For this purpose, local commanders acting without orders from Assad counts as “not Assad”: 70%

As of 4/23/2017
1. In terms of final winner of the French election, my odds are 15/60/10/15 for Fillon/Macron/Le Pen/Melenchon respectively.

As of 11/06/2017
1. Catalonia will not be de facto independent a year from now (ie monopoly on force in its territory): 90% chance
2. No real war (500+ deaths including armed combatants on both sides) in Catalonia in the next year: 2%


1. In December 2016, I bet that most people would misinterpret a New York Times article about economists on school vouchers source). The other side later backed out. A survey determined I was right, but I won $0.

2. In December 2016, I bet Edmund N $40 that Donald Trump’s cabinet would be at least 20% minority. Depending on how you count it, it was somewhere around 16% minority. I lost $40.

3. In April 2017, I bet Zack D $500 on a somewhat private point about the inner workings of psychiatric hospitals. It originally appeared that I won, and I received the money. Later it became pretty unclear, so Zack and I reached an agreement to donate his $500 to charity. I neither won nor lost.

4. In April 2017, I bet a little over $200 on PredictIt that Marine Le Pen would not become President of France. Odds were 67/33 at the time. I was right. I won about $100.

5. In January 2018, I bet Andrew H $1000 (donated to charity) that Tom Chivers’ book about the rationalist community would not be “a hit job”. We are still awaiting the book to decide who won.

Comments are closed.