Over at his blog, Scott Alexander has launched a prediction contest. If you want to participate in it, you should probably go there before reading this post.

I have entered the contest, but I have no expectation of winning. At a certain point, prediction is guesswork, and it is a type of guesswork that I have found I am reliably1 bad at performing. But before that, prediction is largely “understanding the situation well”, and I am generally quite good at that.

There are two options for the contest. If you are entering “Full Mode”, I hope these words aid you in your quest. If you would prefer to enter in “Blind Mode”, stop reading now, close the tab, and don’t come back until you have submitted your entry.

Another note: There are 50 questions; I have skipped2 or merged a few of them to keep this post within reasonable length.

That disclaimer: if you want to enter the contest in “Blind Mode”, stop reading now! You must enter in “Full Mode” if you keep reading …

Ukraine Conflict: The war in Ukraine shows no sign of abating3. The war will go colder, but not quite as cold as the post-2014 status quo. Russia should be able to hold at the Dniepr River and the pre-2022 stalemate lines in the Donbas, and there will continue to be skirmishes across those lines. But Ukraine is unlikely to yield because of the attacks on civilian infrastructure beyond the warfront.

We assume wars are short these days, but the Thirty-Years War lasted over three decades. Something will happen, but not next year.

Putin as President, and Ukraine controls the city of Zaporizhzhia at 90%.
Ukraine controls the cities of Sevastopol or Luhansk at 10%
A cease-fire, and Kerch Bridge destroyed at 10%.

Nuclear War: I continue to be worried about the possibility of a catastrophe at the Zaporizhzhia nuclear plant. But I can worry about things that are unlikely.

1% for nuclear war (which everyone agrees is unlikely, and is too terrible to give higher odds)
10% for “nuclear weapon test” and “panic4 at Zaporizhzhia”.

Taiwan: Earlier this year, I said Vladimir Putin wasn’t stupid/insane enough to launch a failed invasion of Ukraine, and I was wrong. I am going to say the same thing about Xi Jinping now.

The US security blanket still protects Taiwan well enough. But it is fading. Within the next 20 years, the threat of invasion may become imminent. For 2023, it would be unthinkable folly. 10% for an invasion.

Join NATO: the accession of Sweden and Finland to NATO should finish in 2023. Hungary and Turkiye5 have threatened to block it, but they probably won't. 90% for accession.

Ali Khamenei: One of the many posts I did not write this year was one re-writing a BBC article about the Mahsa Amani protests to instead be about the George Floyd protests. It is quite unlikely they will amount to anything; the actuarial tables for an 83-year-old are more relevant. 15% for his death, resignation, etc.

Any Other War: With the multiple African coups and the (ending?) civil war in Ethiopia, it does not take much imagination to imagine a new war in Africa that would be more deadly than a largely-frozen conflict in Ukraine. Call it 50-50.

COVID in China: China’s zero-covid policy is ending6. Over the winter, they will have multiple large COVID outbreaks, and come out with some way of having voluntary quarantines most people ignore rather than the enforced quarantines they have had so far. 90% for “more than 25 million cases”.

2024 election: Unfortunately, these are “derivatives” - “what will the prediction markets say on 1 JAN 2024”. Presumably nobody will do undetected market manipulation to impact the resolution criteria. But “the consensus opinion will be wrong and will be proved that way in February 2024” is something that has to be considered.

I see Ron DeSantis being the odds-on favorite on the GOP side, and Kamala Harris leading the field in January (but not winning at the convention) on the Dem side.

20% of Joe Biden leading the field. If he runs, he will lead prediction markets until after a few primaries. But he won’t run. Of course, I also said “he won’t run” about Biden in 2019.
10% of Gavin Newsom leading the field.
10% of Donald Trump leading the field. We have reached the point where another Trump nomination is not “likely”, insh’allah.
80% of Ron DeSantis leading the field. Just because it looks like a sure thing doesn’t mean that it isn’t7 a sure thing.

Will Trump Tweet: Probably? He will get desperate by the end of the year, and “use Twitter” is the first ripcord to pull. 90% odds.

Biden Approval: The only way Biden will become popular is if he doesn’t run, gas prices stay low, Congress is mundane, and nothing catastrophic happens. All of those are likely. 65% (aka .9^4) odds.

Trump Indictment: Not going to happen. 10% odds.

UK Politics: I haven’t been paying attention ever since the Liz Truss drama ended, but I thought the plan was for an early election in 2024 rather than 2023. Maybe not? Call it 30% for an early election, and 85% that Rishi Sunak remains PM (he might win8 the election).

Twitter: I would call “Recuse”, but that doesn’t seem to be allowed.

95% that Musk remains owner of Twitter. Nobody is buying it from him, and a court-ordered transfer in bankruptcy wouldn’t happen within a one-year timeframe.
As a private company, I don’t think Twitter will have to report its income or mDAU. 50% for both as failed markets.

Inflation: The impact of the 2021 stimulus (and, more generally, COVID) on the economy is fading. 4 percent inflation seems a reasonable pick-em number, so we go 50%.

Crypto-currency: This balloon is still deflating. My long-term price projection for Bitcoin is $7000-10000. While the ability for suckers to inflate the price should have ended, it is possibly my prediction is wrong. 10% for it being above $30k.

As far as Tether … I have concluded that the real story is that Tether isn’t insolvent, and all the billions they report are from various rich people who thought crypto was the next big thing. Even so, it’s 50-50 that it collapses.

Elon Musk still loves crypto, and will probably (90%) do something at Twitter in that sense.

Starship: Perhaps the lasting legacy of the Musk-Twitter episode will be that “rocket science” isn’t actually that hard, at least once you have precision manufacturing and computers. 90% chance of success. The specific condition is one I thought was already true, but … meh.

COVID: With the vaccine and/or having gotten COVID before, COVID is not much worse than the flu. But that means that, probably, it will kill as many people next year as this year, and it will evade vaccines. 90% for “at least 50% of casualties” and 80% for “new variant evades vaccines”.

Artificial Intelligence: DALL-E for videos is coming. Either in 2023 or 2024. 80% for 2023.

As far as “GPT-4”: this is a marketing question, rather than a tech question. But, the inability to have a tech breakthrough could guide the marketing decision. 75% that something called GPT-4 is released.

The same odds (75%) for image compositionality - it doesn’t seem that hard9, but whether or not somebody actually does it (and gets Scott’s attention) is a different question.

And, as far as “AI wins a programming competition”? Call it 80%. 90% that it is capable of doing so by August, and 90% that somebody triggers that news cycle.

and One More Thing: I don’t do comments here normally, but to avoid spoilers in Scott’s thread, they are enabled here for now.

1

But, not so reliably bad that “doing the opposite” works for me in any way.

2

As a reminder, the Newslettr will not discuss markets about assassinations or Supreme Court decisions. Also, not being located in San Francisco, I have no way of knowing if “AR headset competitor” or “self-driving car expansion” will actually happen in 2023.

3

There is talk of … a lot of things in the press. For example, claims that “Russia is planning another advance towards Kiev through Belarussian territory” have been seen in the press, but that is almost certainly propaganda.

4

I would put the odds of a true catastrophe as 1% rather than 10%. But there needn’t be a catastrophe for a panic to trigger an evacuation, which is what is being asked.

5

Turkiye did change its name with the United Nations this year. People haven’t quite picked up on it. Regardless of any other concerns, avoiding the confusion with the bird is reason enough to adopt the new spelling. The umlaut hasn’t taken off, so we won’t bother. And don’t even mention the dotless-I.

6

With the end of zero-covid, there will be a flood of surplus deaths among the elderly. Presumably we are at the point where the Chinese public will be less upset about the deaths than the possibility of another year of lockdown. If they are, we can say the zero-covid policy “worked” even though it is currently being abandoned as a failure.

7

Did you read that as “just because it looks like a sure thing doesn’t mean it is a sure thing”? Sometimes you can over-correct. Sometimes, it looks like a sure thing, and it is.

8

Unless Sunak dies or has an unthinkable scandal, he will remain PM throughout 2023 if there is no election.

9

There are two issues here. The first is the lack of training data, which various startups such as https://infinity.ai/ can already solve. The second is the fractal (self-similar) nature of certain types of neural networks, which … probably won’t be a problem.

the Tisatsar Newslettr

The 2023 Prediction Dispatch