Forecasting newsletter: US elections
Forecasting newsletter: US elections
Huge US election volumes, Robinhood added presidential prediction markets, Richard Ngo on why he is not a Bayesian.
Highlights
Huge US election volumes
Robinhood added presidential prediction markets.
Richard Ngo on why he is not a Bayesian
Index
Current takeaways from the 2024 US election
Platforms
Polymarket
Kalshi
Metaculus
Manifold
Other platforms
Projects
Research
Odds and ends
Current takeaways from the 2024 US election
Polymarket beat legacy institutions at processing information, in real time and in general. It was just much faster at calling states, and more confident earlier on the correct outcome.
The OG (a) prediction markets community, the community which has been betting on politics and increasing their bankroll since PredictIt, was on the wrong side of 50%—1 (a), 2 (a), 3 (a), 4 (a), 5 (a). It was the democratic, open-to-all (a) nature of it, the Frenchman (a) who (a) was convinced that mainstream polls were pretty tortured and bet ~$45M, what moved Polymarket to the right side of 50/50.
Polls seem like a garbage in garbage out kind of situation these days. How do you get a representative sample? The answer is maybe that you don’t.
Polymarket will live. They were useful (a) to the Trump campaign, which has a much warmer perspective on crypto. The federal government isn’t going to prosecute them, nor bettors. Regulatory agencies, like the CFTC and the SEC, which have taken such a prominent role in recent editions of this newsletter, don’t really matter now, as they will be aligned with financial innovation rather than opposed to it.
NYT/Siena really fucked up with their last poll (a) and the coverage of it. So did Ann Selzer (a). Some prediction market bettors might have thought that you could do the bounded distrust (a), but in hindsight it turns out that you can’t. Looking back, to the extent you trust these institutions, they can ratchet their deceptiveness (a) (from misleading headlines, incomplete stories, incomplete quotes out of context, not reporting on important stories, etc.) for clicks and hopium, to shape the information landscape for a managerial class that… will no longer be in power in America.
Elon Musk and Peter Thiel look like geniuses. In contrast Dustin Moskovitz couldn’t get SB 1047 (a) passed despite being the second largest (a) contributor to Democrats, and has maybe been in-hindsight suboptimally positioning his philanthropic bets by reportedly not funding any slightly right of center work (a). These are also bets that matter!
These are a bit tentative, and I’ll have better takes after I marinate on election results a bit more. In the meantime I welcome nitpicks.
Platforms
Polymarket
The election captured an influx of new interest in prediction markets. There is $550M (a) locked in crypto prediction contracts, and ~94% of it is on Polymarket. As a side effect, Polymarket reached the top on various charts (a) of most popular apps (a). That’s it, that’s the high level story. Now the question is whether Polymarket will be able to sustain the attention of these new followers, and whether it will retain its core idealistic vision of providing better information or whether it will descend into a common-denominator sportsbook.
There is some interest on the Frenchman who bet millions that Trump would win, and visibly shifted the odds. A writeup about him here (a), and a video interview with him here (a). Elon Musk retweeted a tweet (a) about a Polymarket market to his millions of followers.
Polymarket’s success has inspired many copycats, normally with unclear stories for why they are better. In a funny twist, a blockchain consultancy is now offering a “Polymarket clone script” (a); Polymarket in a box.
As of early October, only 12.7% of wallets in Polymarket (a) had reached a profit. This might be a slight under-count if a small fraction of users is doing arbitrage with other platforms, or if users have more than one wallet, but it’s still pretty low.
Kalshi
Kalshi started paying about 4% interest on positions (a), added more elections markets (a), and USDC (a) deposits (USDC is a crypto stablecoin). Still they didn’t beat Polymarket.
Metaculus
Metaculus is now open (a) source (a), and has updated its API (a). Some of their research finds that top forecasters were much better than bots (a) with $30K was on the line (for the bot creators), and they started a small forecasting tournament (a) with a $2K pool in collaboration with GiveWell.
Manifold
Manifold continues to add monetary “sweepstakes” markets (a), but they had a relatively meager $350K (a) locked up in their money markets
Manifold’s newsletter looked at Alan Lichtman’s 13 keys to the White House (a). There was a funny moment when Lichtman tweeted (a) that Nate Silver “doesn’t have the faintest idea how to turn the keys”. On top of that, his prediction was wrong (a).
Other platforms
Robinhood added (a) presidential prediction markets.
Iowa Electronic Markets are somehow still alive and producing papers (a)
Good Judgment Open sat at 61% dems (a) as on the 5th of November.
Projects
Adjacent news, a small outlet which integrates prediction markets and news, and offers some good data API (a), had their launch (a), where they explain what they’re about and what they offer.
At Sentinel, we monitored the Israel/Iran situation, got a $130K grant (a) from the Survival and Flourishing Fund, improved the tech (a) stack (a) we use to prospect for news, continued posting our minutes (a), and put out a post looking at precursors of existential risk (a): more specifically, at the chance some items had of killing more than 1M people in any one of the next 10 years.
The RAND Forecasting Initiative seems a bit more active. Two posts on their blog talk about how some forecasting questions feed into their analysis on China sanctions (a) or Russian propaganda in Germany (a) caught my attention.
Sage Forecasting (a) is looking for a member of the technical staff to build forecasting tools. Salary $100-$150K.
Nathan Young open-sourced (a) the code for findingconsensus.ai (a). If you want to present people’s perspectives on a topic, the github repository might come in handy.
David Glidden reports that the DC forecasting meetup was well-attended, with 26 participants. Their next meeting will be in December (a). They received $2.5K from Manifold as a sponsor. Hopefully after the Trump win, they will be in a much better position to push for their policy priorities to a more receptive administration.
Research
An old paper (a) by Robin Hanson points out that a manipulator can increase accuracy by increasing the returns to informed trading.
Richard Ngo has an article exploring why he is not a Bayesian (a) (also on LessWrong (a)). He boldly puts forward the view that a) we should assign truth-values that are intermediate between true and false, and b) we should reason in terms of models rather than propositions.
To me, fuzzy truth values are very interesting because I see how they can save a lot of effort to a computationally bounded agent. For instance, forecasting on a precise specification of “catastrophic risk” can get very bogged down on what exactly is a catastrophic risk. But you could also accept it’s a fuzzy category, and forecast much faster. And if you are a small creature wanting to make sense of a big world, you can find yourself in a situation where you either take that kind of shortcut or you are just not able to make sense of enough of the world.
Manifold (a), Bloomberg (a) and the Swift Institute (a) had conditional forecasts around the US election in some form or other.
Odds and ends
Rootclaim snared (a) a conspiracy theorist into escrowing $1M to wagered on whether a debate presented to a mutually agreed set of judges will convince them on that the COVID mRNA vaccines save more lives than they killed.
An op-ed on Time magazine (a) attacked prediction markets.
The UK’s Flood Forecasting Centre (a) looks at its work 2023-2024. They seem to have better processes than Spain’s AEMET, which wasn’t able to warn people of flooding in Valencia.
Here (a) is an interview with Jaime Sevilla, the director of Epoch, an org which gathers data on AI trends.
Jane Street has a forecasting competition (a) to predict financial markets from real world data. Deadline January 6th, 2025. If you are a very particular type of nerd, this could be your gateway to an extremely well paid job. Includes $105K in monetary prices.
Russia has pursued a cold policy of self-interest. We could have wished that the Russian Armies should be standing on their present line as the friends and allies of Poland, instead of as invaders. But that the Russian Armies should stand on this line was clearly necessary for the safety of Russia against the Nazi menace.
At any rate the line is there, and an Eastern Front has been created which Nazi Germany does not dare assail. When Herr von Ribbentrop was summoned to Moscow last week, it was to learn the fact, and to accept the fact, that the Nazi designs upon the Baltic states and upon the Ukraine must come to a dead stop.
I cannot forecast to you the action of Russia. It is a middle wrapped in mystery inside an enigma; but perhaps there is a key. That key is Russian national interest. It cannot be in accordance with the interest or safety of Russia that Germany should plant itself upon the shores of the Black Sea, or that it should overrun the Balkan states and subjugate the Slavonic peoples of Southeastern Europe. That would be contrary to the historic life interests of Russia.
But here these interests of Russia fall into the same channel as the interests of Britain and France. None of these three powers can afford to see Rumania, Yugoslavia, Bulgaria, and above all Turkey, put under the German heel.
Through the fog of confusion and uncertainty we may discern quite plainly the community of interests which exists between England, France and Russia to prevent the Nazis from carrying the flames of war into the Balkans and Turkey. Thus, at some risk of being proved wrong by events, I will proclaim tonight my conviction that the second great fact of the first month of the war is that Hitler, and all that Hitler stands for, have been and are being warned off the East and Southeast of Europe.
— Winston Churchill, First Lord of the Admiralty, Excerpt from Radio address from London (a), October 1, 1939
This newsletter is sponsored by the Open Philanthropy Foundation. Thanks to Mick Bransfield for item suggestions.