Welcome to Nonrival, the newsletter where readers make predictions.
How it works
- On Sundays, read the newsletter and make a forecast by clicking a link at the bottom.
- On Wednesdays, see what other readers predicted and how your forecast compares.
- Over time, you’ll get scores based on how accurate your forecasts are.
In this issue
- Results from Sunday's forecast on the war in Ukraine
- Breaking down the risk of nuclear war
Thanks for forecasting. Send feedback to email@example.com.
No end in sight in Ukraine
How likely is a peace deal between Ukraine and Russia?
- Nonrival readers' average forecast was a 19% chance of a deal by August 19 (the median is just 10%).
- That mirrors Good Judgment Open, which also gives a 20% chance of a deal before August 19.
- On Manifold, the market gives a 22% chance of a "lasting cease-fire" by end of 2023.
Unfortunately, it seems highly likely that this war will grind on. Political scientist Chris Blattman, whom I interviewed about the year-mark in Ukraine, has noted that most wars are short. But as one astute reader noted in their forecast rationale, of wars that last a year, most stretch on over more than a decade.
How your forecast compares
- You said there was a [022623_FINAL GOES HERE]% chance of an agreement to end the current conflict in Ukraine by August 19.
- You predicted that the average of readers' forecasts would be [022623_CROWD GOES HERE]%. The actual average was 19%. You were closer than [022623_CROWD_RANK GOES HERE]% of readers.
What other readers said
Breaking down the risk of nuclear war
How likely is it that Putin will use nuclear weapons, and what might make him more and less likely to do so? Those can seem like impossible questions to answer, but they're incredibly important. Last year, several forecasting platforms and groups published estimates on the likelihood of Putin using nukes.
Peter Scoblic, a nuclear expert and a senior fellow at the international security program at New America, took a different approach: He tried to break down that giant question into several more manageable ones. The result was the "Red Lines in Ukraine" project on the forecasting platform Metaculus, where Scoblic is director of nuclear risk.
“My initial interest was in developing a better sense of what effect US policies would have on the threat of nuclear escalation,” Scoblic told Nonrival. In addition to having forecasters tackle the nuclear question directly, the Red Lines project also asks things like: Will the US provide ATACMS (a tactical missile system) to Ukraine? And: Will Ukraine control Crimea?
These questions allow for conditional forecasting: If X happens, how likely is Y? For example, Metaculus forecasters currently put a 2% chance on Russia detonating a nuclear weapon in Ukraine this year. But if Ukraine controls a significant chunk of Crimea, forecasters put the risk of Russia using a nuke at 10%. And conditional on the US providing ATACMS, the risk of Russia using a nuke is estimated to be 5%.
“The ultimate hope would be that we can estimate not only the likelihood of nuclear use but we can provide valuable information to policymakers about the potential implications of policy decisions they make," says Scoblic.
But these conditional forecasts are tricky to interpret. They're not reflecting cause-and-effect so much as (hopefully) serving as an early warning system. The US sending ATACMS to Ukraine doesn't necessarily itself increase the risk of Putin using a nuke from 2% to 5%—that remains impossible to measure. But, to the extent one thinks crowd forecasting platforms can make accurate predictions, the decision to send ATACMS would suggest that the situation has changed such that the nuclear risk has risen. The hope is for forecasts like this to be one input into the policymaking process.
Metaculus forecasters' overall prediction of Russia using a nuke has, happily, decreased since the question launched last fall, from the neighborhood of 10% to 2%. "[Forecasters] believe that the use of nuclear weapons in Ukraine is less likely," says Scoblic. Whether or not those numbers are precisely meaningful for such an unpredictable topic, "at the very least it’s a metric of concern.”
Scoblic cautions, though, that that number might rise again as events on the ground change. And that's the point of the Red Lines project: to serve as a warning system for anyone monitoring nuclear risk.
- Permission slip culture is hurting America
- Sun Belt cities are dominating the post-pandemic recovery
- Can you predict NBA All-Stars five years ahead of time?
- Ben Evans' massive annual slide deck on tech trends
- What are the odds H5N1 is worse than COVID?
- Can social scientists forecast societal change? (The more they rely on data the better they do)
- Maybe the yield curve's inversion doesn't predict recessions after all
- Will Google search include a chatbot by end of June? Manifold says 41% chance
- Disney's Bob Iger seems at least somewhat open to selling Hulu