Prior (Poisson process): Russia has invaded three times in the past 20 years (Georgia 2008, Ukraine 2014, Ukraine 2022). Call that \(\frac{3 + 1 invasions}{22 + 1 year} \approx \) 20%/year. Remove a 1/6th because we are 1/6 of the way through the year to get \(\approx\) 17% In the past there's been 3-month+ lead times where Russia's amassed troops around another country. If we say Russia does not have such a troop build-up now and a 3 month lead-time is required 80% of the time, we're down to 13% chance of an invasion this year. But perhaps Russia _...
Wikipedia lists 31 events since 2000 (28 since 2002), 9 of which had fatalities. Model as a poisson process, and our probability density should be an exponential distribution with a mean of ~8 months and a median of \(\ln(2) 8 months \approx 5.5 months\) . You can get a fairly good approximation of that with the square slider at Now and circle slider at 7 months from now. Irrelevant side note: the January 28 bridge collapse got a lot of play, and was one of the bigger news stories of the week. But it had no deaths, and bridge collapses happen ~ every...
@(optimaloption) I really like your base rating on prob. of spillovers. Conditioning on Europe worries me a bit: if you have a formal dataset and conditioning on Europe makes a difference, I'd want to drill into that. If you're conditioning on "wars I can think of which that seem relevant", then conditioning on Europe seems more reasonable, with the caveat that wars you can think of might be biased towards big ones. A reason to adjust down from the Europe-spillover average: most countries in Europe have alliances. Ukraine does not. If I told you "A ...

I thought the same thing. Either axions or (arguably) sterile neutrinos. See Weakly Interacting Sub-eV Particles