![](/_next/image?url=https%3A%2F%2Ffirebasestorage.googleapis.com%2Fv0%2Fb%2Fmantic-markets.appspot.com%2Fo%2Fdream%252FwjeLIpQpAd.png%3Falt%3Dmedia%26token%3D56804e36-64e3-4ebe-bde0-c90c58aef700&w=3840&q=75)
This resolves YES if I find convincing evidence that forecasters have improved performance when they use Fermi Estimation for calculating their probabilities. To be clear, I'm not looking for Fermi Estimation as a means of discovering some underlying fact: "What's the chance of a second Emu War in Australia in 2025? Well first I need to estimate the number of Emus..." but rather directly modelling the final probability as a sort of Fermi estimate equation.
This evidence may already exist today. If so, all that's sufficient to resolve this market is to link that evidence here. However, if positive evidence comes in, I will wait for contradicting evidence before resolving and weigh the balance of the evidence. I'm not looking for one-off anecdotes either -- I know Nate Silver goofed on Donald Trump using something like this method, for instance. Please link actual studies or at least something with multiple repeated instances that we can scrutinize and quantify.
To be clear: the spirit of this market is to try to gather some empirical evidence for better understanding the so-called "multi-stage fallacy" which takes a dim view of using Fermi Estimation to arrive at probabilities -- the argument is that you can add in as many stages as you want to artificially lower the percentage to nothing, because any number of probabilities multiplied together shrink fast. Also concerns about correlation/dependence between probabilities.
My hunch is we should have enough evidence to arrive at a better gut understanding of this that outperforms armchair theorizing. Either it works or it doesn't, and there's already plenty of incentive for people to have tried it. And maybe there's better or worse ways to use it, which seems important.
Curious minds want to know.