Thanks for this post.
Hey! I think this is cool. May I suggest “How many people in Kings County, NY, will be confirmed to have died from COVID-19 during September?” as a question?
I have a forecasting newsletter with ~150 subscribers; I’ll make sure to mention this post when it gets sent at the end of this month.
Foretold has a public API; requests can be made to it from anything that sends requests. This would require some work.
Personally, I’ve used Foretold, Google Sheets, CSVs, an R script, and my own bash script (PredictResolveTally) (which writes to a csv.).
Personally, I like my own setup best (it does work at the 5 second level), but I think you’d be better off just using a CSV, and then analyzing your results every so often with the programming language of your choice. For the analysis part, this is a Python library I’m looking forward to using.
Browsing Wikipedia, a similar effort was the 1985 book Tools for thought, (available here), though I haven’t read it.
As an heavy predictit user
As an heavy predictit user
Could you say more about this? What is your ranking in PredictIt / what is your track record? In particular, GJOpen, for example, doesn’t expect Trump to win
This might be of interest: https://www.kill-the-newsletter.com/.
I’ll let you know. 30%-ish.
Now there is a pure version of what you were looking for! Corona Information Markets
Consider the futures of humanity which I would, upon reflection, endorse as among the best of utopias, and consider the simplest Turing Machines which encode them. If you apply (some function which turns their states after n steps into a real number and concatenate them), would the output of such calculation belong to (this randomly chosen half of the real numbers)?
I’m sure this can be worded more carefully, but right now this may force the oracle to simulate all the futures of humanity which I would consider to be among the best of utopias.
Machiavelli’s The Prince, and various other texts.
predict.replicationmarkets.com will be looking into predicting viable treatments in the future, according to their last newsletter, but right now it’s mostly hypothetical. Rewards are monetary.
pandemic.metaculus.com has several prediction tournaments with monetary prizes. Example.
foretold.io has two active covid communities. No monetary prizes, but predictions are used by epidemicforecasting.org
gjopen.com has plenty of covid questions, but the only reward is reputational.
Augur.net may have some markets, but they’re transitioning to v2.0 and markets are temporarily unavailable.
Can you give some more intuitions as to why allowing finite support is among your criteria?
I can imagine a definition which, lacking this criterion, is still useful, and requiring to have infinite support might be a useful reminder that 0 and 1 are not probabilit(y densities). Further, whereas requiring infinite support might risk analyzing absurd outcomes, it may also allow us to consider, and thus reach maximally great futures.
Reality is inherently bounded—I can confidently assert that there is no possible risk today that would endanger a trillion lives, because I am confident the number of people on the planet is well below that.
Consider that the number of animal lives is probably greater than one trillion, and you didn’t specify *human* lives. You could also consider future lives, or abstruse moral realism theories. Your definition of personhood (moral personhood?) could change. Having finite support considered harmful (?).
Here is another point by @jacobjacob, which I’m copying here in order for it not to be lost in the mists of time:
Though just realised this has some problems if you expected predictors to be better than the evaluators: e.g. they’re like “one the event happens everjacobyone will see I was right, but up until then no one will believe me, so I’ll just lose points by predicting against the evaluators” (edited)
Maybe in that case you could eventually also score the evaluators based on the final outcome… or kind of re-compensate people who were wronged the first time…
Another point in favor of such a set-up would be that aspiring superforecasters get much, much more information when they see ~[the prediction of a superforecaster would have made having their information]; a point vs a distribution. I’d expect that this means that market participants would get better, faster.