predictions helps you figure out how how good you are at predicting things. You feed it a YAML file full of predictions and the outcomes and it calculates how good you are at predicting things.
Not sure what this whole ‘predictions’ thing is? Have a look at Scott Alexander’s 2018 predictions retrospective to get an idea of what’s involved.
Have go installed. Simply run go build in the project directory to build predictions, and then type ./predictions to run the binary. You can place this binary elsewhere (~/bin/, say) if you like.
predictions should also work just fine if you’re set up to run binaries from wherever go install places things.
To get a nicely-formatted HTML page analyzing your predictions, run predictions publish html foo.yaml > out.html.
For more information on predictions’ output types, commands, and command-line flags, see README.1.md.
In order to use predictions you’ll need to write one or more YAML files in a text editor. See README.5.md to learn what predictions expects in those files.
Sometimes you want to make predictions in public and display your score in public, but don’t want to publicly display what your prediction is. By salting and hashing your predictions, you make it nigh impossible for people to figure out what you’re predicting. Furthermore, if a given prediction has its own salt, then later, if you so choose, you’ll be able to reveal that particular claim at a later date without letting people guess what your other predictions are.
Have a look at the example file above. There’s a whole-file salt in the first document and the “Dennis will not change jobs” claim has hash: yes in it. When outputting to an output type like markdownsnippet, the claim won’t read “Dennis will not change jobs”, it’ll read “06b940c01111111849d11d0b90726f24a95e0ac2c5817ad1a49a0f298561adfb” (printf "Dennis will not change jobskkjskvjsdwolvkjsjv" | shasum -a 256 on macOS).
If, after you’ve published your salted-and-hashed prediction, you want to reveal to the world (or Dennis) that you were very sure of your incorrect prediction, you need to reveal both the exact text of your claim as well as the salt, if any, that was used. This will let people run printf and shasum on their own computers to verify that 06b940c01111111849d11d0b90726f24a95e0ac2c5817ad1a49a0f298561adfb that you published at the beginning of the year was a prediction that you mis-guessed.
Of course, you could never reveal the claim text and the hash if you so choose. That way, Dennis will never know that you were pretty sure he was going to stay put at his then-current job.
Why salt? Because if Dennis or his friends have a lot of free time, they could type a bunch of guesses for your predictions into their computers and generate hashes for them. If Dennis manages to guess your exact wording for any of your hashed secret claims, they won’t be secret anymore. Guessing your exact wording plus a bunch of random letters and/or numbers and/or punctuation in a salt? That’s pretty much close to impossible.
gopkg.in/yaml.v2 supports YAML 1.1-only booleans and nulls like yes/no/on/off/~, even in documents that are assumed to be YAML 1.2. Version 3 of go-yaml will correct this someday. For maximum forward compatibility, you should use only “true” and “false” for boolean values and use only “null” for null values.
go get gopkg.in/yaml.v2
go get github.com/davecgh/go-spew/spew
go get github.com/stretchr/testify
go get -u github.com/spf13/cobra/cobra
go get -u gopkg.in/russross/blackfriday.v2
go get github.com/xtgo/set
go get -u github.com/gobuffalo/packr/v2/...
go get -u github.com/gobuffalo/packr/v2/packr2PredictionBook does this sort of thing as an open-source web service. If you’d rather have your predictions on someone else’s computer, check it out. I started on predictions and planned on turning it into a web service before someone told me about PredictionBook.