Ongoing model residual analysis
The no-brainer use of the world’s only microprediction oracle. See Chapter 9 of the book.
Steps:
- Publish model residuals (see publishing docs)
(I guess if you are lazy you could send me a really long skinny CSV, though live is the quintessential use case)
Why?
Someone, somewhere might have deployed an algorithm that finds signal in your noise. Or they might in the future.
What happens:
- You’ll add to the list of streams.
- Algorithms (Python, Julia, R mostly) fight to predict your residuals (distributionally)
- More arrive all the time. See github/microprediction for an explanation of how new methods trickle in from the Python ecosystem. But anyone can deploy algorithms using R, Julia or whatever as well.
- The ongoing battles produce beautiful community cumulative distribution functions, such as the CDF representing the 1-hour ahead forecasts of the logarithm of META price changes.
- You may glean quite a lot from the outcome of that fight, especially if the winners aren’t “null hypothesis” algorithms.
Optional steps:
-
(optional) Submit a distributional prediction of your own residuals (see prediction docs)
-
Check with compliance.
-
Use set_repository() method to point people to a page containing information.
-
Add to the rewards for good prediction determined daily.
Why is Intech allowing other funds to use this?
Read the book.
-+-
Documentation map
View source