What is this about?
This newsletter covers topics on forecasting, with a focus on human judgemental forecasting. This is where humans try to come up with models of the world, and then make their models fight in platforms such as Metaculus, Good Judgement Open, CSET-foretell, various prediction markets, etc.
I also cover broader topics in forecasting, such as those in machine learning, weather forecasting, various crypto prediction platforms, presidential election forecasting, etc. The newsletter goes out monthly, and generally contains the following sections:
Highlights.
Prediction Markets & Forecasting Platforms.
In The News: Forecasting in mainstream media.
Blog Posts.
(optionally: Hard To Categorize: Various miscellany)
Long Content: Where I summarize a paper or present content which I expect to be relevant or interesting in the long-term.
(optionally): Negative Examples: Overconfident predictors, easily avoidable failures; Corrections.
I also maintain metaforecast.org, a search engine for probabilities, and this database of prediction markets.
When should you subscribe to this newsletter?
If you’re a forecaster interested in keeping up with opportunities in the area, such as new prediction markets. If this is the case, you should also subscribe to newsletters from the individual forecasting platforms.
If you’re interested in forecasting because it is adjacent to your field. For example, I keep an eye for incentive failures in forecasting systems, and aligning predictive systems might serve as a toy problem before aligning more powerful machine learning systems. There are also some interesting combinations of forecasting + a random field.
If you’re interested in getting better models of the world about diverse areas then you might want to also consider spending a couple of hours a week becoming a hobbyist forecaster; I’d expect it to have better returns on investment than only reading about it. Good platforms for this purpose are Metaculus and Good Judgment Open.
Acknowledgements
As of early 2023, the organization that I otherwise work for, the Quantified Uncertainty Research Institute, is partially subsidizing this newsletter. Thanks also to the paying subscribers, who give me a signal that this newsletter is producing some value.
