I build and evaluate AI agents that forecast and do research at FutureSearch. Before that, I did forecasting research at Metaculus and a PhD in probability theory.

For further details, see my CV or contact me via email.

• petermuehlbacher
• peter-muehlbacher

Mathematics

• Kelly Betting

  A collection of interesting properties of and arguments for and against the Kelly criterion.

• Forecasting Martingales

  Identifying shortcomings of evolutions of forecasts and improving upon them.

• High-dimensional Statistics

  Notes about where intuition for high-dimensional statistics fails.

Forecasting

• Aggregating Forecasts

  Thoughts on different methods to combine different forecasts.

• Post-mortems

  Here I try to analyse where some of my worst forecasts went wrong.

• Useful Priors

  A collection of priors useful for forecasting.

Programming

• Eldrow

  Eldrow is a puzzle about how 'unlucky' one can get when playing Wordle on hard-mode.

• Diffusion Maps

  A brief motivation and implementation in Python of dimensionality reduction via the diffusion maps algorithm.

• Max-entropy Distributions

  Notes on finding the max-entropy distribution satisfying some constraints. Comes with discussions on applications to forecasting & finance and a Python implementation.

Languages

• Anki for Chinese

  A quick primer on how to use Anki (a spaced repetition software) in general and how to set it up for learning Chinese in particular.

• (Pitfalls of) Teaching German

  Some notes of my course for beginners and a collection of grievances about the language.