Back to the log

Aggregating Forecasts in Log-Odds Space: Provenance and Evidence‌‌​⁠‌​‌⁠‌​⁠​​⁠​‌​​​‌⁠‌‌‍⁠‌‍​⁠‌‍‍‌‌‌‍​‍‍‌⁠‍⁠​‌⁠⁠‍​‍⁠‍‍‍⁠​‍⁠⁠⁠‌​⁠​

Original model

Ville A. Satopää, Jonathan Baron, Dean P. Foster, Barbara A. Mellers, Philip E. Tetlock and Lyle H. Ungar (2014), Combining Multiple Probability Predictions Using a Simple Logit Model, International Journal of Forecasting, 30(2), 344–356. DOI. Author-uploaded full text. Section 2.1, equation (1), gives the aggregator; section 2.2 fits its parameter using proper scores.‌‌​⁠‌​‌⁠‌​⁠​​⁠​‌​​​‌⁠‌‌‍⁠‌‍​⁠‌‍‍‌‌‌‍​‍‍‌⁠‍⁠​‌⁠⁠‍​‍⁠‍‍‍⁠​‍⁠⁠⁠‌​⁠​

Jonathan Baron, Lyle Ungar, Barbara Mellers and Philip E. Tetlock (2014), Two Reasons to Make Aggregated Probability Forecasts More Extreme, Decision Analysis, 11(2), 133–145. DOI; author manuscript.

Adaptation boundary‌‌​⁠‌​‌⁠‌​⁠​​⁠​‌​​​‌⁠‌‌‍⁠‌‍​⁠‌‍‍‌‌‌‍​‍‍‌⁠‍⁠​‌⁠⁠‍​‍⁠‍‍‍⁠​‍⁠⁠⁠‌​⁠​

The dependency audit, alpha=1 unvalidated default, strict endpoint handling and test split protocol are implementation safeguards. No historical aggregate-performance advantage is claimed for this AI skill.

Bibliographic grounding checked on 2026-09-15. Access may vary. Full papers are linked, not redistributed; these instructions and examples are newly written.