Apparently some major changes on Range and Forecasting Accuracy caused it to be re-submitted.
Quick summary:
Code was rewritten in Python
Moved Results section to the top, and rewrote it, it should be easier to understand
Expanded the illustrative example
Added logistic/exponential curve-fits to all section, which enable to extrapolate Brier scores to longer-range questions (under certain assumptions)
This allows to estimate how far into the future we can see before our forecasts become uniformly random
Unfortunately, there is no single nice number I can give for our predictive horizon (...yet)
P-values in some places!
Current theme: default
Less Wrong (text)
Less Wrong (link)
Apparently some major changes on Range and Forecasting Accuracy caused it to be re-submitted.
Quick summary:
Code was rewritten in Python
Moved Results section to the top, and rewrote it, it should be easier to understand
Expanded the illustrative example
Added logistic/exponential curve-fits to all section, which enable to extrapolate Brier scores to longer-range questions (under certain assumptions)
This allows to estimate how far into the future we can see before our forecasts become uniformly random
Unfortunately, there is no single nice number I can give for our predictive horizon (...yet)
P-values in some places!