What is Brier score?

Mean squared error of probability forecasts; 0.25 is the score of a coin flip.

For each call the Brier score is (probability − outcome)², averaged over calls. Lower is better, and 0.25 is what you get by always saying 50%.

It rewards being right and punishes being confidently wrong more than being unsure. The direction gate compares against that 0.25 line directly.

Model gateOur model
The rule that hides a forecast when the model could not beat a coin flip on it.
CalibrationOur model
How closely stated confidence matches the real hit rate of calls made at that confidence.
Walk-forward backtestOur model
Testing the model only on data after each training window, the way it would have been used.

See it on a coin page