Log Loss
Log loss is a proper scoring rule that strongly penalizes extreme probabilities when outcomes go the other way.
Definition
Log loss is a proper scoring rule used to evaluate probabilistic forecasts. It penalizes confident wrong predictions more than Brier score, especially near 0 or 1.
Why it matters
Log loss discourages overconfidence. It is particularly useful when you want a strong penalty for assigning extreme probabilities without sufficient evidence.
Common pitfalls
Single event focus: Like other scoring rules, log loss is most meaningful over many forecasts.
Comparing across different event sets: Difficulty differences matter.
Learn more
For scoring workflows and examples, see BrierScore.com.