Brier score calculator
A score on its own means nothing — this one comes with the number it has to beat.
Result
- Brier scoreLower is better; 0.25 is a coin flip
- 0.1275
- Base-rate referenceWhat forecasting the base rate every time would have scored
- 0.2500
- Skill vs. that referenceYou added information
- 49.0 %
- Forecasts scored
- 6
- Events that happened
- 50 %
- BiasAverage forecast above the base rate
- +5.0 %
A Brier score without its base-rate reference is half a fact: 0.20 is excellent on coin-flip events and poor on near-certainties.
What the number means
The Brier score is the average squared distance between what you said and what happened. Say 90 % and it happens, you took a 0.01 hit; say 90 % and it does not, you took a 0.81 hit. Perfect foresight scores 0, and calling everything a coin flip scores 0.25.
The squaring is the point. It is what makes confident errors expensive and stops the score from rewarding hedging everything to 50 %.
Why the raw score is close to meaningless alone
A Brier score of 0.20 sounds mediocre and could be excellent — or terrible. If the events you forecast happened about half the time, 0.20 is genuinely good. If 95 % of them happened, then simply saying “95 %” every single time would have scored 0.0475, and your 0.20 is far worse than knowing nothing but the base rate.
That is why this calculator always shows the base-rate reference next to the score, and the skill number derived from it. A positive skill means you beat the base rate; a negative one means you would have done better with no information at all. Any accuracy claim that quotes a Brier score without its reference is telling you half a fact.
Bias is a separate question from accuracy
The bias line compares your average forecast against the share of events that actually happened. A large positive bias means you are systematically too confident that things will happen; a negative one means the reverse. You can be well calibrated on average and still have a poor Brier score, and you can have a decent Brier score while being consistently biased in one direction — they measure different failures.
Common questions
What is a good Brier score?
It depends entirely on the base rate of the events you scored. 0.25 is what you get for calling everything 50/50, and anything below that beats a coin flip — but if the events almost always happened, the bar is far lower. Compare against the base-rate reference, not against a fixed threshold.
How is the Brier score calculated?
Take each forecast, subtract the outcome (1 if it happened, 0 if not), square the difference, and average across all forecasts. Lower is better, 0 is perfect.
What is the skill score?
One minus your score divided by the score you would have got by forecasting the base rate every time. Positive means you added information; zero means you matched the base rate; negative means you did worse than knowing nothing but how often these events happen.
Read more
These calculators do arithmetic on numbers you supply. They are not advice, and nothing here is a recommendation to take a position.