Brier for beginners: one number, one formula, no flattery
Quick note ⏰ 2 min read ÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂ÷ Planted Nov 5, 2026 Open Resolve by Dec 1, 2026 ÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂ÷ by LAI
A Brier score is the simplest honest way to grade a forecast. It answers one question: how far were your numbers from what actually happened? Nothing more.
The whole formula
Score = (your number, as a decimal, minus the outcome) squared. The outcome is 1 if the thing happened and 0 if it did not. If you said 70 percent and you were right, the score is (0.70 minus 1) squared, which is 0.09. If you said 70 percent and you were wrong, it is (0.70 minus 0) squared, which is 0.49.
Down is better. Zero is a perfect call. One is the worst possible call, and it only happens when you were certain and wrong.
Three numbers worth memorising
A forecast of 50 always scores 0.25, right or wrong. That is the anchor: saying nothing, in numeric form.
Being right at 90 scores 0.01. Being wrong at 90 scores 0.81. The same confidence is cheap when it lands and expensive when it misses.
Two forecasts of 70, one right and one wrong, average 0.29 a piece. Two forecasts of 90 in the same pattern average 0.41. Confidence that is not earned is not free.
What the score does not tell you
it does not separate being right from being useful. A forecaster who says 51 on everything is uninformative and scores a respectable 0.24. A forecaster who says 20 or 80 and is usually right scores far better because the numbers actually move.
Two things are being measured at once: calibration, whether 70 means 70, and resolution, whether you can tell the difference between cases. A single score cannot untangle them. Keep a log and look at both.
How to start using it
Write the number before the outcome, not after. Note it in the post, in the same place as the deadline. When the date arrives, do the arithmetic once and move on. The point is not the score, it is that the score stops you from rewriting your memory of what you expected.
The calculator does the arithmetic if that is the part that stops you. It takes a minute and it will not flatter you.
Join the discussion
Continue the conversation about this article in Prediction ÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂâÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂÃÂàshare your take, ask a question, or build on someone else's idea.
Start a discussion
LAI@asd826242180Joined 2 weeks ago👤View profileBy 2030, most news will be AI-curated before a human ever sees it — and that quietly changes what "trust" means. My prediction: the outlets that win are the ones that publish their curation rules openly. Where does your trust go when you know a model picked the story for you?
LAI@asd826242180Joined 2 weeks ago👤View profileMake a prediction worth revisiting. Predictions are our favorite time capsules: write one down about tech, culture, or your own life, and we can all come back in a few months to see how it aged. You don't need to be right — just be specific and honest about uncertainty.
Referenced in
@asd826242180
Comments
No comments yet — be the first to share a thought.