“How accurate are you?” deserves a number, not an adjective. That number is the Brier score. Here's what it is, why it's the honest way to grade anyone who deals in probabilities, and what 0.070 across 67 sealed calls actually means — no statistics background required.
When someone asks a forecaster 'how accurate are you?', the honest answer is a number, not an adjective. 'Usually right' is what every pundit says and it means nothing, because it hides how hard the calls were and how confident the person was when they made them. The number that doesn't let you hide is the Brier score. This explains what it is, why it's the fairest way to grade anyone who trades in probabilities, and what mine — 0.070 across 67 closed calls — actually means, with no statistics background assumed.
The first idea to accept is that a good forecast is not 'right' or 'wrong' — it's a probability. 'There's an 80% chance of X' isn't falsified when X fails to happen; 20% things happen all the time. So grading forecasters can't just count hits and misses. It has to reward confidence when you're right and punish confidence when you're wrong, in proportion. That's exactly what Glenn Brier's 1950 scoring rule does.
The mechanic is simple enough to do on paper. For each call, take the probability you assigned to the thing that actually happened, subtract it from reality (1 if it happened, 0 if it didn't), square the difference, and average that over all your calls. In symbols: the mean of (p − o)². The squaring is the whole trick — it makes big confident errors hurt far more than small hedged ones.