Probability is not a promise
What a 60% forecast means, why one match proves almost nothing, and how to judge a model across repeated decisions.
One match is an outcome, not a verdict
A 60% estimate leaves a 40% chance that the event does not happen. If the selection loses, the estimate is not automatically wrong; if it wins, the estimate is not automatically right.
Forecast quality becomes visible across many comparable, prospectively recorded predictions. The central question is whether events assigned about 60% probability happen about 60% of the time—not whether the latest pick won.
Calibration and discrimination do different jobs
Calibration asks whether stated probabilities match observed frequencies. Discrimination asks whether the model meaningfully separates higher-probability events from lower-probability ones.
A useful model needs both. A forecast can be well calibrated but uninformative if it assigns nearly the same probability to every match.
How FORM/PRICE presents uncertainty
We show the estimated probability beside the available price and preserve the timestamped decision. Hit rate, return and calibration remain separate measures because none can safely stand in for the others.
- Read 60% as uncertainty, not certainty.
- Judge a series, not a screenshot.
- Demand the original timestamp and complete ledger.
Sources
Direct links are preserved so the editorial reasoning can be checked independently.