accuracy and calibration

Accuracy refers to how often a system’s predictions or decisions are correct overall—like a weather forecast correctly predicting rain on most days. Calibration involves how well a system's confidence levels match the actual outcomes—if a forecast says there's a 70% chance of rain, it should rain roughly 70% of the time when such predictions are made. Good accuracy means reliable correctness, while good calibration ensures that the system's confidence estimates are trustworthy. Both are important for assessing and improving decision-making systems, especially when their predictions impact real-world outcomes.