Why accuracy is usually the wrong number
On an imbalanced problem — moderation, fraud, anomaly detection — accuracy is dominated by the majority class. A classifier that always predicts negative on a five percent positive rate scores ninety-five percent and is worthless.
The calculator states the base rate and what a trivial always-negative classifier would score, so you can see immediately whether your accuracy figure means anything.
Precision against recall is a product decision
High precision, lower recall means missing cases rather than raising false alarms — right when a false positive is expensive, such as auto-rejecting a legitimate order.
High recall, lower precision means catching nearly everything and flagging some innocents — right when a miss is expensive, such as a safety filter. Neither is better in general, and the model cannot make the choice for you.