Precision or Recall? Classification Metric Game

Game mode

Difficulty

Scenario category

Practice mode gives explanations, unlimited retries, and no life penalty.

Score0

Streak0

Lives∞

Round0 / 10

PlayerPlayer 1

Time--

Ready Intermediate

Start a game to receive your first classification scenario.

You will compare the harm caused by false positives and false negatives, choose a metric, and tune the decision threshold.

Positive class—

Positive base rate—

False-positive cost—

False-negative cost—

Review capacity—

Safety level—

Business objective—

Data balance—

1. Choose the most suitable metric

2. Identify the more costly classification error

False positive False negative Both are similarly costly

3. Tune the operating threshold

Classification threshold 0.50

More recallMore precision

Precision—

Recall—

F1-score—

Accuracy—

Specificity—

Balanced accuracy—

Predicted positives—

Estimated cost—

Animated case simulation

Circles represent cases. Shape outlines distinguish actual positives from actual negatives without relying only on colour.

Confusion matrix

Predicted positive

Predicted negative

Actual positive

TP0

FN0

Actual negative

FP0

TN0

Precision–recall threshold trade-off

Metric reference panel

Use the formulas and decision rules below when comparing classification risks.

Metric	Formula	Best used when
Precision	TP / (TP + FP)	False positives are expensive, disruptive, or limited by review capacity.
Recall	TP / (TP + FN)	Missing a positive case creates the greatest safety or business risk.
F1-score	2PR / (P + R)	Precision and recall both matter and one balanced score is needed.
Specificity	TN / (TN + FP)	Correctly rejecting negative cases is the main operational requirement.
Accuracy	(TP + TN) / N	Classes and error costs are reasonably balanced.
Balanced accuracy	(Recall + Specificity) / 2	Classes are imbalanced and both classes need equal attention.
PR-AUC	Area under P–R curve	The positive class is rare and ranking quality matters across thresholds.
ROC-AUC	Area under ROC curve	Overall discrimination is compared across many thresholds and classes are not extremely rare.

Fast rule: choose recall when missed positives are worse; choose precision when false alarms are worse.

Custom scenario builder

Create a reusable fictional classification challenge. Saved scenarios appear in future rounds.

Scenario name

Performance dashboard

Review metric choices, response speed, threshold quality, and category performance.

Round	Scenario	Chosen	Correct	Threshold	Points	Time
No completed rounds yet.

Accessibility and game settings

High-contrast mode

Reduced motion

Larger text

Sound feedback

Use three-life mode outside practice

Round count

Player 1 name

Player 2 name

Keyboard controls: keys 1–8 select metrics, arrow keys adjust threshold, H uses a hint, and Enter submits or advances.