Game setup
Choose the puzzle and dataset
Progress and best scores are stored in this browser.
Use 0 for no timer.
Read predictions against actual outcomes
True positive: predicted positive and actually positive. False positive: predicted positive but actually negative.
True negative: predicted negative and actually negative. False negative: predicted negative but actually positive.
Accuracy = (TP + TN) / Total
Precision = TP / (TP + FP)
Recall = TP / (TP + FN)
Specificity = TN / (TN + FP)
F1 = 2 × Precision × Recall / (Precision + Recall)
MCC = (TP×TN − FP×FN) / √((TP+FP)(TP+FN)(TN+FP)(TN+FN))
- Keyboard users can classify records with the accessible dropdown table.
- Canvas cards can be dragged with mouse or touch.
- The positive class is shown before each puzzle.
Live session
Scoreboard and progress
Score0Best: 0
Puzzle0 / 0Not started
Time—Current round
Streak0Correct rounds
Lives3Expert mode uses lives
Session accuracy0%0 correct
Current challenge
Start a game to generate a puzzle
The positive class will appear here.
Accessible record assignments
These controls mirror the canvas and provide a keyboard-friendly alternative.
| Record | Actual | Prediction | Confidence | Your category |
|---|
Classify this result
Record
Count every record in the dataset, then enter the four matrix values.
Actual positive
Actual negative
Predicted positive
Predicted negative
| ID | Actual | Predicted | Score |
|---|
One displayed outcome label is incorrect. Select that row.
| Select | ID | Actual | Predicted | Displayed label |
|---|
Configure the game, then select Start game.
Current model
Live confusion matrix and metrics
Actual positive
Actual negative
Predicted positive
True positive0
False positive0
Predicted negative
False negative0
True negative0
Threshold laboratory
Tune classification decisions and error costs
Current decision cost: 0.00
Costs update with the threshold.
Analytics
Performance, outcome counts, and round history
Dataset tools
Import, create, and export records
Accepted actual values: 1, 0, positive, or negative. Scores must be between 0 and 1.
Session report
Learning review and recommendations
Complete at least one puzzle to generate a learning report.