- Waiting for prediction.
- Waiting for prediction.
| ID | Study | Attendance | Assignments | Previous | Actual | Predicted | Error |
|---|---|---|---|---|---|---|---|
| Generate a dataset to view records. | |||||||
| Time | Model | Prediction | Confidence | Points |
|---|---|---|---|---|
| No saved attempts. | ||||
Regression and prediction
Regression estimates a numeric outcome from input features. Predictions are informed guesses, not guaranteed results. Uncertainty should always be shown.
Training and testing data
Training data helps a model learn patterns. Testing data checks performance on unseen records. A strong model should work beyond memorised examples.
Correlation and causation
A feature may move with exam scores. That relationship does not prove direct causation. Real learning outcomes involve many hidden factors.
Bias, overfitting, and generalisation
Biased data can produce unfair estimates. Overfitting makes training results look unusually strong. Generalisation measures performance on new fictional students.
Feature importance
Importance indicates how strongly a model uses each feature. It does not measure personal worth. Feature effects can change across datasets and models.
This game uses fictional records for learning. It must not guide real admissions, grading, discipline, or student ranking. Real educational decisions require qualified human review.
Model outputs can reflect hidden assumptions and data bias. Confidence values are simulated and should not be treated as certainty. Do not enter identifiable student information.
Is this a real grading system?
No. It is an educational game using fictional data and simulated models.
Which model is always best?
No model wins every dataset. Compare validation error, stability, and generalisation.
Why does the score change?
Changes can come from model choice, random noise, difficulty, and enabled features.
What does confidence mean?
It describes simulated prediction stability. It is not a real probability of success.
Can I import my own CSV?
Yes, but use fictional or anonymised records. Never upload identifiable student details.
What is overfitting?
Overfitting occurs when training performance is strong but unseen-data performance weakens.
How are game points awarded?
Points reward accurate guesses, useful model choices, streaks, and improvement planning.
Does studying always increase predictions?
Usually, but diminishing returns and other features can change the simulated effect.
Where is progress stored?
History and achievements are stored locally in your browser using local storage.