Level
Trainee
Accuracy
—
Best error
—
Rounds
0
Model RMSE
—
XP
0
Game setup
Choose a challenge, dataset, model, and difficulty.
Interactive fictional neighbourhood
Select a house on the canvas or generate another property.
Click a property
Property details
Score: 7/10
Score: 7/10
Score: 6/10
Make your prediction
Objective: predict within 10% of the fictional sale price.
$
$
$
Model laboratory
Level: 5
8%
3%
0%
Strength: 3
Training progressIdle
Market event
No active event
Trigger an event to test how your prediction adapts.
Market impact: 0%
Renovation simulator
Budget $75,000
Spent$0
Estimated value gain$0
Estimated ROI0%
Current property explanation
| Round | Property | Your price | Actual | Error | Score | Model |
|---|---|---|---|---|---|---|
| No completed rounds yet. | ||||||
Complete rounds or save the current property for comparison.
| # | Player | Score | Accuracy | Rounds |
|---|
Area, location, condition, access, demand, and amenities contribute differently. The game displays positive and negative feature effects for each fictional property.
Training data helps estimate relationships. Test data checks performance on unseen examples. A large training-to-test gap can indicate overfitting.
MAE measures typical absolute error. RMSE penalises larger mistakes. R-squared summarises how much fictional price variation the simulated model explains.
High complexity may memorise noise. Outliers distort metrics. Missing values reduce information. Data leakage can make evaluation look unrealistically strong.
A feature can move with prices without causing them. This game demonstrates associations only and never claims real causal effects.
Data import and game storage
Imported data remains in your browser and is used only for this session.