Player dashboard
Choose a model, study the simulated history, and forecast hidden future temperatures.
Game setup
Configure the challenge, simulated climate, historical dataset, and prediction target.
Weather and data controls
These variables shape historical records and hidden future conditions.
Forecast model laboratory
Select a forecasting method and tune its educational simulation controls.
Interactive weather timeline
The Konva canvas animates conditions, forecast markers, and day-night transitions.
Prediction results
Actual temperatures remain hidden until you reveal the result.
| Forecast day | Predicted | Confidence range | Actual | Error | Condition |
|---|---|---|---|---|---|
| No forecast available. | |||||
Important factors: Temperature history, seasonality, humidity, pressure, cloud coverage, and wind.
Plotly.js forecast laboratory
Zoom, pan, inspect points, compare models, and export chart images using Plotly controls.
Model tournament
Run every forecasting method against the same hidden future dataset.
| Rank | Model | MAE | RMSE | Trend accuracy | Score |
|---|---|---|---|---|---|
| Tournament not run. | |||||
Achievements and learning progress
Local leaderboard
| Rank | Score | Model | Mode | Date |
|---|
Forecasting learning guide
Core concepts
Time-series forecasting: Uses ordered historical observations to estimate future values.
Trend: A persistent increase or decrease across time.
Seasonality: A repeating pattern linked to days, months, or seasons.
Forecast horizon: The number of future steps predicted.
Confidence interval: A range expressing simulated forecast uncertainty.
Reliable evaluation
MAE: Average absolute difference between predictions and actual values.
RMSE: Penalises larger errors more strongly than MAE.
Overfitting: A model memorises history but performs poorly on unseen days.
Data leakage: Future information accidentally enters model training.
Uncertainty: Weather systems contain noise and unexpected changes.
How to play
- Choose a game mode, difficulty, climate zone, and weather scenario.
- Generate a historical dataset with hidden future temperatures.
- Inspect the timeline and Plotly historical chart.
- Select and tune a forecasting model.
- Run the forecast, then reveal actual values for scoring.
- Compare models, earn XP, unlock achievements, and improve your best score.
Frequently asked questions
Does this game use real forecasts?
No. It creates fictional historical and future weather data for education.
Which model is always best?
No model always wins. Performance changes with trends, seasonality, noise, and sudden events.
Why can longer forecasts be less accurate?
Uncertainty accumulates as the forecast horizon extends farther into the future.
What does missing data do?
It removes historical observations, forcing the simulator to interpolate before training.
How is the score calculated?
The game combines MAE, RMSE, trend accuracy, confidence coverage, speed, and bonuses.
Where is progress stored?
Progress and leaderboard entries are stored locally in your browser.