Start hereWhy can’t time-ordered data be shuffled like ordinary rows?
Forecasting predicts the future from information available in the past. Temporal order creates dependence, trend, seasonality and a strict information boundary.
Building interactive view…
Technical lensFormalise what the visual is doing
Lag features, differencing, autoregression and state-space/decomposition methods model temporal structure. Validation must roll forward rather than sample future observations into training.
Technical questionUse a tiny case to make the mechanism observable. Lag features, differencing, autoregression and state-space/decomposition methods model temporal structure. Validation must roll forward rather than sample future observations into training. Verify one intermediate quantity, state change or mapping independently; then predict the consequence of this change: Randomly split a time series and identify how future patterns leak into training.
Practitioner lensUse it responsibly
Always define forecast horizon and prediction origin. Compare against naive seasonal baselines.
Transfer testTransfer this idea to a new example and justify each decision using this practitioner rule: Always define forecast horizon and prediction origin. Compare against naive seasonal baselines. Then explain what should change if you deliberately test: Randomly split a time series and identify how future patterns leak into training.
Worked explorationUse the visual as an experiment, not decoration
Forecast month t+1 using only data through t. Compare this with a random train/test split that places future observations in training. The random split leaks temporal information and produces an optimistic evaluation.
Technical lens
Lag features, differencing, autoregression and state-space/decomposition methods model temporal structure. Validation must roll forward rather than sample future observations into training.
Practitioner check
Always define forecast horizon and prediction origin. Compare against naive seasonal baselines.
Prediction before interactionRandomly split a time series and identify how future patterns leak into training.
Exploration walkthroughTurn the interaction into an evidence trail
Forecast month t+1 using only data through t. Compare this with a random train/test split that places future observations in training. The random split leaks temporal information and produces an optimistic evaluation. Before moving the control, state your prediction. After the visual changes, name the specific state, statistic, boundary or mapping that changed and explain why that change is consistent—or inconsistent—with your prediction.
- Record one observable quantity before the interaction and the same quantity afterwards.
- Change one factor at a time so the causal effect of the control is inspectable.
- Use an edge or failure case to discover where the concept stops behaving as the simple story suggests.
