Carregando…
Foz do Iguaçu · Ciudad del Este · Puerto IguazúSaturday, July 4, 2026
Transparency of the numbers
The border movement forecast comes from real data — weekday, holidays in all 3 countries, weather and river flow. Here we show the history, the accuracy (forecast vs measured) and the open methodology. No black box.
Forecast movement history
Series recorded 3×/day — the basis of our predictive capability.
Proprietary history · grows every day
collecting since 15 Jan 2024
Append-only time series: nothing is deleted or overwritten. Each day becomes a record you can't recover backwards — forecast vs actual, accuracy, reports and questions. It's what accumulates and can't be copied.
↓ download series (CSV · CC BY)Validation against public visitation 2025
| Month | Forecast (2024) | Actual (2025) | Match |
|---|---|---|---|
| Jan | high | high | ✓ |
| Feb | low | low | ✓ |
| Mar | low | low | ✓ |
| Apr | low | low | ✓ |
| May | low | low | ✓ |
| Jun | low | low | ✓ |
| Jul | high | low | ✗ |
| Aug | high | high | ✓ |
| Sep | low | low | ✓ |
| Oct | high | high | ✓ |
| Nov | high | high | ✓ |
| Dec | high | high | ✓ |
Out-of-sample validation: the seasonality learned in 2024 predicted 2025 and matched 11 of 12 months (high × low). Compared against Foz Airport passengers (ANAC) — an independent source that is not a model input. Validates the seasonal/weekly PATTERN, not the hour-by-hour count (that isn't public — it's what Smart CTI would integrate). source: ANAC (open data) · cross-checked with ICMBio/PNI and Itaipu → · open source (lib/backtest.ts) →
Forecast accuracy · forecast × actual
1 reports — reports received. The rate appears once forecast and actual coincide on the same day (updates on the next capture).
Forecast (3×/day heuristic) × actual (community “how is it now?” reports, day's modal level). 60-day rolling window — crowd-sourced, not an official count. how we calculate →