Stoplight Timer

Stoplight Timer is an offline, learning-based traffic light countdown.
It does NOT query any government API or cloud service. Instead, it learns
the cycle of YOUR specific intersection from a few taps, then predicts when
the light will change next.

HOW IT WORKS
• Tap "Red → Green" when the light turns green
• Tap "Green → Red" when it turns red
• The algorithm measures green and red durations, infers the full cycle,
and uses your most recent tap as a phase anchor
• Adaptive clustering automatically separates rush-hour from off-peak when
enough data accumulates
• Integer cycle snapping (e.g. 215 s, not 215.04 s) prevents fractional
drift from accumulating across days

ACCURACY DEPENDS ON YOUR DATA
This is a statistical learning tool. The more transitions you record, the
better it predicts:
• 5-10 samples → ±5-15 s prediction accuracy
• 10-25 samples → ±2-8 s
• 25-50 samples across 3-5 days → ±2-5 s
• 50+ samples spanning multiple time-of-day buckets → ±1-3 s

A common pattern: spend the first session observing 5-10 transitions at
the same intersection, then add 1-2 a day for a week. After that, the
prediction should match real switches within seconds.

PRIVACY
• 100% local — all data lives in the app sandbox on your device
• No internet permission, no analytics, no ads, no tracking
• Backup and restore via JSON files YOU control

WORKS BEST FOR
• Fixed-time signal plans (most urban intersections)
• Coordinated arterials with morning / evening cycle differences
• Multi-intersection tracking (separate stations per location)

LIMITATIONS
• Vehicle-actuated lights with no traffic give variable cycles — the
algorithm produces a reasonable average but not a precise countdown
• SCATS adaptive intersections need a few fresh samples each week as
the central system fine-tunes plans
• If the city changes signal timing, old data becomes stale; record a
few transitions to re-learn

FEATURES
• Multi-station support (track multiple intersections)
• Auto rush-hour detection (1D K-Means clustering)
• Robust statistics resistant to mis-taps (MAD outlier rejection)
• Backup / restore as JSON
• Light and dark Material 3 theme
• Completely free, no ads, no in-app purchases

This is open-development hobby software. Issues / suggestions welcome via
the support contact.