MetraX Forecast Scorecard

2026-08-20 to 2026-08-26 · measured on 489,559 logged prediction moments · updated 2026-08-27

55.5%
MetraX forecasts within 1 minute of actual arrival
136s
average forecast error
+6.9 pts
accuracy added over reading the raw feed naively

Head to head with the feed's own ETA

horizon outnMetra's native ETA within-1-minMetraX app within-1-min
[0,5) min3795199.4%99.0%
[5,15) min4087476.1%79.9%
[15,30) min824938.7%44.2%
[30,60) min135720.9%25.1%

Same moments, same stops, paired rows only (series since Aug 21, 6 day(s)). The MetraX column replays the shipped predictor under its live anchor policy — when a train is minutes away, MetraX intentionally carries the feed's own number, so the columns converge at short horizons; the model's added value appears from 15 minutes out, where a rider is deciding when to leave. GPS anchoring is not included in this replay, so the MetraX column slightly understates the live app.

What each predictive element contributes

elementwithin-1-min lifterror change
Route history+7.1 pts−8.6%
Delay momentum+4.3 pts−15.7%
Signal freshness+0.2 pts−1.8%
Full blend (shipped)+6.9 pts−16.9%

Each element measured against the naive baseline on identical rows (paired), trained only on data prior to each scored day.

How this is measured

Every metric on this page is computed from MetraX's own logged observations replayed against realized arrivals — no hand-picked examples, no survivorship. "Within 1 minute" means the displayed arrival minute was at most one minute off. The shipped-model series begins 2026-08-21.