The Lie Board

Transit agencies post arrival times. Trains arrive when they arrive. The gap between those two things is what I call the lie — and it turns out some lines lie much more than others.

The Idea

I wanted to build a live leaderboard of transit system reliability that answered a simple question: if the board says the train arrives in 5 minutes, how often does it actually arrive within ±2 minutes of that?

The answer varies wildly by line, time of day, and day of week. The MBTA Green Line is a particularly enthusiastic liar.

How It Works

The project pulls real-time arrival data from transit agency APIs (MBTA, MTA, CTA) and compares predicted arrival times to actual arrival events. Discrepancies are bucketed, ranked, and visualized.

The leaderboard updates every hour. Lines are scored 0-100, where 100 means every prediction was accurate within the margin, and 0 means the predictions were essentially random.

What I Found

The Green Line's C branch has the worst accuracy during peak hours — predictions are off by more than 5 minutes roughly 40% of the time. The MBTA's heavy rail lines (Red, Orange, Blue) perform significantly better.

The NYC Subway's worst performers are the above-ground sections, where weather and grade crossings introduce variance that the scheduling system doesn't adequately account for.

Stack

Data pipeline in Python, APIs from MBTA V3, MTA GTFS-RT, and CTA. Frontend visualization in D3 and Next.js.