howbadisthetrain

NYC · methodology

Where these numbers come from

Every score on this site is computed from the MTA's own published performance data. This page lists the exact datasets, the definitions, and — more importantly — the limits.

What the score is

The score is the MTA's Customer Journey Time Performancefor the most recent published month, divided by ten. That's it. No weighting, no secret blend, no editorial adjustment.

Customer Journey Time Performance is the share of rider trips that finish within five minutes of the scheduled time, measured by the MTA across every trip on the line. So 83.0% becomes a score of 8.3, and 78.2% becomes 7.8. If you want the raw percentage, every page shows it directly — the out-of-ten figure is a presentation choice, not a transformation.

Earlier versions of this site blended five invented inputs into a "vibe score" and described it as such. That is no longer true, and this page has been rewritten to match what the code actually does.

The datasets

Four datasets, all published monthly by the MTA as open data, all per subway line:

  • Customer Journey Time Performance — the headline metric and the basis of the score. Split by peak and off-peak.
  • Terminal On-Time Performance — the share of trips arriving on schedule at the terminal. Weekday figures.
  • Wait Assessment— the share of platform waits that fall within the scheduled headway. This catches the "the train is running, you just waited twelve minutes" problem, which on-time performance alone hides.
  • Trains Delayed by Cause— delays grouped into the MTA's six cause categories. This is where the "biggest delay cause" line and the reliability ranking's cause bars come from.

All of it comes from MTA open data on data.ny.gov. No API key, no scraping, no third-party aggregator. The snapshot behind the current build was generated 2026-10-05, and the latest published month is August 2026.

The status labels

The badge on each line — On time, Delays, or Bad — comes from plain thresholds on the same percentage, and nothing else:

  • On time — 85% or more of trips within five minutes of schedule.
  • Delays — 80% to 85%.
  • Bad — under 80%.

These thresholds are round numbers, disclosed so you can disagree with them, and they are not tuned to favour or punish any particular line.

What this site does not claim

  • It is not real-time. The MTA publishes monthly, about two months behind. The latest month on the site is August 2026, and it will not change until the MTA publishes the next one. There is no live feed behind this, and the site no longer says "ranked right now" anywhere.
  • A monthly average hides bad days. A line at 85% for the month still had individual trips that were much worse.
  • It says nothing about crowding. The MTA does not publish per-car crowding data, so you will not find a crowding number here. We would rather omit it than invent one.
  • J and Z are reported jointly. The MTA measures them as a single series because they share a track and a stopping pattern. Both pages show the same figures rather than a split we made up.
  • SIR is not covered.The Staten Island Railway is not in the subway metrics datasets. It gets an explicit "no data" page rather than a borrowed number.
  • The shuttles are combined. The MTA reports the three shuttle services separately; this site combines them into the single S the map shows, weighted by riders.

The "vs. driving" numbers are estimates

The cost and time comparison on each line page is a rough order-of-magnitude nudge, not a quote. Uber and parking costs vary wildly by borough, time of day, and whether you are a regular. We use NYC TLC averages and a mid-borough surcharge. Do not treat it as a number for any specific trip.

Refresh cadence

The MTA publishes a new month of data roughly two months after the fact. When it does, scripts/build-line-history.py is re-run, the resulting artifact is committed, and the site rebuilds. That is the whole pipeline — it is a script, not a service, and you can read it.

Corrections

If a number here disagrees with the MTA's published data, the number here is wrong and we want to know. Get in touch — we read every email.

Source: MTA open data (data.ny.gov), datasets r7qk-6tcy, s666-h6b7, f6rf-2a3t, 9zbp-wz3y.