MTA, close the 3rd Ave L stop.

Every stop costs the people on board time: the train slows down, opens its doors, and speeds back up. Cutting a stop gives that time back to everyone riding through, but the people who used it now have to walk to another stop. This page weighs the two.

hours saved = time saved on trains − extra walking × walking penalty

The number next to each train is the total hours riders would save on an average weekday if that stop closed. Above zero, riders come out ahead. Below zero, the stop is worth keeping.

Time saved on trains. We take every weekday subway trip in the MTA’s ridership estimates (January to August 2026) and send it through the real weekday schedule, catching whichever train gets there first. Then we remove the stop and do it again. The difference is the time saved. Riders passing through get there sooner; riders who changed trains at that stop may have to change somewhere slower, which counts against the cut.

Extra walking. People who started or ended their trip at the stop walk to another one instead. We measure that walk on real streets from the station entrances, averaged over where people live and work nearby.

The settings. Stop time is how long each stop delays the people on board. The walking penalty reflects that a minute on foot feels longer than a minute on a train; planners usually use 1.5 to 2. Walking speed turns distance into time. You can also choose whether riders walk to the next stop on the same line, the nearest station on any line, or, as a worst case, the full distance between stations.

Left out. Big transfer hubs like Times Sq, where the ridership data can’t tell which station people used, and the last stop on each line. The model also ignores crowding, delays, new riders drawn by faster trips, and buses.

Right of zero the scale is even. Left of zero it is compressed, so a few very busy stops fit on the screen. Exact hours are in the right-hand column.

Methods and sources

Ridership. The MTA Subway Origin-Destination Ridership Estimate gives average trips between every pair of station complexes, by hour and day of week. We use Monday to Friday for January to August 2026, averaged into one typical weekday of about 4.05 million trips.

Schedule and routing. Trips are routed on the MTA’s GTFS weekday timetable with RAPTOR, a standard algorithm that finds the earliest arrival across every scheduled train, allowing up to three transfers. For each hour, each trip departs at six evenly spaced times and the results are averaged, so waiting on the platform is included. Transfer times come from the MTA’s own transfer table (usually 3 minutes).

Cutting a stop. The stop is removed from every train that passes through it. Riders still on board after the skipped stop arrive earlier by the stop time; the rest of the timetable is unchanged, so waits elsewhere stay the same. Every trip is then routed again, and the change in total travel time for everyone who didn’t use the stop is the time saved. We run this at stop times of 0, 30 and 70 seconds and interpolate between them for the slider.

Walking. Distances are shortest paths on the OpenStreetMap street and sidewalk network, measured from the MTA’s station entrance locations. Each street point within 1.2 km that is closer to this stop than to any other station counts as its area. For each point, the extra walk is the distance to the nearest other suitable station minus the distance to this one. Points are weighted by residents from the 2020 Census and jobs from Census LEHD (2022), and averaged. Riders using the stop come from the ridership data.

Which stops are scored. Station details come from the MTA Subway Stations dataset. A station complex with more than one station (like Times Sq) is a hub and is left out, because ridership is only reported per complex. Terminals are the first or last stop of any daytime route pattern. That leaves 356 scored stops.

Credit. The timetable routing approach was inspired by Anita’s NYC Subway Riders map.