Ride the Line
NYC Subway vs Tokyo’s Yamanote — Same problem, different toolkits
Welcome Aboard
Every city has a train story. This one starts in Tokyo and ends with a delayed R train.
Before my solo trip to Tokyo in December 2025, someone on my feed had already solved my first problem for me: download Google Maps, it makes navigating the station exits so much easier. I filed it away and forgot about it until I landed. Then the train came exactly when it said it would, and the one after that, and the one after that. Two weeks of riding that loop around the city without once pulling out my phone to check if something was running late, because nothing ever was. A stranger on the internet had handed me the most reliable transit experience of my life, and I didn’t fully appreciate it until I came home and opened the MTA app. Delayed.
That contrast is what this project is actually about. Not a complaint but a real question: why does one system do what it says it is going to do, and the other treat its own schedule like a rough suggestion? The answer goes deeper than punctuality. It gets into how each system is physically built, how demand moves through it, how often trains actually come, and maybe most surprisingly, what each city chooses to measure and share about how it is performing at all.
Four things this piece is going to get into:
- Why the physical layout of each system, a branching web versus a closed loop, shapes almost everything else
- Where and when each city actually moves its riders, and what the data gap between them reveals
- What “frequent” really means and why consistent service matters more than peak service
- What reliability data each city publishes, what it keeps quiet, and what both of those choices tell you
The two systems at the center of this are the MTA’s Brooklyn-Manhattan Transit group, the BMT, which covers the J, Z, L, M, N, Q, R, and W lines across Brooklyn, Manhattan, and Queens, and JR East’s Yamanote Line, a single closed loop running around the heart of Tokyo.
One note before we go: these two systems are not perfectly comparable, and this project doesn’t pretend otherwise. The MTA publishes hourly ridership for every station, monthly on-time performance by line, and full schedule data as open data. JR East publishes annual station totals and schedules, but no hourly ridership and no delay archives. Where direct comparison isn’t possible, the charts show what is comparable and name the gap. That asymmetry is a finding, not just a limitation.
Stop 1 — The Networks
Before any numbers, it helps to understand what kind of thing each system actually is, because they are solving the same problem with completely different structures, and that structural difference shapes every comparison that follows.
The BMT is a web. Eight lines branch across Brooklyn, Manhattan, and Queens, sharing trunks, splitting at junctions, serving dozens of distinct neighborhoods across roughly 80 miles of track. The Yamanote is a loop: one line, 29 stations, 21 miles, running clockwise and counterclockwise around Tokyo’s commercial and cultural core without a beginning or an end.
That is not a small design distinction. A loop has no terminal bottlenecks and no directional asymmetry. When demand moves clockwise, the counterclockwise service does not run empty. A delay anywhere on the Yamanote propagates around the circuit, but there is no single failure point that can cascade across multiple branches simultaneously. The BMT’s branching structure gives it reach and flexibility that a loop never could, but also a fragility: one problem on a shared trunk affects every line running through it.
The maps below make the contrast visible. Look at the BMT and you see a system built to reach everywhere. Look at the Yamanote and you see a system built to run perfectly.
MTA Brooklyn-Manhattan Transit (BMT) Map
The BMT network across three boroughs. Eight lines, shared infrastructure, and the operational complexity that comes with a century of incremental expansion. Toggle individual lines to isolate shared trunk segments.
JR East Yamanote Line Map
The Yamanote loop around central Tokyo. 26 of its 29 stations rank among JR East’s 100 busiest. The closed loop means every station benefits equally from the same service pattern. There is no branch that gets less.
Stop 2 — Demand
Once you see how each system is built, the next question is who is actually using it and when. This is where the data starts telling a story, and also where the limits of what Tokyo is willing to share become impossible to ignore.
On the BMT, the MTA publishes hourly ridership for every station complex. What that data shows is a system shaped almost entirely by the commute: demand builds sharply into the 8am rush, collapses through midday, spikes again at 5pm, and drops off steeply after. Weekends are a different system entirely, with lower overall volume, a later peak around noon, and a much flatter curve as commuters give way to leisure riders.
Average BMT ridership by hour of day, split by day type. The shaded bands show the 10th to 90th percentile range across individual days. The shape of the curve is consistent, but the volume varies day to day. The commute signature is unmistakable. Data: MTA Open Data, 2025.
Yamanote Station Demand
For the Yamanote, JR East does not publish hourly ridership publicly. What they do publish is annual daily average boardings per station, and those numbers alone tell a significant story. Shinjuku handles over 666,000 daily boardings in the fiscal year of 2024, making it one of the busiest train stations in the world. The top 12 stations on the loop collectively move more people per day than many entire transit systems. Every single Yamanote station grew year-over-year from FY2023 to FY2024, continuing Tokyo’s post-COVID recovery across the board.
Daily average boardings per station, FY2024. Stations ordered by ridership. This is not a chart about outliers. Nearly every station on the loop operates at volumes that would make it a major hub on any other system. Data: JR East official passenger statistics.
Year-over-year comparison for the 12 highest-volume Yamanote stations. Every station grew. This reflects not just post-COVID recovery but a system operating closer to its pre-pandemic ceiling than any comparable metro network. Data: JR East official passenger statistics.
Year-over-year ridership growth rate by station. The diverging format makes it easy to see which stations are recovering fastest and whether growth is concentrated at major hubs or spread across the loop. Data: JR East official passenger statistics.
Stop 3 — Frequency
Demand tells you who shows up. Frequency tells you what happens when they get to the platform, and this is where the structural difference between a branching network and a closed loop becomes most concrete.
On the BMT, the MTA scales service to match demand. More trains during rush hours, fewer overnight, a step down on weekends compared to weekdays. That is a rational approach to running a complex multi-line system, but it means riders feel the difference. Waiting four minutes at 8:30am versus twelve minutes at 10am is not just an inconvenience. It changes how you plan your day and how much margin you have when something goes wrong.
BMT scheduled trips per hour by day type. The system throttles up and down in direct response to demand. The gap between weekday peaks and weekend shoulders is the cost of running a branching network that has to balance service across eight lines sharing the same infrastructure. Data: MTA GTFS static schedule files.
The Yamanote barely moves. From early morning through late evening on weekdays, trains arrive every three to four minutes. Many Tokyo commuters do not check a schedule. They walk to the platform.
Scheduled departures per hour at Shinjuku, by day type. The flatness here is the architectural story. Because every Yamanote train completes the full circuit, there is no empty repositioning, no dead-heading, no direction with systematically less service than the other. The loop supplies itself evenly. Data: JR East GTFS via ODPT.
Stop 4 — Reliability
Frequency gets you to the platform. Reliability determines whether the train that is supposed to come actually does, and this is where the comparison gets genuinely complicated, because reliability does not mean the same thing in both cities, and that difference is itself part of the story.
In New York, the MTA measures on-time performance at the terminal: did the train complete its full run within the allowable time window? By that standard, the BMT runs around 86 to 88% on time on weekdays, with weekends performing somewhat better. The numbers have softened over 2025. The primary constraint is well-documented: legacy fixed-block signals, the same aging infrastructure driving the MTA’s ongoing modernization program. In a branching network, a signal issue on one trunk does not just delay one train. It cascades. One late train becomes twenty.
BMT monthly terminal on-time performance by day type, 2025. Circles mark each month’s peak performance. The metric is publicly available and updated regularly. The MTA cannot obscure this story because it chose to make the data open. The softening trend over 2025 reflects a system contending with signal infrastructure that predates the riders using it. Data: MTA OTP reports.
In Tokyo, the standard is departure within seconds of the posted schedule. Not minutes. Seconds. JR East does not publish line-level delay statistics publicly, but independent reporting puts average delays on non-Shinkansen lines at around 50 seconds. The Yamanote runs under D-ATC, Digital Automatic Train Control, a system that continuously calculates safe following distances and removes human reaction time as the hard floor on how close trains can run to each other. That is why three-minute headways are maintainable: the system does not depend on a driver’s judgment call.
Median minutes between trains by hour of day. The shaded band shows the interquartile range. The schedule is not just consistent on average, it is stable. A dotted line marks the 4-minute peak headway that holds across nearly the entire service day. When riders do not need to check a schedule, the system has already solved half the reliability problem. Data: JR East GTFS via ODPT.
The metrics are not directly comparable. Different definitions, different standards, different institutional cultures around what “on time” even means. But the direction of the gap is clear, and so is what is driving it on the New York side: signal technology built for a different era, running a network that has grown far beyond what that technology was designed to handle.
So What?
I started this in Tokyo, surprised that a train schedule could just be true. Two weeks of the Yamanote being exactly where it said it would be, and then home to New York, where that is still a pleasant surprise when it happens.
But the closer I looked at the data, the more the story shifted. This is not really about punctuality. It is about three things that kept surfacing across every section of this project:
Structure determines almost everything. The loop versus the web is not just a geographic difference. It is the reason the Yamanote runs flat headways all day, why there is no directional dead-heading, why a delay does not cascade the way it does on a shared trunk. Every operational advantage the Yamanote has, in frequency, in consistency, in the way it absorbs disruption, traces back to that one architectural decision made more than a century ago.
Consistent frequency is its own form of reliability. When trains come every three minutes, a delay is an inconvenience. When they come every twelve, it is a disruption that ripples through your whole day. The BMT’s peak service is genuinely strong. The problem is what happens in the valleys, and what those valleys mean for any rider who cannot build their schedule around the peaks.
What cities choose to measure tells you what they are accountable for. The MTA publishes hourly ridership, terminal OTP, and full schedule data as open data. JR East publishes annual station totals and scheduled service, and almost nothing about actual operational performance in real time. New York lets you watch the system struggle. Tokyo keeps its operational record private. Neither is obviously the right call, but the difference matters: open data creates the conditions for public accountability, and accountability creates pressure for improvement.
NYC cannot copy-paste Tokyo. The scale is different, the history is different, the capital and labor constraints are different. But the underlying principle is not Tokyo-specific: invest in signal and train control technology, reduce the cascading failures that turn one late train into twenty, and the system gets more reliable regardless of whether it is a loop or a web.
The gap between these two systems is real. What this project tried to show is that it is also explainable, and that explanation matters more than the complaint.