What's a "clean run"?

A clean run is how a streetcar moves along a segment of track when things go well: the car stops at its passenger stops, but isn't held up by red lights, traffic, turning cars or crowds.

How it's measured. Each route is split into segments, each centred on one stop and measured in 25 m slices. For every segment we take the fastest 10% of real daytime trips (06:00 to midnight). So it's not a theoretical top speed; it's what real cars actually manage there when nothing holds them up, stop included.

Delay is how much longer the typical (average) trip takes than that clean run, shown as seconds lost per 100 m. Tall peaks are where the extra time goes: usually traffic signals, busy stops and turns.

Smoothing. Positions arrive only every ~30 seconds, so a waiting car's time lands in whichever 25 m slice it happened to be reported in. The ridge and the tooltip both average over about 125 m (five slices) so the picture isn't just noise.

Peak height grows with the square root of the delay, so a handful of very bad spots don't drown out everything else and small delays still show. A spot with four times the delay stands twice as tall; the colour and the tooltip give the actual value.

Whole routes. Adding up every segment's clean run gives a route's clean-run total: a trip no single car actually makes, a bit like adding up the best time on every segment of a ride, a run nobody really did. So route cards don't present it as a trip time; they show how much a typical trip loses against it. Typical time divided by clean-run total is what traffic engineers call the travel time index (1.00 = no delay). Operating speed in the routes chart is route length over typical trip time, stops and delays included, the speed transit agencies usually report.

Stops, signals and everything else

The clean run includes a quick stop, so extra boarding time at a busy stop counts as delay. That's a different problem from red lights or traffic, so every spot is labelled using the City of Toronto's traffic signal list:

• Signal: up to about 100 m before a traffic signal (a streetcar is 30 m long and often queues behind another): red lights and queues.
• Stop: around a passenger stop with no signal: mostly boarding.
• Stop and signal: both at once. Most Toronto stops are at intersections, and with positions only every ~30 seconds there's no way to split "doors open" from "waiting for green" there.
• Elsewhere: between stops and signals: traffic, turning cars, slow zones and long queues backing up from a signal.

The tooltip also says whether a signal has transit priority.

Turn bans. Zoom in and a small icon marks signals where the City's traffic by-law bans turns; hover for which turns, which directions and when, and whether streetcars are exempt. "No turn on red" rules are left out.

Transit signal priority

The panel compares the time a typical streetcar loses approaching signals with and without transit signal priority: the median across signal approaches (the ~100 m before each signal, each direction separately, with at least 20 passes). It follows the route, direction and time-of-day filters. Read it with care:

• Priority wasn't placed at random. It was installed on purpose, likely where streetcars had the most trouble. Similar delay could mean priority is pulling bad spots back to average, or that it does little. This data can't tell those apart.
• "Priority" isn't one thing. The City's open data only says whether a signal has transit priority, not what kind: extending a green, cutting a red short, changing the phase order, a queue jump, or only helping cars running late. Very different strategies are lumped together here, and there's no public data to separate them.
• Stops at the signal. Where a stop sits at the signal, boarding time swamps whatever priority does: a car loading passengers can't use an early green. So signals with and without a stop are compared separately; the no-stop row is the fairer test of priority.
• Early days. With little data, a few bad days at a handful of signals can swing either number.

Clean-run speed view

The delay view compares each spot with its own clean run, so slowness that every car shares is invisible there. The clean-run speed view shows that instead: how fast a clean run moves through each part of the track (the fastest 10% of daytime passes), leaving out time stopped at stops. Slow spots away from stops point to curves, switches and special trackwork, or slow orders on worn track. Near stops cars are braking and pulling away, so those parts are faded.

Waits at stops

Delay is about the car; riders also feel spacing. Two cars arriving together and then a long gap leave far more people waiting than evenly spaced cars, even if every car runs on time between stops. The Waits view measures that the way London and North American transit agencies do:

• Headway: the time between one car and the next on the same route, leaving the same stop. Every vehicle on the route counts, buses included, since riders board those too.
• Average wait: for someone turning up at a random moment. Long gaps count for more: you're more likely to arrive during one, and you wait longer when you do.
• Excess wait (Transport for London's headline measure for frequent routes): the average wait minus the wait if cars had come exactly as the TTC's schedule says. It splits into uneven spacing (the cars that ran were bunched) and missing cars (fewer ran than scheduled, for example short turns).
• Gaps between cars: each stop's tooltip shows how the gaps were spread, in minutes, with the scheduled gap marked. Evenly run service is one tall bar at the schedule; bunching shows as a pile of very short gaps plus a tail of long ones.
• Bunched: a car leaving less than half the scheduled gap behind the one ahead. Big gap: more than 1.5× the scheduled gap.

Only frequent service counts: hours scheduled every 12 minutes or better. On infrequent service people time their trips to the schedule, so on-time performance matters more there. The schedule is the TTC's published one in effect that day. A stop where fewer than a third of scheduled cars showed up that day is treated as not served (a diversion, or buses off the track) and left out; its tooltip says on how many days. Departure times come from positions about every 30 seconds, so each headway is good to within about 10 to 20 seconds; a car the feed lost track of shows up as a longer gap.

Signals view

For every signal on the streetcar network, and each direction separately, this times how long streetcars take through the last 100 m before it (to just past the stop line), and compares that with a clean approach: the quickest 10% of daytime passes. A streetcar that takes more than 90 seconds longer than that has most likely missed the first green it could have made: 90 s is about a typical Toronto signal cycle. Each signal's real cycle isn't published in usable form, and it varies by intersection and time of day (some adjust it constantly), so 90 s is a stand-in, not that signal's cycle. The tooltip also gives the share held over 1 minute and over 2 minutes, so a finding doesn't hang on the exact number.

The held 90 s+ rate is the share of streetcars that were held that long. On most approaches it's around 1%; a few stand out. Two ranked lists pick those out: signals with transit priority where streetcars are still often held (priority may not be suited, may fire too late, or queues may block it), and signals without priority where it might be worth having. Busier signals count for more.

• Where, not why. A long wait can be a red light, a queue spilling back from the next signal (flagged when one is within about 120 m), or traffic. The City's data only says whether a signal has priority, not what kind or whether it fired.
• Stops in the approach. About two-thirds of approaches have a stop in them, so boarding time is mixed in. They're shown faded and left out of the rankings.
• Placement isn't random. Priority was put where streetcars had trouble, so comparing signals with and without it describes, it doesn't prove.
• Coarse timing. Positions arrive about every 30 seconds, so approach times are good to about ±15 s; the held-over-a-cycle rate is the robust number, not the exact seconds.

Faded track

Track where no streetcar has run for a few days is drawn faded: a diversion, a closure, or buses running instead. Its numbers are real but from before then, and the tooltip says when a streetcar was last seen there, with any recent TTC alert for the route. If a stretch is closed for two weeks or more and then reopens, only what's measured after the reopening counts, so rebuilt track isn't averaged with how it was before.

What this doesn't show

• Why. It shows where, when and how much time is lost, not the cause. The stop and signal labels and the TTC's alerts are clues, not explanations.
• Slowness every car shares. The clean run comes from real cars at each spot, so anything that slows every car is built into it and doesn't show as delay: slow orders on worn track, switches and special trackwork, tight curves, low speed limits. A spot can be slow for everyone and still look green here.
• Short layovers. Standing still for over 3 minutes is left out, but shorter waits near the ends of a line can still show up as delay at the last few stops.

Worth knowing

• Quiet stops. If more than about 1 in 10 cars skip a stop (nobody waiting, nobody getting off), the clean run doesn't include stopping there, so that stop shows a little more delay than it otherwise would.
• Overnight. Overnight cars are measured against the same daytime clean run. They often beat it, so overnight delay is mostly near zero.
• Left out: cars off their usual track (diversions), buses on streetcar routes, and standing still for more than 3 minutes (layovers at the end of the line).
• Early days. With only a few days of data the clean run moves around. It settles as data builds up; each route's card shows how much its clean-run total moved in the last two weeks.