Buses follow the official route shape, not dots drifting off the street.
Know when the bus actually arrives.
Live buses on the map, ETAs with a range when we're unsure, and a heads-up when a line leaves its route.
We follow the bus, not the timetable.
Every vehicle reports its position continuously. We match that trace to the official route pattern, work out where it actually is along the line, and estimate the rest from how that stretch of road has behaved recently.
- 01Match to the route
The GPS trace is matched to one official route pattern.
- 02Measure the road
Historical run times and dwell for that exact stretch, this hour, this weekday.
- 03Show the range
A time window when we have a live bus. Scheduled when we don't.
Before you walk to the stop.
A window when there's a live bus. Scheduled when there isn't.
Reading the live feed
Alerts when a line you follow diverts, and when it goes back on route.
Routes turn at corners. A diversion is easy to spot.
lines across the network
arrivals in benchmark
More reliable ETAs than the official feed. 70.5 seconds less average error.
In this benchmark view, BusTO v1.1 cuts 70.5 seconds from average error and 154.3 seconds from difficult arrivals compared with the GTT reference.
70.5 seconds less average error than the GTT reference.
Deep ETA TCN v1Current passenger model; 42.6% lower mean error than GTT in its matched test.
ETA v2The official realtime field provided through MaTO.
MaTO realtimeBusTO v1.1 improves BusTO v1
15.2% lower error
5.5% lower error
11.9% lower error
Arrivals inside the prediction window
of predictions (higher is better)+1.9 percentage points
+1.8 percentage points
The improvement, in real seconds
BusTO v1.1 and v1 use the larger 19–23 August held-out cohort. GTT is shown as a gray reference from the separate 6–7 August exact-match benchmark; see the methodology note below.
70.5 seconds less average error than GTT
78.6 seconds closer than GTT on a typical prediction
154.3 seconds less error than GTT on difficult arrivals
How often the prediction lands close
25.1 percentage points more within one minute than GTT
28.0 percentage points more within two minutes than GTT
The three values are shown together for quick comparison, but GTT comes from a different matched cohort. BusTO v1.1 and v1 used 44,944 predictions and 14,536 arrivals from 19–23 August 2026. BusTO v1 versus GTT Realtime used 6,754 exact trip-and-stop pairs across 854 arrivals on lines 4, 9, and 68 from 6–7 August, where the matched mean errors were 106.9s and 186.1s respectively. The official field may represent estimated departure rather than physical passage, so that result remains provisional until GTT/5T confirms its semantics. Full methodology and guardrails →
ETAs are guesses. We say so.
Buses due soon are usually within a minute. Farther out, the range gets wider. The app shows that instead of faking precision.
Live bus on this trip, close enough for a tight ETA.
About 3-5 minutes. A window, not a fake exact time.
Nothing on the map for this trip. Timetable time, clearly marked.
Several buses off route in the same corridor. One alone counts as possible, not confirmed.
Check before you leave.
Open the live map, pick your stop, and see if the bus is coming or already off route.