User-worked with stop lights level crossing on the Wales route (ELR SWM2).
Information only. Always follow the signals and barriers at the crossing. Predictions may be wrong.
| 09:56 | 🚄 1V04 | Nottingham → Cardiff Central Down (NE → SW) | passed |
| 09:57 | 🚛 6B13 | Robeston Sdgs → Westerleigh Puma DBC Up (SW → NE) | no show |
| 09:57 | 🚆 2G54 | Cardiff Central → Cheltenham Spa Up (SW → NE) | passed |
| 10:07 | 🚄 1S43 | Cardiff Central → Edinburgh Up (SW → NE) | -1 min |
| 10:19 | 🚆 2G51 | Cheltenham Spa → Cardiff Central Down (NE → SW) | approaching |
On today’s timetable, Three Gates crossing is predicted to be closed for about 2 h 4 min between 06:00 and 22:00, across roughly 65 closures. The longest single predicted closure is 1 minute. Freight paths scheduled “as required” (most don’t run on any given day) aren’t counted here; when one does run, the live timeline above shows it.
Longest closure so far today: 0.6 min (00:14–00:14, 1 train), timed from trains actually seen on the signalling wire: the window edges use the same per-type margins as the forecast, since barrier positions aren't in any feed we can read.
Closures cluster around 09:22–10:22 in the morning and around 17:38–18:38 in the evening. Gaps between closures are longest outside those windows. Other days run a different timetable, so the peaks move.
The crossing is near Chepstow, Monmouthshire.
Three Gates is a user-worked with stop lights crossing (UWCMSL) on the Gloucester–Llanwern Exchange Sdgs line (Wales route · ELR SWM2). Forecasts here combine the working timetable with live train-movement data (and live signalling where the crossing has it), so they reflect actual running including delays and freight, though they can still be wrong when the network is disrupted.
We score this page against what actually happened on the railway: over the last 7 days that covered 419 trains through this crossing.
100.0% of observed passes were being warned about on this page when they passed, and 98.3% of the warnings we could verify came true.
Typical timing error was ±4s (9 in 10 within 60s). This scores when trains pass: barrier movements aren't in the public feeds we can score against.