Traffic · Forecast record
How accurate are our forecasts?
We forecast when the centre lane of the Lions Gate Bridge will next switch, and how long the southbound queue will be. Every forecast is checked afterwards against what really happened.
Published Updated
Scored against what the lane really did
95%
of the time, the lane had switched by our “expected by” time
Our aim is nine in ten. Before 10 am we fall short: 88%.
92%
the same over the twelve weeks before, each week forecast only from earlier days
2.1min
typical miss on when the lane opens on a weekday morning
Weekend mornings vary more: about 20 minutes.
Our forecast and the published times
The province publishes approximate weekday times for the centre lane (Counterflow, Government of British Columbia). They list no switches in between. Operators switch the lane by hand, as traffic needs. On Tuesday 3 November 2026, that meant 19 switches.
Published times
What the lane did, Tue 3 Nov
Our forecasts that day
98%
of the time from 6:00 to 9:30 am, the lane ran as published
65.5%
of the time from 3:00 to 7:00 pm. The rest of the time it was southbound or switching.
27
switches on a typical weekday. The published times list none.
When does the lane first switch on a weekday morning?
How long will the southbound delay be an hour from now?
The record, day by day
Southbound delay, forecast an hour ahead, Wed 28 October 2026
- What happened
- Our forecast, an hour before
- Our likely range
Every day: how often the lane had switched by our “expected by” time
Does it get better with more data?
A little, and mostly in the first three months.
With one month of data, the model’s best single guess of the switch time missed by about two minutes more than with five to eight months. After three months, more data changed little. The “expected by” time held about as often either way.
- Test days 28 October to 9 November
- Test days 5 to 27 October
How it works
We keep the record
The province publishes the bridge’s lane signals and road detectors. We save them every few minutes, day and night, since February 2026.
People switch the lane
No timer decides. Operators switch the lane by hand, about 30 times a day. And people have habits.
We learn the habits
From thousands of past switches, the model learned how the bridge looked just before each one. Then it compares the bridge now.
We give a time we can keep
Not an exact minute, because a person decides: a time the switch should come by, nine times in ten. When we know less, the card says so.
We score ourselves
Every forecast is checked against what the lane did. Bad days stay on this page.
Questions
How accurate is the Lions Gate Bridge centre-lane forecast?
Over the last 13 days (28 October to 9 November 2026), the lane had switched by our “expected by” time on 309 of 325 cards checked: 95%. Over the twelve weeks before (3 August to 27 October 2026), with each week forecast only from earlier days, it was 92% of 9,243 checks. Our aim is nine in ten. Before 10 am we fall short (88% in both), so the home card marks a morning forecast “low confidence”.
Does the province say when the centre lane will switch?
No. It publishes approximate weekday times: two lanes southbound from 6:00 to 9:30 am, and mostly two lanes northbound from 3:00 to 7:00 pm. Operators switch the lane by hand, around the clock, “based on traffic, visibility and weather conditions” (Government of British Columbia). On 55 weekdays from 10 August to 9 November 2026, the lane changed direction a median of 27 times a day, and from 3:00 to 7:00 pm it was northbound 65.5% of the time. Is our forecast better than the published times? Not for the first switch of the morning: on those weekdays, 9:30 am and our usual time for that weekday both missed by a median of about 13 minutes. What the published times do not give is the time of the next switch. For the lane right now and for closures, see DriveBC.
Is this an official forecast?
No. The Lions Gate Bridge Museum is independent of the province and of DriveBC. Our forecasts are estimates. For closures and road conditions, DriveBC is the official source.
Does the forecast get better with more data?
A little, and mostly in the first three months. We built the forecast from one, two, three, five and seven to eight months of earlier data and tested each on the same days it had never seen. With one month, its best single guess of the switch time missed by about two minutes more than with five to eight months (on 28 October to 9 November 2026, 10.5 against 8.7 minutes). After three months, more data changed little. The “expected by” time held about as often either way. The record we learn from started in February 2026, so eight months is all there is.
When does the centre lane open in the morning?
On weekdays, at about 6 am, southbound first: it opened between 5:50 and 6:10 am on 141 of 151 weekday mornings from 18 February to 8 November 2026. Weekends are later and less regular: on Saturdays the median was 8:08 am and on Sundays 8:20 am, mostly northbound first. How close our opening time came, from 1 April to 9 November 2026: on 124 weekday mornings, a median miss of 2.1 minutes; on Saturdays 20 minutes (27 mornings) and on Sundays 21.6 (26 mornings).
When does the centre lane first switch on a weekday?
Usually between 9:35 and 9:45 am: that was the median first switch on each weekday from 19 February to 6 November 2026 (Sunday: 9:55 am). Since late October it has often come after 10 am. When the home card says “usually”, it describes past days. It is not a forecast.
The centre lane is closed in the daytime. How long will it be?
Usually a few minutes. Most daytime closures are a switch: the lane closes, then opens the other way. Of 4,499 switches from 19 February to 7 November 2026, half took 3 minutes or less and 99% took less than 10. A closure for an incident or for maintenance is longer: of 42 such closures, half took 30 minutes or more.
What do Clear, Slow, Congested, Heavy and Severe mean?
They are the museum’s five levels for the extra minutes a trip takes over a clear run. Southbound: Clear under 3, Slow 3 to 8, Congested 8 to 15, Heavy 15 to 25, Severe 25 or more. Northbound, on the bridge deck only: under 1, 1 to 2, 2 to 4, 4 to 8, 8 or more. From 18 February to 28 October 2026, a weekday afternoon (3 to 6 pm) southbound was Heavy 43% of the time.
How do you check the forecasts?
Two ways. We run the live forecasting code again on past days, using only what was known at each moment, on days its model had never seen (28 October to 9 November 2026). And for the twelve weeks before, each week was forecast by a version built only from the days before that week. Each forecast is then compared with the lane record of that day. This is not a log of what the site showed: no such log is kept yet. The delay forecast is compared with the sign’s own reading an hour later, and with our own estimate of the delay; neither is an independent measurement.
Can I use these numbers?
Yes, under CC BY 4.0. Download the weekly summary below. It holds summary figures only, not detector readings. Please credit Lions Gate Bridge Museum (lionsgatebridge.org).
All the numbers
| What | Number | Sample |
|---|---|---|
| Of the time, the lane had switched by our “expected by” time | 95% | 28 October to 9 November 2026 · 325 cards checked |
| The same over the twelve weeks before, each week forecast only from earlier days | 92% | 3 August to 27 October 2026 · 9,243 checks |
| Typical miss on when the lane opens on a weekday morning | 2.1 minutes | 1 April to 9 November 2026 · 124 weekday mornings |
| The lane had switched by the “expected by” time, the last 13 days | 95.1% (309 of 325) | 28 October to 9 November 2026 · 325 cases |
| The same, before 10 am | 87.8% | 28 October to 9 November 2026 · 49 cases |
| The same, from 10 am | 96.4% | 28 October to 9 November 2026 · 276 cases |
| Twelve weeks, each week forecast only from earlier days | 91.6% (8,465 of 9,243) | 3 August to 27 October 2026 · 9,243 cases |
| Twelve weeks, before 10 am and from 10 am | 87.7% and 92.9% | 3 August to 27 October 2026 · 9,243 cases |
| The best single guess (not on the card): median miss | 6.8 minutes (always guessing the usual wait: 7.0) | 28 October to 9 November 2026 · 271 cases |
| Weekday morning opening: median miss, and within 10 minutes | 2.1 minutes; 92% | 1 April to 9 November 2026 · 124 cases |
| Saturday and Sunday mornings: median miss | 20.0 and 21.6 minutes | 1 April to 9 November 2026 · 53 cases |
| Southbound delay an hour ahead, against the sign’s reading an hour later: median miss | 6.0 minutes (the sign’s number at the time: 5.0; the usual for that hour: 6.0) | 28 October to 9 November 2026 · 659 cases |
| The same, nine in ten within | 17 minutes (the sign’s number at the time: 20) | 28 October to 9 November 2026 · 659 cases |
| Southbound delay an hour ahead, against our own estimate: median miss | 1.0 minute | 28 October to 9 November 2026 · 651 cases |
| First switch of a weekday: median miss | 13.4 minutes for 9:30 am; 13.5 for our usual time | 10 August to 9 November 2026 · 55 cases |
| Weekday afternoons, which way the lane runs 15 minutes later: right | the published “northbound” 80%; our forecast 72% | 28 October to 9 November 2026 · 317 cases |
| The same, 30 minutes later (the lane often switches twice in half an hour; our forecast gives only the next switch) | the published “northbound” 82%; our forecast 48% | 28 October to 9 November 2026 · 303 cases |
- These are the live code run again on past days, and a rolling check: not a log of what the site showed.
- Neighbouring checks look at the same switch, so the counts overstate the independent evidence. The day is the real unit.
- One road detector our estimates read has been marked by the official page as giving invalid data most of the time since 26 October 2026. Our own checks found its readings no worse when marked, but that is not independent proof.
Download the weekly summary (CSV) Summary figures only, with the number of cases behind each. Please credit Lions Gate Bridge Museum (lionsgatebridge.org).
