2022 rider data

[alwyn]

I've finally got round to cleaning up and anonymising the 2022 rider data, and it's now available to anyone who wants it:

https://www.dropbox.com/s/cjm9gy089zdtwiv/LEL2022%20anonymous%20rider%20data.csv?dl=0

Feel free to use this how you see fit. However if you do something cool with the data (Ivan) then please share it here or on our facebook group.

[Wycombewheeler]

Looking through that data, it seems there are about 890 finishers (number of people with a start time and a finish time inside the time limit)
although this is muddied somewhat, as there are people in those 890 with other controls missing, and also there are (bizarrely) people with no start time who are not counted in the 890.

There are some strange records indeed, including one with a series of stops on the way north, then nothing, then 3 days later a finish time

There were approximately 140 people who benefitted from the additional time allowance (start to finish between 125 hours and 128 hours) and a further 49 people with a total time of over 128 hours

1511 starters, so the additional time moved the completion rate from ~50% to ~60%

[The French Tandem]

There are some strange records indeed, including one with a series of stops on the way north, then nothing, then 3 days later a finish time

Quite a lot of riders decided to pack in Barnard Castle northbound. Among  them, some decided to just ride their bicycle back to Debden at their own pace, knowing that they would have plenty of time before the closure of the Debden control. Have some of them been mistakenly considered as finishers? This might explain some of the strange records.

A

[alwyn]

There are some strange records indeed, including one with a series of stops on the way north, then nothing, then 3 days later a finish time

Quite a lot of riders decided to pack in Barnard Castle northbound. Among  them, some decided to just ride their bicycle back to Debden at their own pace, knowing that they would have plenty of time before the closure of the Debden control. Have some of them been mistakenly considered as finishers? This might explain some of the strange records.

A

This is operational data to help us run the event and analyse rider flows afterwards. We also use it as backup in case there is a gap on the brevet. However the brevet is the principle proof of passage, and they will be checked three times in total to make sure there are no anomalies.

AUK are currently carrying out check 3; our work is done.

[Wycombewheeler]

Quite a lot of riders decided to pack in Barnard Castle northbound. Among  them, some decided to just ride their bicycle back to Debden at their own pace, knowing that they would have plenty of time before the closure of the Debden control. Have some of them been mistakenly considered as finishers? This might explain some of the strange records.

A

And I would have done likewise if I had packed before reaching Dunfermline, but I wouldn't have presented my card for stamping at the finish, although maybe they would have preferred that as a way of knowing which people are no longer still out on the course somewhere.


In my rough count people with missing records (other than start/finish) are included in the 890, some of these may be errors who do have completed brevet cards. It's also possible that some people without a start or finish time recorded but all other records intact will turn out to be finishers with good brevet cards and a computer anomaly, I expect the number of finishers will be close to 890, once everything is considered. I might be wrong and there might be fewer.

[Flâneur]

Interesting what happens about these null start times. Cards were stamped at the start but no written times added, presumably cards wouldn’t have been stamped for riders before their allocated start time.

Anyway I’ve found my record and it’s all complete so all good 

[sibi]

And I would have done likewise if I had packed before reaching Dunfermline, but I wouldn't have presented my card for stamping at the finish, although maybe they would have preferred that as a way of knowing which people are no longer still out on the course somewhere.

After my knee decided that I should pack at Malton I asked on the way back if I should have my card stamped. They said yes because the people following me on the net would know where I am and did not get lost. However I skipped some controls like Louth as I took the direct way from Hessle to Boston.

[alwyn]

Interesting what happens about these null start times. Cards were stamped at the start but no written times added, presumably cards wouldn’t have been stamped for riders before their allocated start time.

The rider's start time was pre-printed, so no need to write it again on the card. A null start time on the online tracking is irrelevant because the card is the primary record.

During our side of validation we used the online record to fill in any gaps on the brevet card. After this we had about a dozen queries, most of which we could solve right away. After that we were left with four riders we had to contact to get a explanation from them about the gap. Of these, we've left two at the mercy of AUK to make a decision, and two we decided to not put forward for validation.

[The French Tandem]

During our side of validation we used the online record to fill in any gaps on the brevet card. After this we had about a dozen queries, most of which we could solve right away. After that we were left with four riders we had to contact to get a explanation from them about the gap. Of these, we've left two at the mercy of AUK to make a decision, and two we decided to not put forward for validation.

Hats off to you for all this hidden work! Events like LEL could not possibly exist without some people who accept to spend hours and days for such  kind of tedious work.

A

[nicktick]

I'm not surprised by some missing controls in the data - my online tracking missed out Boston on the way home (much to the consternation of my father who was following a lot more closely than my somewhat dismissively unconcerned wife!), but I definitely got the card stamped, and the guy on the computer made the right noises and moved his fingers a bit when I handed him the card & mumbled something vaguely coherent.
So, based on a sample of one, there were errors in 100% of the computer based records...

[grams]

The process on the computer meant you searched for a rider, brought up their page, verbally confirmed their name, then clicked the Check In button. It was *very* easy to forget the last step. Usually I caught it when the next rider arrived, but not always.

Possibly something for the IT people to think about for next time.

[Wycombewheeler]

I'm not surprised by some missing controls in the data - my online tracking missed out Boston on the way home (much to the consternation of my father who was following a lot more closely than my somewhat dismissively unconcerned wife!), but I definitely got the card stamped, and the guy on the computer made the right noises and moved his fingers a bit when I handed him the card & mumbled something vaguely coherent.
So, based on a sample of one, there were errors in 100% of the computer based records...

Still better than my revisited on pbp, where must of the controls failed to see my timing chip.

Response of the volunteers,  don't forget to get your card stamped" and a grave expression.

[abc123]

I have been doing a fairly deep dive into the LEL data and using it as an opportunity to try and learn some new stuff. I have posted a few examples of charts created from the data below (without context) but this is not the best site for displaying images. If you are interested, please visit London Edinburgh London in Charts where there are more images in higher resolution, and I have done a full write up about what the charts represent and how they were created.

[TomAplomb]

Fantastic work abc123! Those are some great figures. What tools/language did you use to create them?

[Wycombewheeler]

Love the animation, was hoping to be able to slow it down on the linked website, but it wasn't there.

Pretty sure this shows I was bumping up against the rear of the bulge after my late start, until I decided to just ease up a bit, and accept finishing in 125 hours is fine.

[abc123]

Love the animation, was hoping to be able to slow it down on the linked website, but it wasn't there.

Pretty sure this shows I was bumping up against the rear of the bulge after my late start, until I decided to just ease up a bit, and accept finishing in 125 hours is fine.

The higher frame rate version on the site is a very large file, so it might take a while to load when you open the page. I am working on trying to reduce it to a more reasonable size.

Update: I've replaced the animation on the website with a smaller file (and also updated the version above) and there's also a copy on YouTube here:
https://youtu.be/vB9h9TPQEWs

The animation does highlight the bulge on the northward leg, and how thin it becomes stretched after the turn. At one point on the return leg there is a chain of riders spanning almost the entire distance between Dunfermline and London.

[abc123]

Fantastic work abc123! Those are some great figures. What tools/language did you use to create them?

All the charts were created using Python modules: Pandas and GeoPandas to engineer the data and Matplotlib/Seaborn for plotting.

[Ajax Bay]

Bravo! No, I'll go further: BZ!

[TomAplomb]

Fantastic work abc123! Those are some great figures. What tools/language did you use to create them?

All the charts were created using Python modules: Pandas and GeoPandas to engineer the data and Matplotlib/Seaborn for plotting.

Nice. Do you have a link to a GitHub repo? No worries if not - I'm just curious.

[Zed43]

Interesting analysis 

I found that I could get a reasonable estimate of my real average speed by starting with a base of 80% of my Stage 1 speed and applying simple weighting factors to reduce the speed more for the harder and later stages.

I would love to see what factors you used / the formula you ended up with.

It spurred me to have another look at the GPS data of my rides in 2017 and 2022

	2017	2022
total distance	1456 km	1539 km
total climb	9272	11351
start time	12:45	11:15
time limit	116h 40m	128h 20m
total time	101h 50m	118h 30m
time in controls	35h 5m (34,5%)	41h 40m (35,2%)
time moving	63h 35m (62,4%)	72h 5m (60,8%)
time stopped	3h 10m	4h 45m
time riding 23:00-06:00	18h 25m	20h 0m
moving average	22.9 kph	21.4 kph
normalized power	137 W	129 W
average temperature	15.6o	19.8o

Time stopped (between controls) are generally very short and comparable between editions, though in 2022 I had a few longer ones that made the difference:

• 15m secret control
• 10m toilet break High Force
• 30m fix mechanical in Edinburgh
• 40m cafe between Brampton / Barnard Castle
• 20m ice cream between Louth / Boston (dealing with heatwave)

2022 was pretty much solo, with some "social" group riding; in 2017 I was in some trains traversing the Fens (both North and South bound) that got my average speed up.

[edit: fixed formatting]

[abc123]

I received the latest edition of Arrivée this week, and I was struck by an anecdotal remark recounted in Nick Tickner's article about his experience of LEL 2022:
"anyone who needs to sleep at Louth probably isn't going to make it".
Is this true? What does the data show? For the following charts, I looked for the first control where each rider stopped for at least 3 hours, and assumed this location was their first sleep stop.

It looks like quite a number of riders who stopped at Louth actually went on to finish successfully in time, but the success rate clearly improves for riders who made it further north on the first day.
Distance travelled is only part of the story, and the second chart presents the same data in terms of how long the first day was for riders in the two groups. The distributions are very similar, showing that both groups of riders actually put in similar shifts on the first day.
More info and more charts here : LEL in charts

[Wycombewheeler]

How do you know how long people stopped for?

If course the Louth comment is probably handed down wisdom, from a time when the last start grotto was not as late as it was this time

[abc123]

There's more detail on the linked site above, but in simple terms I used the stage 1 average speed of each rider to estimate average speeds for every rider on all the subsequent stages. The estimated average speeds can be used to calculate an estimated riding time for each stage, and subtracting this time from the recorded arrival time intervals gives you the time spent at controls.
With the estimated control stopping times, you can calculate the total number of riders at a control at any point in time, but it also allows you to do other stuff, like the animation I posted previously, or identifying where riders stopped to sleep.

[Ajax Bay]

How do you know how long people stopped for?

If course the Louth comment is probably handed down wisdom, from a time when the last start grotto was not as late as it was this time

Pretty sure the 'if you stop in Louth you'll be up against it' message was broadcast in 2017, when the last start was at 1600 (later than this year). But there was some confusion on the this message as 'they' also said 'no problem if you get out of time on the early controls, provided you make it up later'.
A cohort took this advice, even if they didn't have a mid/late afternoon start, and rested at Louth (in the dark hours). They were just about 'in time' on arrival in Edinburgh, but the forecast half gales on Thursday either encouraged DNFing at (say) Bampton (going south) or later, or actually dramatically slowed the 'low available power' element of the 'bulge' on the fens.

[Wycombewheeler]

There's more detail on the linked site above, but in simple terms I used the stage 1 average speed of each rider to estimate average speeds for every rider on all the subsequent stages. The estimated average speeds can be used to calculate an estimated riding time for each stage, and subtracting this time from the recorded arrival time intervals gives you the time spent at controls.
With the estimated control stopping times, you can calculate the total number of riders at a control at any point in time, but it also allows you to do other stuff, like the animation I posted previously, or identifying where riders stopped to sleep.

Impressive amount of data crunching