Author Topic: [LEL17] Managing the spikes at controls  (Read 33318 times)

CrinklyLion

  • The one with devious, cake-pushing ways....
Re: Managing the spikes at controls
« Reply #50 on: 07 August, 2017, 11:36:24 am »
But in the control dashboard (not public tracking) you could see stuff like how many riders north or south-bound you had had through - so you could know how far through the field you'd got, which was really handy for stock control/ordering.  We did have a little cheer when we realised we could tell the kitchen 'less than a thousand to go!' - didn't tell them it was 999 though.  And the flow charts (can't remember if that was in a different bit of the site?) predicted number of arrivals in the next hour/2 hours/3 hours - obviously their accuracy would be impacted by riders not being scanned into/out of dorms or back out onto the road since all the system could know about a rider is that they had checked in to a given control and were therefore either somewhere in that control or somewhere on the road towards the next one - but they did give us a clue about when the bulge might turn up.

ETA - and there's no way where I was that we could have spared a volunteer to head count consistently throughout the event IMO, and if it isn't done consistently then what's the point?  Bear in mind that our gates were a significant distance from anyone else in a place with atrocious mobile signal - so someone would have had to run up that hill and back down again every half hour while someone else stayed there counting.  Or riders could just check out on the system that was already in place.  Yes, the bag drop thing mucked it up for the last 500ish km for some riders.  I'm sure that can be avoided next time.

CrinklyLion

  • The one with devious, cake-pushing ways....
Re: Managing the spikes at controls
« Reply #51 on: 07 August, 2017, 11:45:18 am »
Oh - and if we'd head counted people going past the desk as exiters, we'd have had about 4000 riders exiting due to some of the faffing yoyos.

Re: Managing the spikes at controls
« Reply #52 on: 07 August, 2017, 12:36:07 pm »
In lieu of mandatory checkouts, a simple exit head count by a volunteer on the gate, forwarded (phone, text or email) to the next control every hour or half-hour was all that was needed to have a good picture of the likely inflow and when and have time to prepare for it.

(Given some of the IT glitches, it may have been useful for each control to have been manually forwarding a heads-in/heads-in count to the next control anyway as a backup anyway)

Great idea - if you have enough volunteers at your control ...

hellymedic

  • Just do it!
Re: Managing the spikes at controls
« Reply #53 on: 07 August, 2017, 01:18:15 pm »
I doubt there would be a sufficient volunteer pool to be able to supply an exit headcounter 24/7 at umpteen controls. It would be BORING, potentially cold and thankless.

Re: Managing the spikes at controls
« Reply #54 on: 07 August, 2017, 01:37:04 pm »
At some point chips/timing-mats or GPS trackers[1] will become cheap enough that they could be considered for a future LEL.

I'm not suggesting they replace the Brevet Card as proof-of-passage (that should continue to be driven by human interaction) but automation can take away the bulk of the work for rider tracking by automating the collection of the raw data.

1. SPOT trackers are expensive and difficult to obtain in the numbers required (transcontinental race have said they had to limit entrants to 300 due to the lack of availability of trackers). GPS trackers with cellular data updates would be good enough for the UK despite the patchy mobile data coverage (I don't think there's any control where there isn't mobile data availability in a 5 mile run in to the control and within 5 miles leaving a control. That would be enough to build some good rider tracking and prediction systems).
"Yes please" said Squirrel "biscuits are our favourite things."

simonp

Re: Managing the spikes at controls
« Reply #55 on: 07 August, 2017, 02:04:05 pm »
I can't remeber anything about Louth. Spalding food was reasonable quality for the main but could have done with more. The lentil soup was a piss take.

Everywhere else that I reached the food was great and the volunteers were fantastic throughout.

hulver

  • I am a mole and I live in a hole.
Re: Managing the spikes at controls
« Reply #56 on: 07 August, 2017, 03:22:03 pm »
We had advance warning of the very first riders at Pocklington showing up early, because I was looking at the #lel2017 tag on Instagram and saw a photo posted from the humber bridge. I warned the kitchen staff, and they put some sandwiches and donuts out just before the first riders turned up.  ;D

I don't think that's scale-able to the whole event though.

mattc

  • n.b. have grown beard since photo taken
    • Didcot Audaxes
Re: Managing the spikes at controls
« Reply #57 on: 07 August, 2017, 04:17:11 pm »
At some point chips/timing-mats or GPS trackers[1] will become cheap enough that they could be considered for a future LEL.

I'm not suggesting they replace the Brevet Card as proof-of-passage (that should continue to be driven by human interaction) but automation can take away the bulk of the work for rider tracking by automating the collection of the raw data.

1. SPOT trackers are expensive and difficult to obtain in the numbers required (transcontinental race have said they had to limit entrants to 300 due to the lack of availability of trackers). GPS trackers with cellular data updates would be good enough for the UK despite the patchy mobile data coverage (I don't think there's any control where there isn't mobile data availability in a 5 mile run in to the control and within 5 miles leaving a control. That would be enough to build some good rider tracking and prediction systems).

I think trackers could be really beneficial even now - we don't need EVERY rider to have one, just a representative sample. Watching the dots flow towards your control would be very useful (and it's a much more intuitive view than the current barcharts).

There will always be quite a few riders _wanting_ to be tracked publically, so that would ease the cost.
Has never ridden RAAM
---------
No.11  Because of the great host of those who dislike the least appearance of "swank " when they travel the roads and lanes. - From Kuklos' 39 Articles

Re: Managing the spikes at controls
« Reply #58 on: 08 August, 2017, 08:28:27 am »
I really don't understand why an event like LEL would use chips and timing mats. It's such a hassle compared to just keying a number into a device at a control. That way you need a few dozen devices (a phone will do it) as opposed to a few dozen readers, thousands of chips and bracelets, plus the devices anyway.

The easiest way to get trackers for an event like LEL is to have something made bespoke - basically a battery attached to a gps chip and a SIM card/SMS chip. But Chinese factories won't get out of bed for fewer than 10,000 devices, so that doesn't work for us either.

The easiest solutions would be to build an app with a very low battery drain, or just give spot trackers to the 5am group. But I'm not sure LEL should really push the whole 5am group any more to the fore. It's beginning to feel inappropriate.

Re: Managing the spikes at controls
« Reply #59 on: 08 August, 2017, 08:59:37 am »
But I'm not sure LEL should really push the whole 5am group any more to the fore. It's beginning to feel inappropriate.

I thought it was just me thinking  that. Please don't let outside pressure turn it into a race or some sort of "challenge ride".
Audax is unique, with it's volunteer ethos and emphasis on finishing not times.

Brakeless

  • Brakeless
Re: Managing the spikes at controls
« Reply #60 on: 08 August, 2017, 09:21:42 am »
But I'm not sure LEL should really push the whole 5am group any more to the fore. It's beginning to feel inappropriate.

I thought it was just me thinking  that. Please don't let outside pressure turn it into a race or some sort of "challenge ride".
Audax is unique, with it's volunteer ethos and emphasis on finishing not times.

Did the earlier group not get through controls quickly though and keep them a little clearer for later arrivals. I thought getting a bunch of riders ' up the road' was a great idea. I agree that quick riders shouldn't be the emphasis of the ride at all but separating them out a bit is a good idea is it not?

Re: Managing the spikes at controls
« Reply #61 on: 08 August, 2017, 09:55:03 am »
Agree entirely with that part of it and the fact that the early starters accept a lower time limit than the hoi polloi, adding to their pressure.
However in every start wave there were mixed abilities. By the time I had attended to a tyre problem on the first leg, I had been passed by the fast riders of the group setting of after me and soon after, the main bunch from that wave arrived as well. 

But in alwyn's post there was a hint(maybe I misinterpreted it) that the fast group was being treated as some sort of elite , somehow better than the rest . There will always be someone faster than me,and I really don't mind the fact that the fastest rider group is assembled on previous experience, but I still maintain that we risk eroding the potential goodwill of volunteers if LEL becomes a pseudo race, even at the front of the ride.
If someone is told that the first rider was through here yesterday ,it can have a negative effect on their morale whereas others just say "So what".
Some of the social media posts seemed to emphasise the riders position not just their progress.

Ride London and many sportives ask riders to state their expected journey time on their entry form and then set starting groups accordingly.
Accepting that LEL is bigger than your average AUK event, is that the way to go?

mmmmartin

  • BPB 1/1: PBP 0/1
    • FNRttC
Re: Managing the spikes at controls
« Reply #62 on: 08 August, 2017, 10:32:16 am »
You are at cross-purposes here. Danial is talking specifically of the invitation-only 5am starters. This is a tiny part of the ride. The creation of the 100 hour group was a brilliant idea designed to get faster riders away and on the road to allow an increase in numbers by 500 riders but not create a bigger bulge at controls. This was a good plan but like all plans, it didn't survive contact with the enemy, namely the riders. Some entered the 100 hour group simply to get a place in the event knowing they weren't going to make it round in time but didn't mind that, others entered but didn't get anywhere near the 100 hours. Hence the increase in the overall DNF figure, which was increased by the 60 per cent DNF in the 100 hour group. Behind that headline figure the DNF rate is about the same as last time and the same as PBP.

If you wanted to lower the DNF figure (but why would you?) the easiest way would be to put the final scan desk at 1,401k, maybe in a rented house or van/pub/tent. This would take nearly three hours off riders' times so many who are three hours OOT will reach that scanning desk within their time. Then they'd pootle back to Loughton at their own speed. You'd simply move the desk and its concomitant volunteers from the reception area at the school with no need for extra resources.
Besides, it wouldn't be audacious if success were guaranteed.

Brakeless

  • Brakeless
Re: Managing the spikes at controls
« Reply #63 on: 08 August, 2017, 10:35:45 am »
I was in the 5am group. I joined it knowing that I certainly wouldn't be the quickest but I was confident I could get round in under 100 hours. My main motivation was that I would be ahead of the bulge and hopefully have quieter and quicker to negaotiate controls, I wasn't the only one in the group thinking this way and I ended up yoyoing riders from the 5am group the whole way round. I didn't feel like we were treated as any different to the rest of the field, I certainly hoped we weren't. I think we did benefit enormously from quiet controls but at the same time I don't think I saw any riders from our start group needing any 'extra ' help at at controls due to fatigue, general unprepardness etc and hopefully we were pretty 'light' on controls. I'd be interested if the view about the early starters is different to my experience from someone looking in rather than out though. If there are similar 100 hour groups next time I think a big emphasis will need to be on this years statistics as getting round in 100 hours was a big ask even for the fittest according to the finishing figures. I have nothing negative to write about this years event, all the volunteers were stars, all the riders I encountered were friendly. The talk of a pop up control between Pocklington and Louth is a great idea, I cycled round Barton on Humber at 6.30am trying to find some breakfast but nothing doing. If there is a hall in Barton it would be perfect.

αdαmsκι

  • Instagram @ucfaaay Strava @ucfaaay
  • Look haggard. It sells.
Re: Managing the spikes at controls
« Reply #64 on: 08 August, 2017, 10:49:28 am »
If you wanted to lower the DNF figure (but why would you?) the easiest way would be to put the final scan desk at 1,401k,

You made a suggestion after Windsor Chester Windsor 2015 of having a finish control at 601 km. If I remember correctly it was explained why that wouldn't work, and the same stands true in this case.

The ride finishes at Loughton, not somewhere in Essex. And as we don't have mandatory routes in the UK how would anyone decide exactly where to place that 1,401 km control?

What on earth am I doing here on this beautiful day?! This is the only life I've got!!

https://tyredandhungry.wordpress.com/

simonp

Re: Managing the spikes at controls
« Reply #65 on: 08 August, 2017, 11:01:20 am »
Was in the 7.30 start. Given I'd survived vedettes on fixed I'd give it a go.

It was a very lonely ride due to the smaller field and the gap between final 100h and first 117h group. I was t going as well as in 2013 either. I only saw one other rider between Pocklington and Thirsk. He'd made about 5h on me by then.

I did have a period between 80h and 90h fields in 2015 where I didn't have anyone to ride with (everyone who caught me was too fast for a while) but I found a nice group overnight to Villaines which made it a lot easier on the head and rode with many different people at various times and enjoyed myself. I did enjoy the company of Alan Parkinson this year. Always good value.

If I'd spent more time on the bike after BCM I'd have had a faster pace but it would still have been on my own. The training I've been doing has been split between rowing and cycling. I've hit 3m power PBs recently but I expect my endurance is not what it was - I've not finished a 1000km+ event since last PBP. Immediate future will be about rowing though I might drag my sorry carcass around a DIY 400k later this month just to get the 10th SR monkey off my back.

Re: Managing the spikes at controls
« Reply #66 on: 08 August, 2017, 11:06:20 am »
Some entered the 100 hour group simply to get a place in the event knowing they weren't going to make it round in time but didn't mind that, others entered but didn't get anywhere near the 100 hours. Hence the increase in the overall DNF figure, which was increased by the 60 per cent DNF in the 100 hour group. Behind that headline figure the DNF rate is about the same as last time and the same as PBP.

I know alwyn is against qualifiying rides but maybe they should be part of the deal for 100h groups?

Brakeless

  • Brakeless
Re: Managing the spikes at controls
« Reply #67 on: 08 August, 2017, 11:38:09 am »
Some entered the 100 hour group simply to get a place in the event knowing they weren't going to make it round in time but didn't mind that, others entered but didn't get anywhere near the 100 hours. Hence the increase in the overall DNF figure, which was increased by the 60 per cent DNF in the 100 hour group. Behind that headline figure the DNF rate is about the same as last time and the same as PBP.

I know alwyn is against qualifiying rides but maybe they should be part of the deal for 100h groups?

I think you had to have previous 'form' to be in the 5am group. I had a sub 100 time for 2013 and an OK time for Mille Cymru which I assumed is why I got into the 5am group. Alwyn can answer better than me.

Re: Managing the spikes at controls
« Reply #68 on: 08 August, 2017, 11:55:22 am »
Some entered the 100 hour group simply to get a place in the event knowing they weren't going to make it round in time but didn't mind that, others entered but didn't get anywhere near the 100 hours. Hence the increase in the overall DNF figure, which was increased by the 60 per cent DNF in the 100 hour group. Behind that headline figure the DNF rate is about the same as last time and the same as PBP.

I know alwyn is against qualifiying rides but maybe they should be part of the deal for 100h groups?

I think you had to have previous 'form' to be in the 5am group. I had a sub 100 time for 2013 and an OK time for Mille Cymru which I assumed is why I got into the 5am group. Alwyn can answer better than me.

Later posts from alwyn suggest that either previous form, or sufficient self-confidence to blag it (either convincingly or amusingly) were enough for a 5am start.

When the dust settles and the organising team have had a few more nights' sleep, I'd quite like to see a fairly detailed set of figure around DNFs (notably those 100h riders who were HD by virtue of finishing >100h but <117h) - I'm sure that they will come in due time, albeit not necessarily made fully public.

Re: Managing the spikes at controls
« Reply #69 on: 08 August, 2017, 11:59:15 am »
If you wanted to lower the DNF figure (but why would you?) the easiest way would be to put the final scan desk at 1,401k,

You made a suggestion after Windsor Chester Windsor 2015 of having a finish control at 601 km. If I remember correctly it was explained why that wouldn't work, and the same stands true in this case.

The ride finishes at Loughton, not somewhere in Essex. And as we don't have mandatory routes in the UK how would anyone decide exactly where to place that 1,401 km control?




Then a case of trimming the distance from way over to say 1405k
There is 4 yrs in which to achieve that.

Re: Managing the spikes at controls
« Reply #70 on: 08 August, 2017, 12:06:45 pm »
But WHY?  Why is it necessary to have more riders finishing?  It is an Audax - a very hard one.  The responsibility for getting home is the rider's, not the organisers'.  Certainly volunteers do all they can to help but paying the entry fee doesn't guarantee a finish.  The correct way to decrease the DNF numbers is for more riders to finish.

Peter

Fidgetbuzz

  • L sp MOON. 1st R sp MARS . At X SO sp STARS
Re: Managing the spikes at controls
« Reply #71 on: 08 August, 2017, 12:09:40 pm »

[/quote]
Then a case of trimming the distance from way over to say 1405k
There is 4 yrs in which to achieve that.
[/quote]

Very very difficult to achieve .. given location of available controls .. absolute requirement for a safe route . Distance was just 2.5% over .. well under permitted amount.
I was an accountant until I discovered Audax !!

Re: Managing the spikes at controls
« Reply #72 on: 08 August, 2017, 12:11:56 pm »
I really don't understand why an event like LEL would use chips and timing mats. It's such a hassle compared to just keying a number into a device at a control. That way you need a few dozen devices (a phone will do it) as opposed to a few dozen readers, thousands of chips and bracelets, plus the devices anyway.

I agree that tracking arrival times is easily done as it is now, but the question was about improving estimation of rider arrivals to give the kitchens a chance to ramp up in time and not waste food by having it ready when not enough riders are in. The implication is that rider departure time is more useful which is trickier to reliably obtain from controls.

Mats/chips may not be specifically required, it was just an example of what technology exists now that could help automate the rider tracking (and specifically control exit which currently requires cooperation from the riders).

Having a good idea of where riders are (without GPS trackers) requires:-
a) Details of departure times
b) A good model (basing things on a riders previous speeds between controls, group speeds between controls to account for differing weather/terrain, etc)

Having tracked linger time for a large percentage of riders on LEL2017 it may be possible to create a reasonable model for future LELs based on this data that just requires control arrival times (which are collected like they are now with no extra technology required). From there you might be able to extrapolate to have a reasonable prediction for the bulk of riders.

I'd be interested in the anonymised dataset (including drop bag checkout/checkin) to see what modelling could be done. The idea would be to analyse the complete set to extract behavioural patterns and identify different rider types (stamp and go riders, touring riders, longer stops at bag drop locations. riders up against time limits) and temporal patterns (average stop times at different times of day, average speed drops due to fatigue/terrain, etc) to build up the model. Then replay the minimal dataset (bag drop locations, rider arrival times) in chronological order in an attempt to predict the arrival times of riders at subsequent controls. I could also see how well a model worked against the LEL2013 dataset.

The easiest way to get trackers for an event like LEL is to have something made bespoke - basically a battery attached to a gps chip and a SIM card/SMS chip. But Chinese factories won't get out of bed for fewer than 10,000 devices, so that doesn't work for us either.

There are plenty of trackers that exist that are battery+GPS+SIM, but I'm not sure how easy it would be to get 1500+ of them (including PAYG data only SIM cards). Also the tracker is more expensive if it can't be used for future rides. It has to last 7 or so days on a single battery and the battery has to be replaceable. The service it reports to must have some form of API for bulk API collection, and it has to be reliable (and waterproof)!

The easiest solutions would be to build an app with a very low battery drain, or just give spot trackers to the 5am group.

Unless a mobile is powered/charged by something along the way then not running in airplane mode between controls will drain the battery far quicker than any app. In areas of low/no signal the phone boosts transmission power in the hope of contacting a base station, this is the major battery drain.

It would have to be an app that was capable of periodically disabling airplane mode, waiting for a possible signal, sending position update and then settling back into airplane mode. AIUI airplane mode cannot be disabled by an app.

Even then we have many foreign riders that may find roaming data access on their phone prohibitively expensive and swapping to a local SIM stops them from being in contact with their friends/family.

For those that do power their phone on the go, given the numerous existing apps I can see some service in the future which gathers the data from the various services (Garmin Connect, Strava, FindMyPhone, etc), normalises it, and presents it over one consistent API. Then riders could register with this service providing access to their data to LEL and it is taken from there...
"Yes please" said Squirrel "biscuits are our favourite things."

Re: Managing the spikes at controls
« Reply #73 on: 08 August, 2017, 12:21:38 pm »
Are we losing the point of what Audax is, a (largely)unsupported long distance bike ride over a given time and distance, with controls as set points? Whether it's 1400, 1405km of whatever, no one is forced to enter if they don't like the arrangements?

By the way, for tracking I use the track my iPhone app, works like a charm and keeps Mrs P happy. Not perfect I know but works for us

A

hulver

  • I am a mole and I live in a hole.
Re: Managing the spikes at controls
« Reply #74 on: 08 August, 2017, 12:23:15 pm »

I'd be interested in the anonymised dataset (including drop bag checkout/checkin) to see what modelling could be done. The idea would be to analyse the complete set to extract behavioural patterns and identify different rider types (stamp and go riders, touring riders, longer stops at bag drop locations. riders up against time limits) and temporal patterns (average stop times at different times of day, average speed drops due to fatigue/terrain, etc) to build up the model. Then replay the minimal dataset (bag drop locations, rider arrival times) in chronological order in an attempt to predict the arrival times of riders at subsequent controls. I could also see how well a model worked against the LEL2013 dataset.


I don't think drop bag checkout/checkin would be much use. Certainly at Pocklington all we did was scan bags when they arrived on the van. We didn't scan them when giving them to riders, and after problems with controls scanning out bags further up the route (it was resetting the time in hand figure on the tracking display), we didn't scan them out of the control either.