Clumps are now clusters

After a very active discussion on the Ride Every Tile Strava club (first one I’ve seen with 100 comments actually), it is great to see we have a solution to the “impossible-to-visit squares” problem. And that solution is “max cluster”. The idea of a maximum cluster is very similar to the idea of “max clump” described earlier on this blog . In fact the algorithm is the same, it is just the name that has improved. Although I am biased because I played a role in their creation, I do think clumps are a reasonable solution to the impossible squares problem because

  • No matter how many impossible squares you have, you can join the max clump game. In the worst case you pay a “fine” of 5 points for not visiting a square, and if there are a number of unvisitables next to each other, the cost per square works out less than 5.
  • The cluster metric adapts to whatever topology is in your area. If there’s a coastline that does go north/south or east/west, you can make a pretty pattern by visiting everywhere along the coasts.
  • The max square score is a very slow-growing score (like Eddington number it grows with the square root of effort) so it requires long-term patience to increase. The max cluster on the other hand brings more instant gratification.
  • Even if max square is working out nicely for you, it is still another thing to look at VeloViewer.
  • As a bonus, the map UI in VeloViewer has become much more sophisticated (details at Ben’s post here). Amongst other things, you can see cluster and visited-squares colour the same if you want to, recovering the old UI.

cluster_prod

Using the new flexible UI to show the damage the Squarpocalypse has done!

A further bonus is that some of Ben’s legions of followers on Twitter/Facebook have been attracted Ride Every Tile, and we now have over 180 members (from about 40 two weeks ago). And with that influx, there’s been some great new ideas. In particular “Half-Fast” Mike who rides in and around Tokyo has shared a great idea for getting squares to be displayed on your Garmin Edge as you are out and about – increasing the chances that you will actually bag all the squares you hope to bag on a ride – full details here

Finally, the race to recover lost squares following the Squarpocalypse has heated up. I dragged myself out of bed (again) to do a very similar route to the last time I dragged myself out of bed (ride report here for that one ) except this time I would bag the four squares that I had missed. Despite leaving home 12 minutes later than target (4.42am) I reached work at 8.04am, four minutes later than target – traffic lights aren’t always more red than you think they ought to be. This got me back to 32×32. At the time of writing this actually third on the leaderboard, but the situation is very fluid.

As always, there was something unexpectedly pretty or interesting on a squares ride… I had to loop round the South Mimms services to bag a square just outside Potters Bar. Service stations are normally pretty awful places (Sartre wasn’t talking about services when he said “L’enfer, c’est les autres” but the phrase always comes to mind) and South Mimms, at the junction of the M1 and M25 is pretty typical so it was lovely to discover that the village of South Mymms has many centuries of history and a church that dates to 1136. As usual I regret not stopping to get photos – some at http://stgiles-stmargarets.co.uk/st-giles-history

Strava link for the ride: https://www.strava.com/activities/1045242141

 

 

 

 

Clump examples

[19th June 2017: Updated to add: Ben VeloViewer has now done a better proof-of-concept – check out your own VeloViewer map  for an example! Also see Ben’s Facebook post for more details and for your chance to vote on the name for this new thing. At the moment it looks like “cluster” will win out over “clump”.]

The pictures below show some example “max clumps” for real data. The “max clumps” shown here are calculated using the following rules:

  •  A square is in a clump if
    • You have visited it
    • You have visited all four squares that are adjacent to that square.
  • If two adjacent squares are both in a clump they are in the “same clump”
  • Divide up all your visited squares into clumps. The clump with the most squares in it is called the max clump.

These rules are not set in stone or sent by God or anything else that. They are just some choices that seem to create some nice patterns.

(more commentary follows below the pictures)

clump_pete_bartlett

My max clump

clump_eric_nichols

The max clump of the rider who was North American leader before “Squarepocalypse”

clump_phill_cloke

The max clump of the rider who was UK leader before “Squarepocalypse”

I quite like the middle map – it clearly shows that the cyclist in question has really been up and down everywhere on the New England coast.

All three maps feature lots of cross shapes. These are due to me generating the data in the days immediately after the squarepocalypse. If there is a missing square then not only is that square missing from your clump, but also the four squares adjacent to it aren’t included either This has the effect that a missing square has slightly multiplied effect on your “clump score” – meaning that you should still try hard to visit every square. But if there are truly unreachable squares, then you aren’t crippled from playing the squares game – you just get a 5 point penalty effectively.

Feedback welcome at the Ride Every Tile Strava club .

An alternative measure to “max square”

There has been a discussion about how to take account of “inaccessible squares” in the max square challenge ( https://www.strava.com/clubs/279168/posts/582450 ).

Here is what I wrote so far on this. Plan to expand this with more examples of the clump idea.

“I agree with Nils that it raises the prospect of endless debate about what counts as inaccessible. E.g. someone on Facebook already posed the question about dual carriageways – certainly many people don’t like riding on them for good reasons, but they are not inaccessible. It just adds extra motivation for a 4am Sunday ride.

“Similarly there are very-hard-to-visit places – power stations, large factories, some military areas. I have the germ of an idea that we could set up something like the Association of British Tile Cyclists (or something similar grand, or indeed for Belgium). This could be an “official” entity that can make access requests which may be more successful than individual requests. It would also be an way of setting up occasional group rides which have a stated goal of tiling a particular area (so frustrating when you do a sportive that misses a few out in the middle…. 🙂 )”

“Having said all that, there are going to be areas where even the most determined can’t get to – there are military areas where even the military don’t go – e.g. firing ranges of certain types have a build up of unexploded unordances. You’re not going there, even on an open day. So, the idea of some sort of different metric does have some appeal. But those proposed on Facebook, e.g. “98% of 40×40″, are also unsatisfying. It is difficult to have a leaderboard comparing 100% of 38×38 versus 98% of 40×40. And we all like a leaderboard :-)”

“So I’ve been fumbling around for an alternate metric. Of course one possibility is just “total tiles” as already reported by VeloViewer. Another possibility, is to define a “clump”. A square is in your clump if you been to that square and to three of the four adjacent squares. Then measure how many squares are in your clump. This clump measure rewards people like Phill and Eric who have been to huge swathes of their area, but which aren’t square-shaped. It is also quite robust against a single hole due to an inccessible. But it is doesn’t reward a single track of squares. E.g. if you just cycled from London to Paris you clump size would be near zero as all you have is a long streaky path – you haven’t been “everywhere” in that area. 

Squarpocalypse!

 

Squartasrophe! Squarpocalypse! Squarmageddon!

Yes the date of Wednesday 14th June 2017 will live long in the memory of avid square-fillers. I was casually doing my daily VeloViewer refresh when a glance at my “max square” field on the summary had me choking on the contents of my bidon. Having rather smugly expected to see my recently-achieved “50×50” shining back at me, the colour fell from my face at the sight of a mere “20×20”. Surely some mistake??

I frantically switched to the map view on the activities tab, switched on Explorer mode and zoomed in. There it was in red and slightly paler red… my square was a shriveled up ghost of its former self. Dotted amongst the old, lost square dozens of squares were now showing as unvisited. I zoomed right into them and, sure enough,  none of the ride lines actually DID go into those squares. But the day before they had shown as visited… so.. had I been there or not? What was going on?

In truth Ben Lowe, the magnificent impresario who makes VeloViewer happen, had warned us that the squarpocalypse was coming in this RideEveryTile post (further warnings in this Facebook post ).

Previously the algorithm for deciding what squares you have been in had to cope with the fact that it was only using summary data from your Strava data. This led to the possibility of squares being visited not being counted (e.g. if a road was a bit bendy, nipped into a square and out again, then the summary straight line might not capture this). Now people don’t like missing out on hard-earned squares, so Ben worked-around the issue by allowing a “buffer” of 100m around the squares. But this mean instead of false negatives (visited squares showing as unvisited) it allowed the possibility of false positives (unvisited squares showing as visited just because you’d been within 100 metres of them).

Now Ben has been able to implement a more accurate algorithm that can truly determine whether a particular ride goes through a particular square (I think by looking at all the Strava data, not just the summary). So there is no need for the buffer anymore. So Ben removed it.

And all us tilers discovered how much we had (almost certainly accidentally) been relying on the buffer to grab us a few squares!

[19th June 2017: Edited to add: The above makes it sounds like Ben forced this change upon the tiling community – this is not true – he did get the express 100% approval before going ahead. Also Ben has now put out his own blog post on the new algorithm – his post is sure to be far more accurate on the details than this one]

All is not lost

It is possible to recover some squares without even leaving the house!

Though the buffer is gone, all your previously uploaded rides still have the squares data associated with them that was generated using the old summary-data-only algorithm. This means you can end up with situations as shown in the image below:

VisitedOrNot

The bottom left square is shown as unvisited, but you can tell from the ride track which road I must have been on, and that road takes me through the bottom left square. Unfortunately the old algorithm didn’t capture this. However help is at hand. Simply double-click on the red line to load the details of that ride. Whilst you reminisce about what a lovely ride that was, behind the scenes the data for that ride is being refreshed to use the new accurate algorithm rather than the old. If we now go back to the map page and force-refresh the page (I used Ctrl+F5 in Chrome), we find the square is now showing as visited:

Visited

Of the 2500 squares that previously made up my 50×50 square, 58 of them were showing as unvisited without the buffer. By selectively refreshing a few rides to use the new algorithm, this has come down to 45.

It’s not really the apocalypse is it?

Whilst most of who’ve been building up our max square for many months will surely have some momentary disappointment at it all seemingly disappearing, in truth it is almost the apocalypse – it is a great thing. Here presented to us on a plate is a copper-bottomed excuse to go out and visit more squares. Properly This Time.

Addendum

As dawn broke the day after the apocalypse, I got right on with doing them Properly This Time. By extending my commute by a mere 10 miles, I was able to pick up a square that I’d never been to before (Strava link ; the new square is the little nubbin close to A3/M25 junction)

CS9 : Part 3

CS9 was mentioned tangentially in an email from City Hall following  new(ish) Walking and Cycling Commissioner Will Norman’s interview in the Guardian today. The email (quoted more fully on road.cc ) says “Consultation on Cycle Superhighway 9 (the route from Kensington Olympia to Hounslow) will take place in the summer (including tackling Chiswick Roundabout / Kew Bridge Junction)”.  With Hammersmith gyratory plans now post-consultation ( my notes are here ) it is good to know that the Kew Bridge to A4 run, the next most crazy bit of the A315 commute, will also eventually be tackled.

 

CS11: My notes

TfL is still dithering on whether to go ahead and reduce rat-running in Regent’s Park by closing some of the gates to motor traffic some of the time.

Sadiq Khan is on the record a number of times of saying he will “learn the lessons from earlier schemes” – which seems to delay and delay whilst he tries to figure out how to please everyone.

The face of the opposition to CS11 is “two swimming pools” Jessica Learmond-Criqui. (Articles about the swimming pools issue e.g. here, here  and here tend to give her the name of her husband – i.e. Jessica Stokel). Every anti-CS11 stunt, such as claiming in the The Stage that the eldery won’t be able to go to the theatre because bus route 13 will now stop at Victoria rather than Aldwych to dressing up school children in surgeons’ masks to rage against polluting effects of cycle lanes has been organised by this one person.

Should TfL really be held to ransom by the noisy antics of a single well-connected and well-funded NIMBY?

[Some pro-CS11 folk claim in comments to various news articles that Ms Learmond-Criqui herself drives around the “narrow streets” of Hampstead in a Range Rover or similar “Chelsea Tractor”. I have seen no evidence of this.]

 

Strava and Apple Health bugs

TLDR: if Strava suddenly thinks you’ve changed gender, check your Apple Health settings.

A couple of days ago I got a few of my rides flagged on Strava. It was slightly mystifying as the roads were fairly ordinary rides with no GPS glitches or anything like that to give me virtual badges beyond what I’d earned.
Then I uploaded a ride yesterday and was surprised to take a bunch of top 10s! Wow I must be in unbelievably good February form!! And then I noticed I had picked up a QOM. Yes with a ‘Q’. Checking the settings I saw that Strava did indeed think I was female. My suspicions turned to the japers at work – they know I’m permanently logged into Strava and spend too much time route planning of post-morteming. 

Then today I got an email stating that my QOM on another segment had been “stolen”. Back to the settings, this time on the mobile and I see that every time I try to change to male, it automatically switches back a couple of seconds later to the fairer sex.

Lightbulb moment – a week or so ago I had connected Apple Health to Strava and never having set the former up properly it thought I was female and moreover took precedence over the correct setting in Strava. Fixed now.

But as of today at least, there is a stupid off-by-one bug across Strava and Apple Health so that if I set my Date of Birtu correctly in Health, then it appears as one day too early in Strava… Fortunately this will only effect one day every five years when I’m in the wrong age category a much less disruptive then the various QoMs I’d been advertently pinching – apologies.