Freezing Saddles Heat Maps

[chris_s]

@lordofthemark 134376 wrote:

City of Alexandria dispatcher – See this road (uphill near lord of the mark’s home)? The cyclists on it have consistently been riding under 10MPH. Go clear it.

City of Alexandria employee, on arrival – Whiskey Tango Foxtrot – it is completely clear!

That sort of thing would definitely be a problem. It would also have trouble with congestion, stop lights, etc. It’d be quite good at telling where the trails and such are definitely clear, but it’d have some real issues telling the difference between “not clear” and “people aren’t biking there or are biking there slowly for some unrelated reason”.

You could up the accuracy quite a lot by only comparing to previous ride through that same area by that same person, rather than looking at specific speeds. So things would be based off “this guy went through here 90% slower than he usually does” rather than “this guy went through here very slowly” but then that gets thrown off by people changing their routes due to conditions and requires a lot more people to be submitting rides in order for you to have the critical mass of data necessary.

[Mikey]

What you could probably do is determine the baseline average speed for that stretch of path, road, etc, then track the last 24 hrs departure from baseline. And then color code kind of like the traffic widget on google maps. i.e. Blue – faster than normal. Green 0-5% slower Yellow 5-10% slower, or something like that.

[lordofthemark]

@chris_s 134378 wrote:

That sort of thing would definitely be a problem. It would also have trouble with congestion, stop lights, etc. It’d be quite good at telling where the trails and such are definitely clear, but it’d have some real issues telling the difference between “not clear” and “people aren’t biking there or are biking there slowly for some unrelated reason”.

You could up the accuracy quite a lot by only comparing to previous ride through that same area by that same person, rather than looking at specific speeds. So things would be based off “this guy went through here 90% slower than he usually does” rather than “this guy went through here very slowly” but then that gets thrown off by people changing their routes due to conditions and requires a lot more people to be submitting rides in order for you to have the critical mass of data necessary.

You would definitely need multiple riders. Otherwise it becomes “lordofthemark normally rides that segment at 11MPH, but this week he rode at 9MPH, something is wrong” – “Sure something is wrong, his air pump broke, and he has been riding with underinflated tires” For other people it could be the carbon fiber road bike is in the shop, and they are now riding on the MTB, or whatever.

[S. Arlington Observer]

Also, note that the time stamp is not up to date in real time. (Not sure how it updates.) As of 2 p.m. it does not show my morning ride this a.m. but does show yesterday afternoon’s evening commute. It is great and I love it. But it isn’t actually in real time.

[chris_s]

@S. Arlington Observer 134383 wrote:

Also, note that the time stamp is not up to date in real time. (Not sure how it updates.) As of 2 p.m. it does not show my morning ride this a.m. but does show yesterday afternoon’s evening commute. It is great and I love it. But it isn’t actually in real time.

@chris_s 134289 wrote:

All-time data updates once a day in the middle of the night.
Weekly data updates once a day in the middle of the night.
Last 24 hours data updates hourly on the hour.

There may also be additional lag between when you upload to strava and when hozn processes the data for me to access.

[chris_s]

@chris_s 134384 wrote:

There may also be additional lag between when you upload to strava and when hozn processes the data for me to access.

I went ahead and double-checked, and my auto-updater has been running as expected – it is updating the 24 hour data on-the-hour as expected. Was it the 24 hour data that was stale or are you looking at one of the longer period that update less frequently?

[americancyclo]

@chris_s 134386 wrote:

I went ahead and double-checked, and my auto-updater has been running as expected – it is updating the 24 hour data on-the-hour as expected. Was it the 24 hour data that was stale or are you looking at one of the longer period that update less frequently?

Interesting. Looking at my team’s 24 hour map at 3:11pm, It shows my ride yesterday at 5:12PM (phone app) but not the 4:06PM or this morning’s ride at 7:03AM. (garmin recordings)

[chris_s]

@americancyclo 134387 wrote:

Interesting. Looking at my team’s 24 hour map at 3:11pm, It shows my ride yesterday at 5:12PM (phone app) but not the 4:06PM or this morning’s ride at 7:03AM. (garmin recordings)

I bet we’ve got a timezone issue. Probably I’m speaking local and it’s expecting UTC or something. I’ll check with hozn.

[Vicegrip]

@chris_s 134374 wrote:

hozn pointed out that I could potentially intuit road & trail conditions from the ride data based on a) whether somebody traveled on it and b) what speed they did so. Something like:

No rides in last 24 hours = impassable [though you’d have to account for the fact that there are streets that people rarely ride on even in good conditions]
All rides in last 24 hours under 5 mph = difficult conditions (people are walking their bikes or biking very slow)
All rides in last 24 hours under 10 mph = use caution
More than 50% of rides in last 24 hours over 10 mph = clear

It’d probably have to be tweaked to be more sophisticated than that, but I bet you could get a pretty decent idea of road & trail conditions off of strava ride data automatically, without any manual reporting given a sufficient # of people authorizing the app.

I like it and agree that I think we have that right now with the standard Strava heat map. I (the grumpy old guy) bet the city folks have no idea about it nor would they avail themselves to or act on the info.

Could always hop off the bike where the tarmac ends and the snow remains and walk off the letters “Plow this trail!” with the Garmin in hand. Of course you would have to be fluent in the lost art of cursive. With all this snow I have to say I have been practicing a wee bit.

[hozn]

@chris_s 134386 wrote:

I went ahead and double-checked, and my auto-updater has been running as expected – it is updating the 24 hour data on-the-hour as expected. Was it the 24 hour data that was stale or are you looking at one of the longer period that update less frequently?

I will double check, but we should be getting gps tracks every hour or so (more frequently, I think).

Edit: timezone would definitely be an issue here. We are storing localized times, but I think the parser may indeed be translating naive timestamps to UTC. We can change that to the competition timezone (EST/EDT).

[jrenaut]

Dates and times are a disaster as far as computers are concerned. Everyone knows about the Y2K stuff but the crap that programmers deal with every day is even worse. We really need to switch to a base 10 time system recorded as milliseconds since some significant event a few thousand years ago and be done with it. We can either do it now on our own or have it forced on us when the artificial intelligences take over.

[chris_s]

So it turns out while working on some other stuff, hozn introduced a bug that was causing all recent rides to come back with swapped latitude and longitude. So your new rides were appearing….but in places like Antarctica and Iran rather than where you expected them. Looks like it is fixed now.

[hozn]

Re: dates/times, the computers are just modeling the f’d up world we’ve made for them! . But as chris_s notes this wasn’t the problem here (apparently I had remembered to make the times naive); apparently few systems can agree on whether it is lat,lon or lon,lat. The db geometries and geojson expect (lon, lat) while Strava returns (lat, lon) for coordinates. Hence, this was getting twisted around in several places.

I am now fetching lower resolution streams (to improve performance) and replacing the ride geo tracks. That will continue to happen over the evening and into tomorrow. So there may be more quirks, though I did do some basic inspection of the new data to check that it is not flipped again

At some point (later this weekend), I will add the time component to the API which will enable the animations chris mentioned earlier.

P.S. you may notice there is a “photos” page now on the site where you can see all the competition photos.

[S. Arlington Observer]

@chris_s 134408 wrote:

So it turns out while working on some other stuff, hozn introduced a bug that was causing all recent rides to come back with swapped latitude and longitude. So your new rides were appearing….but in places like Antarctica and Iran rather than where you expected them. Looks like it is fixed now.

I’ve always wanted to go to Antarctica. I guess virtual is better than nothing.

Still in awe of you tech guys.

[hozn]

The geo tracks are all low-resolution now (max 100 points per track, instead of 1000), so heat map loading times should be much faster.

[Author]

Posts