PDA

View Full Version : "Freeze Points" metric



vvill
02-08-2016, 01:46 PM
Hi folks,

You may have a noticed a new "Freeze Points" graph under the Individual (Various) (http://freezingsaddles.com/leaderboard/individual_various) leaderboards. This is a sort of experimental metric that's intended to capture the spirit of riding regularly in the cold, but not necessarily long distances.

Each of your rides is scored based on both the starting temperature of the ride, and your ride distance. Your highest scoring ride of the day is added to your BAFS season total.

However, there are diminishing returns: -
1. You are awarded for riding more up to a "sweet spot" of around about 3-7 miles. You can ride longer of course, but you there isn't much reward.
2. You are awarded more for rides that start at cold temperatures, especially below freezing temperatures. Between about 38-45F you can't score as many, but as you get to 32F and below, you get significantly higher scoring rides. Below about 25F the diminishing returns come back into play - as long as it's around freezing or less, you will add more to your score.


There are no daily ride bonuses and therefore no "sleaze days" in this metric (...although there's no reason you couldn't go out at 12:01am each morning and ride 5 miles!) There may or may not be a pointless prize given away based on this. I haven't spent a lot of time on calibrating/tuning the scoring function so it may be subject to change. Feedback is welcome! Obviously you can "game" your score to some extent, but I think generally it does a decent job rewarding those who ride a non-trivial distance in cold weather on a regular basis which is what I feel BAFS is all about.


Big thanks to hozn (as usual) for adding my code to the website! Temperatures are based on the same data used throughout the site (Weather Underground API).






This shows how many Freeze Points you would get for a range of starting temps/mileages:
http://i.imgur.com/79d0Hhd.png
The shape is not too dissimilar from the NWS windchill chart (http://www.nws.noaa.gov/om/winter/windchill.shtml) although my mileage scale is not linear and I'm not using exponentials. Above 45F you actually get a (small) negative score - I should probably change this to just be 0.

hozn
02-08-2016, 01:52 PM
I was telling vvill we should consider making this the scoring system in future years. It doesn't hugely shake up the leaderboard, but I think it captures the spirit of the competition. (I also kinda like the idea of getting rid of sleaze rides -- or if we keep a daily reward, also make the points factor in temperature.)

I would say adding precipitation could be interesting; however, we've seen this to be pretty unreliable in wunderground data. Plus riding "in the snow" is often a lot harder than riding "while it's snowing".

Personally I love the penalty for > 45º too :) It really changes the competition.

lordofthemark
02-08-2016, 02:12 PM
Nooooooooooooooooooooooo!

vvill
02-08-2016, 02:16 PM
Personally I love the penalty for > 45º too :) It really changes the competition.

Haha, glad to hear it. I love it too, but figured most folks wouldn't want any rides to be negatively scored. Obviously we could take it even further and move it closer to 40F but 40F in windy rainy conditions would be suffering enough for some points.

And on that note I agree that if we took into account wind, precipitation and even daylight vs nighttime as well it would be great, but unfortunately those are not as easily qualified as temperature, and the spirit of "Freezing Saddles" implies just freezing cold weather and riding a bike.



I wouldn't push to make this the main scoring system, just a fun alternative. Maybe more for the hardcore riders, because we're not trying to encourage them to just ride a mile at some point every day, we're encouraging them to ride a few miles every day whatever the temperature may be (and especially on cold days, I guess). There's no reason we couldn't have multiple scoring systems/awards with the excellent framework / database that hozn/jrenaut have set up.

dkel
02-08-2016, 02:23 PM
How fine is the resolution on the temperature data? There are days where the temperature reported on my rides doesn't match what the thermometer outside my window says. Same is true for other weather data, but that's outside the "freeze points" discussion.

vvill
02-08-2016, 02:32 PM
How fine is the resolution on the temperature data? There are days where the temperature reported on my rides doesn't match what the thermometer outside my window says. Same is true for other weather data, but that's outside the "freeze points" discussion.

This probably doesn't your question completely, but the weather data doesn't come from your Strava ride (which comes from your Garmin, or whatever). The Freezing Saddles site uses Weather Underground's data, but I'm not sure about the resolution.

Bilsko
02-08-2016, 03:02 PM
Weather data will always be difficult to get right; not sure how you have the API setup, but I imagine that you query the hourly reported temperature for the same GIS-derived location as the ride-start? So if I start my ride at home in DC, you'll pass that lat/long combo along to wunderground, which will identify the closest monitoring station and then return the temp at ride start...

One issue is that the weather station might not be reflective of my actual location (but its probably good enough for the purposes of this competition)

Another issue is if I have a long ride that takes me to locations that end up being colder than my starting point, then I don't get to capture the benefit for my bafs points calculation.
(ie, if I ride from DC - relatively warm due to heat-island effect - out to Sugarloaf, I might end up with a nice 5-10F drop in temperature. That would make a pretty big difference in the calculation - although it appears that the distance vs temp mechanism would negate the effect of me riding really far to a colder location.

Perhaps introducing a POI into your Garmin/Strava recording for the midpoint of a long ride that would capture the location where it was colder, and have the API gather both locations and report wunderground readings for both...? (ie. it captures a 34F temp at my starting location, then it also gets a 27F temp at Sugarloaf and performs a calculation based on both readings).

Clearly not something to implement this year, but worth consideration for future years.

vvill
02-08-2016, 03:16 PM
Weather data will always be difficult to get right; not sure how you have the API setup, but I imagine that you query the hourly reported temperature for the same GIS-derived location as the ride-start? So if I start my ride at home in DC, you'll pass that lat/long combo along to wunderground, which will identify the closest monitoring station and then return the temp at ride start...

One issue is that the weather station might not be reflective of my actual location (but its probably good enough for the purposes of this competition)

Another issue is if I have a long ride that takes me to locations that end up being colder than my starting point, then I don't get to capture the benefit for my bafs points calculation.
(ie, if I ride from DC - relatively warm due to heat-island effect - out to Sugarloaf, I might end up with a nice 5-10F drop in temperature. That would make a pretty big difference in the calculation - although it appears that the distance vs temp mechanism would negate the effect of me riding really far to a colder location.

Perhaps introducing a POI into your Garmin/Strava recording for the midpoint of a long ride that would capture the location where it was colder, and have the API gather both locations and report wunderground readings for both...? (ie. it captures a 34F temp at my starting location, then it also gets a 27F temp at Sugarloaf and performs a calculation based on both readings).

Clearly not something to implement this year, but worth consideration for future years.

I think there is actually an "average temperature" per ride in the database that could be incorporated quite easily, but as a first run I didn't want too many variables or too complex a query. One of the benefits of diminishing returns on mileage is that the temperature probably won't change much during someone's first 5-8 miles of riding, which is about all that matters for this scoring anyway. Admittedly a 34F vs 27F ride would be a big score difference though.



(hozn/jrenaut could tell you more about how the temperature data is obtained, I think.)

Steve O
02-08-2016, 03:48 PM
Above 45F you actually get a (small) negative score - I should probably change this to just be 0.

I don't know. -3 points for 200 miles at 50 degrees. Sounds okay: we gotta rein in subby somehow.

Steve O
02-08-2016, 03:57 PM
One other wrinkle would be sort of like the 2nd derivative.
The freeze points list still has a lot of the same names at the top.
I noticed that I'm at 87 on the freeze points list, but 71 on the leaderboard. What does that mean? I think it means that I generally start my commutes after sunrise and hence the temp of my typical daily ride is a few degrees higher than others. Also, that I live in closer to the urban heat island than, say, sethpo (68 on freeze points list; 100 on the leaderboard).

What might be interesting is to identify the riders who vary the farthest in their places on the two lists. If you are much higher on the freeze points list than the leaderboard, then you are a cold-weather badass. And vice versa.

consularrider
02-08-2016, 04:41 PM
Well Oh, Steve. I'm #15 on the freeze points list, but #55 on the leaderboard. Yeah, I'm in a generally colder place than ya'll. I've been doing most of my riding during midday.

vvill
02-08-2016, 04:46 PM
What might be interesting is to identify the riders who vary the farthest in their places on the two lists. If you are much higher on the freeze points list than the leaderboard, then you are a cold-weather badass. And vice versa.

A lot of the same names, of course, but as you say, some notable jumps between placings. Matt H, Pete B, Michael P for example all jump up quite a bit - all regular AM commuters, I think.

Just wait 'til we do this by teams... hehehe :rolleyes:

http://vignette2.wikia.nocookie.net/batman/images/8/88/Mr_Freeze_%28Arnold_Schwarzenegger%29_1.jpg



(ie, if I ride from DC - relatively warm due to heat-island effect - out to Sugarloaf, I might end up with a nice 5-10F drop in temperature. That would make a pretty big difference in the calculation - although it appears that the distance vs temp mechanism would negate the effect of me riding really far to a colder location.

I didn't run this on any real data when I created it, so now that I am looking at the results, and re-reading your reply (was rushing earlier - had to get in my kidical ride for the day!), I will say it is definitely a bit temperature sensitive. I could tweak the function to be a little less steep with respect to the weather data. (Though one could of course start a new ride at the colder location if they were seeking to maximize their Freeze Points. Or start a ride, pause it, and finish it when it's warmer, etc.)

(Also - this doesn't work with manual entries since there's no GPS data. So if Strava loses someone's 5 mile ride this weekend at <10F, it's not exactly fair!)

jrenaut
02-08-2016, 04:51 PM
You only have your top ride of the day counted towards your total? That would penalize those like me who do a lot of short rides (to school with the kids, home, back to school, back home, errand, etc). Last Tuesday I rode 19.6 miles and my longest ride was 2.5 miles.

Amalitza
02-08-2016, 04:57 PM
One other wrinkle would be sort of like the 2nd derivative.
The freeze points list still has a lot of the same names at the top.
I noticed that I'm at 87 on the freeze points list, but 71 on the leaderboard. What does that mean? I think it means that I generally start my commutes after sunrise and hence the temp of my typical daily ride is a few degrees higher than others. Also, that I live in closer to the urban heat island than, say, sethpo (68 on freeze points list; 100 on the leaderboard).

What might be interesting is to identify the riders who vary the farthest in their places on the two lists. If you are much higher on the freeze points list than the leaderboard, then you are a cold-weather badass. And vice versa.

I am currently 94th (leaderboard points) and 155th (freeze points), for a difference of 61 places, in the non-badass (ie, sensible bike rider) direction.

hozn
02-08-2016, 05:27 PM
Bilsko hits the nail on the head as far as limitations of our temperature system.

Also for the sake of API economy we use a location name if we can -- e.g. "Vienna, VA" -- rater than lat/lon. Though it appears that wunderground often just resolves these to DCA, whereas we would get more granular data, but no reuse, if we always used lat/lon.

vvill
02-08-2016, 06:24 PM
You only have your top ride of the day counted towards your total? That would penalize those like me who do a lot of short rides (to school with the kids, home, back to school, back home, errand, etc). Last Tuesday I rode 19.6 miles and my longest ride was 2.5 miles.

There is a pause button... or we could double points for #kidical rides. Or whatever.

The concept (as I imagined it) is if you are stopping a ride then you are going inside and therefore get some relief from being out in the cold, so points are awarded accordingly. 30 mins in the cold is a lot tougher than a 5 min sleaze ride (although I fully concede it's not the same with kids or heavy cargo, etc.)


Also for the sake of API economy we use a location name if we can -- e.g. "Vienna, VA" -- rater than lat/lon. Though it appears that wunderground often just resolves these to DCA, whereas we would get more granular data, but no reuse, if we always used lat/lon.

Maybe that explains the snowiest ride thing. DCA under measured Snowzilla, apparently.

jrenaut
02-08-2016, 06:26 PM
There is a pause button... or we could double points for #kidical rides. Or whatever.

The concept (as I imagined it) is if you are stopping a ride then you are going inside and therefore get some relief from being out in the cold, so points are awarded accordingly. 30 mins in the cold is a lot tougher than a 5 min sleaze ride (although I fully concede it's not the same with kids or heavy cargo, etc.)

Ok, that makes sense. You're right, it's the uninterrupted cold that really gets into your soul.

DismalScientist
02-08-2016, 07:10 PM
Well Oh, Steve. I'm #15 on the freeze points list, but #55 on the leaderboard. Yeah, I'm in a generally colder place than ya'll. I've been doing most of my riding during midday.
Perhaps it's reading the temperature in Celsius.:rolleyes:

JonRobot
02-09-2016, 09:47 AM
on the regular leaderboard i'm at 141st place (hooray nearly the middle!), and on the freeze points i'm at 55th place, for a difference of 86 points in the direction of being awesome. that's what i get for commuting around 8am every day.

bikeeveryday
02-09-2016, 10:24 AM
I don't know. -3 points for 200 miles at 50 degrees. Sounds okay: we gotta rein in subby somehow.

I agree, negative points should be awarded when the temperature is over 45F. That isn't even close to Freezing! No points for you! [imagine Soup Nazi voice]

TwoWheelsDC
02-09-2016, 10:29 AM
This system seems more complicated, so it must be better!*

Also, my Freeze Points rank is 107 higher than my regular rank. So I may not always ride, but when I do, it's during the shittiest weather.


*seriously though, I think this better captures the spirit of the competition and fully endorse its use in the future.

americancyclo
02-09-2016, 10:52 AM
Above 45F you actually get a (small) negative score - I should probably change this to just be 0.


I agree, negative points should be awarded when the temperature is over 45F. That isn't even close to Freezing! No points for you! [imagine Soup Nazi voice]

So with this scheme it's better to not ride at all if it's over 45 degrees out, rather than incur negative points?

wheels&wings
02-09-2016, 11:00 AM
Firstly, I’m good with any additional metrics…it is fun to see the data. This is my first winter with an iphone and thus electronically-recorded Strava information. It is totally cool.

Second, like some other commenters, I’m wary about the quality of the temperature/weather data coming from National Airport. Their conditions do not always seem to reflect my experience on the roads and trails.

Third, like some others, I sense that the new metric does not really capture my riding habits so well. For example often I leave work when it’s light out and then detour for groceries and an eldercare stint on my way home. Then I return through the woods late at night. When I left work it was pretty nice out, but when I return home it’s not.

Fourth, I agree that we want to reward hard-core winter riding, but I think we also want to encourage sensible, year-round riding for everyone. It’s an incredible way of life, not just a crazy-fun wintertime game. One way to ride smart in winter is to push the commuting times to take advantage of the warmer or lighter parts of the day. This is not always possible, as we all face a range of work/family scheduling responsibilities. But I don’t think we should penalize these more common-sense choices….especially for those of us for whom goofing off is a rare luxury.

Fifth, just to emphasize, I am delighted with any and all new metrics or bits of information. I feel really spoiled by hozn and others who invented such a data-rich system for Freezing Saddles. And – not insignificantly – my children enjoy the FS site too. They’ve learned all our funny team names and ask me questions like, “What ever happened to 8 Ballin’?” They’re always checking the charts of elevation, distance, and other variables. A huge thank-you for all this awesome data!

bikeeveryday
02-09-2016, 11:16 AM
So with this scheme it's better to not ride at all if it's over 45 degrees out, rather than incur negative points?

Perhaps there could also be a metric for the number of rides incurring negative points, which could be called the "Fair Weather Rider" category. You still get your mileage number, but reduced by the temperature variable. Thus, if you ride 100 miles in 50F weather, you only get 90 points (or something like that)...

bikeeveryday
02-09-2016, 11:18 AM
This system seems more complicated, so it must be better!*

Also, my Freeze Points rank is 107 higher than my regular rank. So I may not always ride, but when I do, it's during the shittiest weather.


*seriously though, I think this better captures the spirit of the competition and fully endorse its use in the future.

My regular point rank is 25, while my freeze point rank is 17. Not sure what that means, but in the scheme of things shows some consistency on my part...

Tania
02-09-2016, 11:23 AM
And – not insignificantly – my children enjoy the FS site too. They’ve learned all our funny team names and ask me questions like, “What ever happened to 8 Ballin’?”

Tell them to fret not, we are plotting a roaring return to the top.

Tania
02-09-2016, 12:16 PM
I don't know whether to love these freeze points (they put me ahead of KayakCyndi! :eek: ) or hate them (my rides are usually either sleaze rides or between 13-50 miles; I rarely hit the sweet spot).

When I checked rank this am I was 61 to 64 or something like that so not much different there for me.

Vicegrip
02-09-2016, 02:04 PM
Looks fun but I would like to see better accuracy if it is used in the points system. The weather data used and my Garmin don't even come close to agreement. I think where I start my 6 am ride is counted as DCA which is way warmer than where I start per the Garmin and the outdoor thermometer. I guess there is no way to use the Garmin data if it is present and default to the weather service if not.

Seems Hozn and I are right next to each other in the freeze points and we ride some of the same trail at a slight time shift.

eminva
02-09-2016, 02:44 PM
Also for the sake of API economy we use a location name if we can -- e.g. "Vienna, VA" -- rater than lat/lon. Though it appears that wunderground often just resolves these to DCA, whereas we would get more granular data, but no reuse, if we always used lat/lon.

Yep, this was my problem last year, and I assume this year -- I can get no traction on the "Coldest Ride" leaderboard. I live in Vienna, where my thermometer tells me it can get quite cold, but last year hozn confirmed my data defaults to DCA. I seem to be similarly mired in mediocrity on the Freeze Points chart.

But more data is always fun, even if it has its limitations. Thanks vvill.

Liz

vvill
02-09-2016, 02:58 PM
Glad to hear people are enjoying it! I remember reading some discussion about using weather data in a calculation and I wanted to try something out.


I am currently 94th (leaderboard points) and 155th (freeze points), for a difference of 61 places, in the non-badass (ie, sensible bike rider) direction.


Third, like some others, I sense that the new metric does not really capture my riding habits so well. For example often I leave work when it’s light out and then detour for groceries and an eldercare stint on my way home. Then I return through the woods late at night. When I left work it was pretty nice out, but when I return home it’s not.

Fourth, I agree that we want to reward hard-core winter riding, but I think we also want to encourage sensible, year-round riding for everyone. It’s an incredible way of life, not just a crazy-fun wintertime game. One way to ride smart in winter is to push the commuting times to take advantage of the warmer or lighter parts of the day. This is not always possible, as we all face a range of work/family scheduling responsibilities. But I don’t think we should penalize these more common-sense choices….especially for those of us for whom goofing off is a rare luxury.

Ha, I'm also more of a sensible bike rider (my rank went down). I tend to avoid very cold (and/or dark) riding times if I can, and rain as well (snow is an exception). I agree we don't want to penalize people - the metric isn't really meant to capture everyone's riding habits, I see it more as a test case/prototype that is highlighting riders that are consistently riding in freezing and sub-freezing temperatures.

I don't think there is any push to switch the scoring system, especially without better temperature data. It's just for fun, and to maybe give some ideas as to what can be done with our data. And nothing would happen unless there's a consensus anyway - the system we adopted from the National Bike Challenge is simple to understand and can be used even without GPS tracking.


I would like to see better accuracy if it is used in the points system.

Agreed. And I think it may be possible to pull a ride temperature number from Strava, but it's usually even worse in my experience. One of my Garmins consistently records 5-12F higher than it should, and they're notoriously slow at adjusting to changes (e.g. from going inside to outside) At least with the weather API everyone is using the same "ugh... DCA" data.



At this point, maybe a version of this could be made that's actually more extreme with greater negative points, precipitation, elevation, etc. since we don't have very good resolution with our weather data. (Elevation data is similarly tricky - good enough for charts and side prizes but between using topo data, and everyone's varying barometric sensors, I wouldn't put too much trust into it.)

Steve O
02-09-2016, 04:07 PM
Just a reminder that the prizes are pointless. So arguing whether or not something is "fairer" or "better" or "more representative of FS" is kind of a pointless exercise as well.
That said, including more ways to look at data allows for even more Pointless Prizes to be awarded. There is no limit on winners. We already have something like 25-30 different awards going, so what's one more? Or 5 more?


Pointless prizes for those who end up in the identical place on both lists
Person who accumulates the most negative points
Highest average "Freeze Points" per mile ridden
etc.


The one downside being that we'll have to reserve the final HH location until midnight.

jrenaut
02-09-2016, 04:19 PM
The one downside being that we'll have to reserve the final HH location until midnight.
This is only a downside because unless we have it very quickly after the last day of the competition, i will likely be on a beach in the southern hemisphere when it happens.

Oh, wait, nevermind. There is no downside to being on a beach in the southern hemisphere mere days after Freezing Saddles ends. Carry on.

hozn
02-09-2016, 05:30 PM
Yeah, my Garmin always reads 5-7° colder than weather.com (which I trust more); I don't think using Strava/Garmin data would be more accurate.

I am considering, though, switching to always use lat/lon and slowly going back to correct previous rides where we used location names. My API key should allow sufficient lat/lon queries since we only do it once per day. It will probably be close, though.

Vicegrip
02-09-2016, 05:36 PM
Just a reminder that the prizes are pointless. So arguing whether or not something is "fairer" or "better" or "more representative of FS" is kind of a pointless exercise as well.
That said, including more ways to look at data allows for even more Pointless Prizes to be awarded. There is no limit on winners. We already have something like 25-30 different awards going, so what's one more? Or 5 more?


Pointless prizes for those who end up in the identical place on both lists
Person who accumulates the most negative points
Highest average "Freeze Points" per mile ridden
etc.


The one downside being that we'll have to reserve the final HH location until midnight.Please to not get me wrong. I like pointless prizes. in fact this year I am striving at 0.00 pointless prizes as a goal in and of itself. On the other hand it was mentioned that thermal gradient might be used as a modifier for team points. Pointless prizes are 1 thing and funny but
Team Points are a whole nother thing. I am all over any way to add to my team points. I am also all over any way to add laughs and granularity to the game.

hozn
02-09-2016, 06:14 PM
Yeah, I proposed using it for official scoring. I wasn't 100%/serious, though I do think a version of this (maybe not givng negative scores for longer rides) could be more interesting than a mileage-based system. Or at least seems to capture better essence of the competition. Rides in cold should be rewarded disproportionately, no?

Temp data isn't perfect, but then neither are things like barometric elevation data and yet Strava loves climbing competitions. There are ways to improve the temp data too. (We should also track windspeed, probably.)

QuikAF77
02-09-2016, 06:46 PM
A lot of the same names, of course, but as you say, some notable jumps between placings. Matt H, Pete B, Michael P for example all jump up quite a bit - all regular AM commuters, I think.

I like this new metric, I jumped from the 30's and 40's to 7th!

hozn
02-09-2016, 07:48 PM
FWIW, I am redoing all weather data using lat/lon, so wunderground should use the closest weather station and not snap rides to regional authorities. The data won't be perfect, of course, but it should be at least the same level of imperfect for everyone (in this region, not so confident about accuracy abroad).
http://uploads.tapatalk-cdn.com/20160210/d8b4d5053e8654b63c4d3ecf06d94633.jpg

jrenaut
02-09-2016, 08:40 PM
What OS are you running? Is that Gnome 3?

hozn
02-09-2016, 09:05 PM
Just to follow up, the rides are all using lat/lon-based weather now. So this is as good as it's gonna get unless we can find a more accurate/better-coverage API for weather data.

@jrenaut, this is just Ubuntu 15.10 / Unity desktop. I'm pretty impatient when it comes to distros these days and enjoy the huge ubuntu community for solving my problems for me :)

Vicegrip
02-09-2016, 09:12 PM
What magic is this? They are speaking in tongues. Burn them!


Thanks again to all the folks for all the work making the game run and be fun.

DrP
02-09-2016, 09:13 PM
So, if you use manual entries for strava, none of these special metrics work? I guess trying to tell from my ride title where I went is tough in an automated manner. But they all are in the DC area, so one might think that could give some basic, even if not perfect numbers (and yes, DCA is all wrong with temp - there are many times through out the year that once you reach the Gravely Point parking lot you heat up until you cross 4MR. And one very memorable occasion of walls of temperature differences. And then the fog days.). Oh well. I guess I pay for not having something recording my every move. Darn.:rolleyes:

jrenaut
02-09-2016, 09:15 PM
@jrenaut, this is just Ubuntu 15.10 / Unity desktop. I'm pretty impatient when it comes to distros these days and enjoy the huge ubuntu community for solving my problems for me :)
I realize that less than 5% of the forum here will care about this, and less than 10% will even get what we're talking about, but threadjacking is a part of this forum and I can't help myself. I like Debian-based distros (and i'm currently running Debian 8 with Mate desktop) because they tend to be lighter than Ubuntu but can still use all the Ubuntu packages. When I finish this contract I'm on and have a minute of computer downtime I'm planning to reinstall my OS (which I like to do every 6 months or so, and would recommend everyone do on a similar schedule) and go back to Crunchbang (or at least the community continuation of the dead OS (https://crunchbangplusplus.org/)).

vvill
02-09-2016, 09:18 PM
Nice!!! I assume this would affect snowiest/rainiest ride data too?


Just a reminder that the prizes are pointless. So arguing whether or not something is "fairer" or "better" or "more representative of FS" is kind of a pointless exercise as well.

Further to Vicegrip's reply: I don't think anyone is arguing at all. It's just discussion, which I don't think is actually pointless. If they were completely pointless, people wouldn't bother coming up with them or competing for them, or awarding them! They're more auxiliary or secondary but that doesn't have the nice alliteration.

I guess the question is do we want to try to take this further than the original basic concept which was just adapted from the National Bike Challenge. It does seem a shame not to do a little more with all the data and keen riders we have. I imagine we'd always keep the existing scoring system (maybe it'd be the "Classic" leaderboard), but perhaps something involving more variables would be more interesting.

hozn
02-09-2016, 09:37 PM
I like Debian-based distros (and i'm currently running Debian 8 with Mate desktop) because they tend to be lighter than Ubuntu but can still use all the Ubuntu packages.

Wait, Ubuntu *is* a debian-based distro! :-) I don't mind the heft since it performs very snappy on my current system (the solid-state drive helps); I use Xubuntu on another older system where I want better performance. I like Debian's stability, but want the latest/greatest packages. I have not tried this in dpkg land, but I assume you could not actually install Ubuntu packages that have been built against newer glibc (or other core Linux shared libs). You might also have to contend with different dep pkg naming, etc. At least that has been my experience mixing packages from different RH-based distros.

Ok, I won't hijack vvill's thread anymore. :-)

Yeah, @vvill, this should affect snow data, though I should have done a before/after comparison. It doesn't look hugely different.

davidq
02-09-2016, 11:10 PM
FWIW, I am redoing all weather data using lat/lon, so wunderground should use the closest weather station and not snap rides to regional authorities. The data won't be perfect, of course, but it should be at least the same level of imperfect for everyone (in this region, not so confident about accuracy abroad).

Wow, nice work hozn.

I have a question for the coldest ride, what's the criteria? I check mine looks like it pick the ride on Feb 6 which marks 33.10F (average?), but I knew this is not true. I have many short ride under frozen temperature, but the one I won't forget is the ride I did on Jan 18 (https://www.strava.com/activities/473181536) because my water bottle totally frozen and I didn't get drink until rest stop. My phone AccuWeather show that I started at 14F and finished at 19F. I am pretty sure the whole DC area should has similar temperature on Jan 18, which is MLK Birthday.

hozn
02-10-2016, 04:49 AM
I will get back to you on that. The criteria is based on ride start location and weather observations over the duration of the ride, but I am not sure what chart you are referring to, so I will check the data.

eminva
02-10-2016, 05:53 AM
Just to follow up, the rides are all using lat/lon-based weather now. So this is as good as it's gonna get unless we can find a more accurate/better-coverage API for weather data.

Thanks! I moved way up on the coldest ride leaderboard (can't remember where I was; now I'm #30). Still neither under- nor over- performing on Freeze Points -- I blame it on those rides in Florida last month. :roll eyes:

Liz

davidq
02-10-2016, 07:07 AM
I will get back to you on that. The criteria is based on ride start location and weather observations over the duration of the ride, but I am not sure what chart you are referring to, so I will check the data.
I mean "individual various" > "coldest ride", but now looks like it got correct.

hozn
02-10-2016, 07:14 AM
Wow, nice work hozn.

I have a question for the coldest ride, what's the criteria? I check mine looks like it pick the ride on Feb 6 which marks 33.10F (average?), but I knew this is not true. I have many short ride under frozen temperature, but the one I won't forget is the ride I did on Jan 18 (https://www.strava.com/activities/473181536) because my water bottle totally frozen and I didn't get drink until rest stop. My phone AccuWeather show that I started at 14F and finished at 19F. I am pretty sure the whole DC area should has similar temperature on Jan 18, which is MLK Birthday.

Perhaps you were looking at this while the data was being reprocessed? Your coldest ride shows on the chart as 15.8F.

Indeed the ride you linked had an avg temperature of 18.14º and average windchill of 3.4º (maybe we should incorporate windchill into these metrics).



mysql> select ride_id, ride_temp_start, ride_temp_end, ride_temp_avg, ride_windchill_start, ride_windchill_end from ride_weather where ride_id = 473181536;
+-----------+-----------------+---------------+---------------+----------------------+--------------------+
| ride_id | ride_temp_start | ride_temp_end | ride_temp_avg | ride_windchill_start | ride_windchill_end |
+-----------+-----------------+---------------+---------------+----------------------+--------------------+
| 473181536 | 15.8 | 19.4 | 18.14 | 3.4 | 4.5 |
+-----------+-----------------+---------------+---------------+----------------------+--------------------+
1 row in set (0.00 sec)


Note: the average temperature is calculated by sampling the observations (for *start* lat/lon) over the duration of the ride. (So it is not just (start_temp+end_temp)/2)

Attached is a shortened (to just a subset of AM observations) version of the weather file data that was used to calculate this in case you (or anyone else) wants to see the raw material here. The full files have observations for at least every hour, so are much larger (too large to upload, apparently).

davidq
02-10-2016, 08:53 AM
hozn, you are amazing.

Any chance add team freezn points. I am not sure how vvil get his magic metric, but I am pretty sure that is against his/our team since most of our teammates Freezing Points ranking are lower than their individual point ranking.:(

Steve O
02-10-2016, 09:51 AM
(maybe we should incorporate windchill into these metrics)

No. Wind chill is mostly irrelevant to bike riding. On a calm day you generate your own wind chill anyway. A morning with 15mph winds from the west makes my wind chill differential 0 (it's a tailwind, so I'm essentially riding in the calm) and hozn's as though it's a 30+mph wind. Like today, for instance. How do we account for that? Do we grant equal wind chill points to GB flying in with the wind at his back wearing a giant smile and hozn, battling against the wind for every inch with a frozen grimace?

We all already account for a 12-20mph wind when we prepare ourselves. Whatever wind is out there may add or subtract from that, but it probably evens out over time. It's built in already, IMO.

Steve O
02-10-2016, 10:18 AM
mysql> select ride_id, ride_temp_start, ride_temp_end, ride_temp_avg, ride_windchill_start, ride_windchill_end from ride_weather where ride_id = 473181536;
+-----------+-----------------+---------------+---------------+----------------------+--------------------+
| ride_id | ride_temp_start | ride_temp_end | ride_temp_avg | ride_windchill_start | ride_windchill_end |
+-----------+-----------------+---------------+---------------+----------------------+--------------------+
| 473181536 | 15.8 | 19.4 | 18.14 | 3.4 | 4.5 |
+-----------+-----------------+---------------+---------------+----------------------+--------------------+
1 row in set (0.00 sec)


Wait, Ubuntu *is* a debian-based distro! :-) I use Xubuntu ..I like Debian's stability, ..I have not tried this in dpkg land, but I assume you could not actually install Ubuntu packages that have been built against newer glibc (or other core Linux shared libs). You might also have to contend with different dep pkg naming, etc. At least that has been my experience mixing packages from different RH-based distros.

I like Debian-based distros (and i'm currently running Debian 8 with Mate desktop) because they tend to be lighter than Ubuntu but can still use all the Ubuntu packages. .. go back to Crunchbang

@jrenaut, this is just Ubuntu 15.10 / Unity desktop. I'm pretty impatient when it comes to distros these days and enjoy the huge ubuntu community

What magic is this? They are speaking in tongues. Burn them!

10929

jrenaut
02-10-2016, 10:21 AM
How do we account for thousands of torch-bearing citizens in our weather calculations?

hozn
02-10-2016, 10:42 AM
No. Wind chill is mostly irrelevant to bike riding. On a calm day you generate your own wind chill anyway. A morning with 15mph winds from the west makes my wind chill differential 0 (it's a tailwind, so I'm essentially riding in the calm) and hozn's as though it's a 30+mph wind. Like today, for instance. How do we account for that? Do we grant equal wind chill points to GB flying in with the wind at his back wearing a giant smile and hozn, battling against the wind for every inch with a frozen grimace?

We all already account for a 12-20mph wind when we prepare ourselves. Whatever wind is out there may add or subtract from that, but it probably evens out over time. It's built in already, IMO.

Yeah, that's fair. I was thinking we'd also track wind speed/direction and ride trajectory ... but that is definitely getting more complicated. (And fetching windspeeds at various points along the ride would become a familiar problem of over-saturating API limitations.)

For my own use, I've thought of making a utility that updates my ride descriptions with windspeed and direction. This would let me normalize those variables to get closer to a question of answering the speed of different tires (e.g. take the Hunter Mill -> Vienna segment for days with different tires and same windspeeds and look at speed/power graph [or averages].) Have to think through that a bit more, but seems like it could work (or work well enough to provide meaningful data).

ewilliams0305
02-10-2016, 12:54 PM
Looks fun but I would like to see better accuracy if it is used in the points system. The weather data used and my Garmin don't even come close to agreement. I think where I start my 6 am ride is counted as DCA which is way warmer than where I start per the Garmin and the outdoor thermometer. I guess there is no way to use the Garmin data if it is present and default to the weather service if not.

Seems Hozn and I are right next to each other in the freeze points and we ride some of the same trail at a slight time shift.

I've also noticed this. For example When I road to WVA and Chambersburg and ...... Stava reported my temp at 32 deg. I can assure you it was warm (32ish) at 4am when I left the city. By the time I was in WVA about 10am it was closer to 26, I know this because the university in shepherdstown has a big LED sign with the temperature displayed.

I'm sure it would be a royal pain in the frozen saddle to have the metrics include such a large geographic area on a single ride. I'm not sure if the weather underground API supports some kind of time/location GET. Neither am I sure if the strava ride parsing gives enough data to get the location of the rider at a given interval or time on the ride. If all this were or is possible, we could use the riders location at (1) hour intervals to grab a temperature every hour for the number of elapsed hours. Average then together and you'd have a very accurate way of determining the average temp.

Even on a 1-1 1/2 hour ride like what vicegrip is describing It would potentially decrease the overall temp of his ride as you'd have 2 temps to average.

All that said, how can I use all this data for my Longest coldest ride? I don't think we've had any rides into the negatives yet so my current absolute zero is still zero.

hozn
02-10-2016, 01:25 PM
I've also noticed this. For example When I road to WVA and Chambersburg and ...... Stava reported my temp at 32 deg. I can assure you it was warm (32ish) at 4am when I left the city. By the time I was in WVA about 10am it was closer to 26, I know this because the university in shepherdstown has a big LED sign with the temperature displayed.

Just to clarify here, temperatures reported by Strava are from your Garmin, not from any weather datasource. (Unless they've changed something.) -- And these numbers almost certainly won't match up with the avg temperature for rides that we calculate from wunderground observations.



I'm sure it would be a royal pain in the frozen saddle to have the metrics include such a large geographic area on a single ride. I'm not sure if the weather underground API supports some kind of time/location GET. Neither am I sure if the strava ride parsing gives enough data to get the location of the rider at a given interval or time on the ride. If all this were or is possible, we could use the riders location at (1) hour intervals to grab a temperature every hour for the number of elapsed hours. Average then together and you'd have a very accurate way of determining the average temp.


We could absolutely do this. The wunderground API returns weather observations (for the entire day, usually in 1-hour increments) based on a latitude and longitude. And now that we're storing full geographic streams for rides, we know exactly where you are at each of those hours during the ride. The "problem" here is that we have 265 riders and I'm limited to making 500 wunderground API requests per day. Hence we only request the weather for a single point per ride (start lat/lon).



Even on a 1-1 1/2 hour ride like what vicegrip is describing It would potentially decrease the overall temp of his ride as you'd have 2 temps to average.

For rides <= 2 hours (which is 92.7% of rides in the database) we could just grab ride end point and use that for the second of the observation readings. This really only applies to commuters who log 2 rides for their commute (most) -- or other people that are ending their ride somewhere else. We might be able to squeeze that into the API rate limits if we make the assumption that we'd be fetching both start/end points anyway (i.e. start_geo is home in the AM and work in the PM, which means we've got a days worth of observations for both locations).

Another thing we could do is decrease the fidelity of our lat/lon points -- i.e. make them fuzzier -- so that people that start out a couple blocks apart don't require separate wunderground API calls. Essentially come up with some grid system that we think is fine-grained enough that weather stations will be the same anyway and then snap lookups to those coordinates. Then we'd probably have enough leftover requests to do multiple requests per ride and improve the data.

... But I think any of that is going to require someone else to offer to do the legwork :)

hozn
02-10-2016, 01:30 PM
So, if you use manual entries for strava, none of these special metrics work? I guess trying to tell from my ride title where I went is tough in an automated manner. But they all are in the DC area, so one might think that could give some basic, even if not perfect numbers (and yes, DCA is all wrong with temp - there are many times through out the year that once you reach the Gravely Point parking lot you heat up until you cross 4MR. And one very memorable occasion of walls of temperature differences. And then the fog days.). Oh well. I guess I pay for not having something recording my every move. Darn.:rolleyes:

Correct, none of this works for manual rides. In fact, manual rides are excluded from many of the non-points leaderboards. Basically I got tired of people forgetting to enter a duration for their manual rides (distance 10, time 0 = infinite speed!) so any metric that has anything to do with time, speed, etc. excludes manual entries.

Subby
02-10-2016, 02:22 PM
2nd overall and 23rd in freeze points. Ride smarter not harder. :D

americancyclo
02-10-2016, 02:34 PM
Just to clarify here, temperatures reported by Strava are from your Garmin, not from any weather datasource. (Unless they've changed something.) -- And these numbers almost certainly won't match up with the avg temperature for rides that we calculate from wunderground observations.



So does this mean that it pulls from the closest station and also pulls temp at the departure time?

Here's a comparison of my data for this morning using my garmin, the Wunderground PWS that are closest to my start and end point, and DCA.




Garmin
Wunderground
DCA


depart
23
29.8
32


arrive
28
30.7
32

hozn
02-10-2016, 03:06 PM
Yes, we can crossref those numbers with the numbers in the system after tonight's run.

jrenaut
02-10-2016, 04:09 PM
Out of curiosity, I put temperature on the main screen of my Garmin 510 (about a year old, firmware updated in the last month or so). Very unscientific test, but the Garmin didn't do too well. It was about 55 when I turned it on, which may be perfectly reasonable in the laundry room where it was sitting for a few hours. Over the course of my 15 minute ride, it dropped steadily until it maybe was close to the actual temperature. It was in my pocket for 10-15 minutes while I retrieved the kids, and it was back up in the 50s when I turned it back on. For the entirety of the ride home, a little more than 15 minutes, it was dropping.

This would suggest that if your ride is less than about 25 minutes, the temperature data is completely meaningless. But as I said, not a terribly scientific test.

hozn
02-10-2016, 07:26 PM
Yeah, the Garmins take a looong time to adjust. Some people put them in the freezer to help accelerate the adjustment. Mine is steady since it sits in the shed on the bike all night. But both my Garmin 510 and 1000 do read low in winter time (unsure in summer; I don't look at the temp). Both are consistently 5-7º colder than weather.com temps. E.g. it was 34º when I left the office and my Garmin said 28º.

DismalScientist
02-10-2016, 07:59 PM
Some people put them in the freezer to help accelerate the adjustment.

Thermodynamic doping? Is there no limit to what level the BAFS Freds won't sink?:rolleyes:

Tim Kelley
02-11-2016, 07:42 AM
Thermodynamic doping? Is there no limit to what level the BAFS Freds won't sink?:rolleyes:

Strava KOMs are easier to get in the summer due to less dense air...

Steve O
02-11-2016, 08:35 AM
Strava KOMs are easier to get in the summer due to less dense air...

And the humidity (humid air is less dense than dry air)

And since air gets thinner at higher altitudes, you tend to speed up towards the top of the hill. Thus longer, higher hills have faster KOMs. ;)

Tim Kelley
02-11-2016, 08:40 AM
http://velonews.competitor.com/2013/03/bikes-and-tech/technical-faq/technical-faq-why-is-riding-in-the-cold-so-hard_277626

vern
02-11-2016, 09:09 AM
And the humidity (humid air is less dense than dry air)

And since air gets thinner at higher altitudes, you tend to speed up towards the top of the hill. Thus longer, higher hills have faster KOMs. ;)

Not thinner at higher altitudes, but less dense. The molecules are not packed as tightly together.

Amalitza
02-11-2016, 11:51 AM
For the record, "sometimes the wind is a headwind and sometimes it's a tailwind so it all evens out" is true in the same sense that "sometimes you're going up the hills and sometimes you're going down the hills, so it all evens out" is true.:rolleyes:

//just returned from a mostly SE out and mostly NW back "recreational ride"//

bentbike33
02-11-2016, 12:38 PM
just returned from a mostly SE out and mostly NW back "recreational ride"

You did that backwards, for today anyway.

Amalitza
02-11-2016, 01:01 PM
oh, it's backwards pretty much every day

Steve O
02-12-2016, 01:26 PM
So, my understanding is that a metric like this is intended to reward those who ride in the gnarliest weather more than those who ride in the clementest weather (I made that word up).

The problem is defining "gnarliest." Is it the cold? Wind? Precip? Ice? Perfect storm combos of these?

One way to get at whether the weather is bad is to use ourselves as proxy. Lookie here:
10952

There isn't much correlation between temperature and number of daily riders. The day with the fewest riders was Jan 23, which I think most of us would agree was the gnarliest riding day this year. But not the coldest.

There's a pretty strong correlation between bad weather and number of riders, so we could devise a formula or bonus system or something based on that rather than trying to parse and define the weather itself.

If we think FS participants are not a good proxy (I would argue that, as a group, we are a decent statistical sample for this purpose), then we can use the Arlington counter data. Staff is pretty convinced that the rider numbers fluctuate in strong correlation to weather conditions: gnarlier weather; fewer riders.

Then we don't have to debate whether a strong headwind is gnarlier than a polar vortex, or if sleet is worse than snow.

Vicegrip
02-12-2016, 02:47 PM
Miles ridden per day rather than riders might be a good metric to look at if doable. On a snow day many people can sleaze a wobbly 1.1. Grade on a curve? On days when the miles totals goes down measurably increase the point value.

jrenaut
02-12-2016, 04:30 PM
Weighting for conditions sounds cool, though I'm not sure it would really make a difference. Making up numbers for an example - let's say on an average day all participants ride 1000 miles. On a day when we all ride 1500 miles, you score .75 points per mile. On a day when we all ride 500 miles, you score 2 points per mile. I suspect our high mileage riders are out there no matter what.

Or maybe we look at your average day vs the overall average? Everyone rides 1000 miles a day avg, you ride 10 average. Today everyone rode 500 but you still hit your 10, so you have a bonus of 5? That might be interesting. Kind of a "the weather does not affect me" chart.

hozn
02-12-2016, 05:51 PM
I think we could come up with a reasonably good "gnarly" score that gives weights to different factors. There is probably a relatively easy way to factor in headwind too, though if we do this by day then wind cardinality (?) does seem to zero itself out and simply wind speed becomes the weight.

The thing we probably can't factor in easily is the presence of snow on the ground. If anyone knows a good data source for that ...? Maybe we could add a relative factor that just assumes that days with fewer riders are by definition gnarlier. Maybe a bit of a chicken/egg problem with calculations :-)

Edit: I realize now (after reading) that my second suggestion is exactly what vicegrip proposed :)

vvill
02-12-2016, 07:31 PM
The problem is defining "gnarliest."

The idea of this first implementation of Freeze Points was to award points based on temperature and a consistent amount of mileage rather than purely miles and a daily bonus (which I feel is perfectly fine for the National Bike Challenge but could be made into something more winter specific). It's a long contest, and there are a lot of cold riding days, so I wanted to capture that sort of "spirit". You can of course ride 1.0 miles on the days where you wouldn't normally ride at all and there's nothing wrong with that, but I think if you really embrace winter riding, you'll end up only sleazing on days where you're too tired or busy, rather than when the conditions are less favourable (although I don't necessarily condone crazy-riding-on-ice-with-23s... hi Subby!)

There would be no curves based on mileage/temperature if it was just supposed to gnarliest, because let's face it, riding 250+ miles in freezing temperatures on Jan 1 would just blow a gnarliest score out of the water (...hi Eric!) Another element was time - I don't think everyone can make the time to ride mega miles, but most people who are regular commuters, utility riders or even just regular recreational riders will easily reach the "sweet spot" of mileage. The current system makes it impossible for time-crunched folks to ever match the really high mileage riders, and I think something like Freeze Points gives everyone a better chance to achieve higher scores.

I certainly don't think the Freeze Points metric is perfect (having seen the full output, I think it's weighted a little too much based on temperature) - but I'm happy it's generated so much discussion and to that end, I enjoy reading this thread. I like that people are thinking of other metrics (and I agree with hozn that we could surely devise a "gnarliest" one based on time of day, precip, wind, temperature, mileage, average speed, VAM, and whatever else, if that's what we want - is it?). I, like many here I imagine, really enjoyed seeing a lot of the Individual/Team leaderboards and data exploration charts when they came up but after a while I couldn't help thinking we could derive even more.

The idea of a self-scaling metric is interesting, although yeah a day's "clemency" rating can't be finalized until everyone's submitted their ride data and obviously it will be skewed with weekends, holidays, etc. and travelers who are riding outside of the DC area. The nice thing about weather is it's out of our control, so it's a more absolute measure, but OTOH a self-scaling metric does avoid having to weigh up the relative influence of each data input - I would agree with Vicegrip that miles/day would be more meaningful than riders/day, although maybe I'd cap the miles that each rider could contribute at about 40-50.

I agree that the temperature vs. riders graph is mostly interesting because it shows how little influence temperature has on this crazy group of riders we have! I think frozen precipitation on roads/trails overnight would be the biggest factor influencing the number of riders the next day.

hozn
02-12-2016, 09:06 PM
Miles ridden per day rather than riders might be a good metric to look at if doable. On a snow day many people can sleaze a wobbly 1.1. Grade on a curve? On days when the miles totals goes down measurably increase the point value.

I do think miles vs. temp is more illustrative. I changed the chart. http://freezingsaddles.com/explore/distance_by_lowtemp

It clearly still lacks an element of "gnarly", though, as there are interesting holes -- like Jan 23.

ewilliams0305
02-14-2016, 06:47 AM
There would be no curves based on mileage/temperature if it was just supposed to gnarliest, because let's face it, riding 250+ miles in freezing temperatures on Jan 1 would just blow a gnarliest score out of the water (...hi Eric!) Another element was time -

I think it's also important to factor in "ride time", something I failed to do with my longest coldest ride award. I realize it's important to not only reward people for big miles and cold days but also ride time. I think it's a heck of a lot easier to ride a metric when it's 20 degrees out if you can ride a metric in 3-4 hours. If this metric took say 4-6 hours it's safe to say your suffering on a completely different level. Another example, shuby road 109 miles at 21 degrees yielding a score of 5, I have NO doubt this was a lot harder than say riding 6 miles at 0 degrees. I'll have to start my own thread about this but I think a good measure would be ride time VS temp.

As VVILL noted my 257 mile ride was averaged at 32 (not cold), but after 12 hours of being outside I can tell you I've never been this cold in my entire life.

Rod Smith
02-14-2016, 06:58 AM
`I like to wait until it warms up. Kudos to those hearty souls wise enough to ride in the middle of the day when the sun is at it's highest above the horizon.