strava power estimate?

M1llh0use
M1llh0use Posts: 863
edited March 2015 in Road general
so those of you with power meters, how close to accurate is the strava power estimate figure?
{insert smartarse comment here}

Comments

  • napoleond
    napoleond Posts: 5,992
    Miles out. And inconsistent. Occasionally the final average may be similar but that's more luck.
    Insta: ATEnduranceCoaching
    ABCC Cycling Coach
  • M1llh0use
    M1llh0use Posts: 863
    under/over/both?
    {insert smartarse comment here}
  • napoleond
    napoleond Posts: 5,992
    M1llh0use wrote:
    under/over/both?

    Both. It's all over the shop and as a result meaningless.
    Insta: ATEnduranceCoaching
    ABCC Cycling Coach
  • M1llh0use
    M1llh0use Posts: 863
    roger.

    Thanks.
    {insert smartarse comment here}
  • FatTed
    FatTed Posts: 1,205
    Nap D if you record a ride with a power meter, how do you know what Strava would estimate the power to be?
    I think on a mountain climb power estimates are reasonable, other times no relation.
  • napoleond
    napoleond Posts: 5,992
    Record with my phone (no power) and my garmin (powertap).

    Did this a few times to see what the estimates were like for a mate.

    Even on a steady climb they can be pretty far out. I'll try it again and see if they've improved the algorithm in the last year or so.
    Insta: ATEnduranceCoaching
    ABCC Cycling Coach
  • DavidJB
    DavidJB Posts: 2,019
    Absolute clown shoes...may as well be a random number generator.
  • rower63
    rower63 Posts: 1,991
    Actually I think the power estimates are about right, with some provisos.
    The vast majority of the power required for any course for a road bike come from climbing and aero drag. Rolling resistance and other energy losses are almost negligible by comparison. The equations commonly used for all these are quite accurate and very well established.
    If you know the gradient profile of a course, and the weight of the rider + bike, and make some standard assumptions about what aero position you're likely to be in (which in turn gives estimates of drag coefficient and frontal area on the basis that Strava "knows" your weight via your profile), then required power is simply a matter of plugging those numbers plus your speed over the segment and seeing what power comes out.
    However, Strava cannot know what wind conditions were like, so any wind at all will throw the estimate out. Also, when a segment is created, it is created from the creator's actual ride data. So if that ride data had a weak signal and dodgy and inaccurate gradient estimates, the power estimates will again be thrown. Furthermore, if the segment contains many changes of gradient, and if Strava's algorithm uses just a single average gradient for its calc (which I suspect it does), then it'll get thrown, because power-vs-time is not linear for different gradients.
    So, as I intimated at the start, I reckon the Estimated Power is correct, if the creator's ride data is accurate, if it's a windless day and if the segment is constant gradient.
    Dolan Titanium ADX 2016
    Ridley Noah FAST 2013
    Bottecchia/Campagnolo 1990
    Carrera Parva Hybrid 2016
    Hoy Sa Calobra 002 2014 [off duty]
    Storck Absolutist 2011 [off duty]
    http://www.slidingseat.net/cycling/cycling.html
  • rower63 wrote:
    Actually I think the power estimates are about right, with some provisos.
    The vast majority of the power required for any course for a road bike come from climbing and aero drag. Rolling resistance and other energy losses are almost negligible by comparison. The equations commonly used for all these are quite accurate and very well established.
    If you know the gradient profile of a course, and the weight of the rider + bike, and make some standard assumptions about what aero position you're likely to be in (which in turn gives estimates of drag coefficient and frontal area on the basis that Strava "knows" your weight via your profile), then required power is simply a matter of plugging those numbers plus your speed over the segment and seeing what power comes out.
    However, Strava cannot know what wind conditions were like, so any wind at all will throw the estimate out. Also, when a segment is created, it is created from the creator's actual ride data. So if that ride data had a weak signal and dodgy and inaccurate gradient estimates, the power estimates will again be thrown. Furthermore, if the segment contains many changes of gradient, and if Strava's algorithm uses just a single average gradient for its calc (which I suspect it does), then it'll get thrown, because power-vs-time is not linear for different gradients.
    So, as I intimated at the start, I reckon the Estimated Power is correct, if the creator's ride data is accurate, if it's a windless day and if the segment is constant gradient.


    seems-legit-2osh0x.jpg
    I'm sorry you don't believe in miracles
  • rower63 wrote:
    Actually I think the power estimates are about right, with some provisos.
    The vast majority of the power required for any course for a road bike come from climbing and aero drag. Rolling resistance and other energy losses are almost negligible by comparison. The equations commonly used for all these are quite accurate and very well established.
    If you know the gradient profile of a course, and the weight of the rider + bike, and make some standard assumptions about what aero position you're likely to be in (which in turn gives estimates of drag coefficient and frontal area on the basis that Strava "knows" your weight via your profile), then required power is simply a matter of plugging those numbers plus your speed over the segment and seeing what power comes out.
    However, Strava cannot know what wind conditions were like, so any wind at all will throw the estimate out. Also, when a segment is created, it is created from the creator's actual ride data. So if that ride data had a weak signal and dodgy and inaccurate gradient estimates, the power estimates will again be thrown. Furthermore, if the segment contains many changes of gradient, and if Strava's algorithm uses just a single average gradient for its calc (which I suspect it does), then it'll get thrown, because power-vs-time is not linear for different gradients.
    So, as I intimated at the start, I reckon the Estimated Power is correct, if the creator's ride data is accurate, if it's a windless day and if the segment is constant gradient.

    So basically, what you are saying is that the algorithms are good until you leave your house.
  • rower63 wrote:
    ..... I reckon the Estimated Power is correct, if the creator's ride data is accurate, if it's a windless day and if the segment is constant gradient.

    So basically never then? ;-)

    Comparing my "power" readings from Strava beforehand and now with a PowerTap I can see that some of them are close, but many are miles out and in either direction. Also if you have for example a road bike and a TT bike then that throws Strava power figures another curve ball.
  • Anonymous
    Anonymous Posts: 79,667
    I was looking at a segment on one of my rides which is a short steep climb of about 70 metres which you have to sprint up. Pretty much the whole top ten times it suggests a power average of 700 watts or so. Except the 1 guy who actually has a meter which it records as 450ish. Pretty much says it all for me.
  • rower63
    rower63 Posts: 1,991
    rower63 wrote:
    ..... I reckon the Estimated Power is correct, if the creator's ride data is accurate, if it's a windless day and if the segment is constant gradient.
    So basically never then? ;-)
    that about sums it up :)
    Dolan Titanium ADX 2016
    Ridley Noah FAST 2013
    Bottecchia/Campagnolo 1990
    Carrera Parva Hybrid 2016
    Hoy Sa Calobra 002 2014 [off duty]
    Storck Absolutist 2011 [off duty]
    http://www.slidingseat.net/cycling/cycling.html
  • Dorset_Boy
    Dorset_Boy Posts: 7,580
    NapoleonD wrote:
    Record with my phone (no power) and my garmin (powertap).

    Did this a few times to see what the estimates were like for a mate.

    Even on a steady climb they can be pretty far out. I'll try it again and see if they've improved the algorithm in the last year or so.

    Does you phone pick up your heart rate and cadence?
    I think you'd need to compare the actual power with power calculated with the aid of HR & Cadence wouldn't you? (Still won't be a particularly accurate estimate due to the variances many have mentioned above!)
  • napoleond
    napoleond Posts: 5,992
    Dorset Boy wrote:
    NapoleonD wrote:
    Record with my phone (no power) and my garmin (powertap).

    Did this a few times to see what the estimates were like for a mate.

    Even on a steady climb they can be pretty far out. I'll try it again and see if they've improved the algorithm in the last year or so.

    Does you phone pick up your heart rate and cadence?
    I think you'd need to compare the actual power with power calculated with the aid of HR & Cadence wouldn't you? (Still won't be a particularly accurate estimate due to the variances many have mentioned above!)
    It doesn't do that though
    Insta: ATEnduranceCoaching
    ABCC Cycling Coach
  • napoleond
    napoleond Posts: 5,992
    edited March 2015
    Did it again for a ride today. Interestingly the overall average wasn't far off, 4 watts! However... That's not the whole tale.

    EDITED due to privacy settings not seeming to work!

    The longer climbs are not *too* far out (within 10w) but everything else is seriously all over the shop as to make it worthless.
    One of the shorter sections was 170w out!

    Obviously this is an N=1 study. One rider, one ride. But the second half of this ride was a real slog after my FTP test (during which my saddle dropped and I lost about 8-10w :( ), Strava estimates that it was a leisurely bimble to the shops.
    Insta: ATEnduranceCoaching
    ABCC Cycling Coach
  • jgsi
    jgsi Posts: 5,062
    The bigger picture so the TSS would be widely as differing as is the NP.
    So a coach seeing low fluffy bunny NP instead of your true NP would probably have you all out tomorrow again because assuming you only needed 24 hours to recuperate from that 'bimble'.
    If people are relying on Strava fluffy bunnies no wonder there are 'I am knackered but not know why' threads on here.
  • FatTed
    FatTed Posts: 1,205
    Be interested in more results from long climbs when all the other variables are less important.
  • RChung
    RChung Posts: 163
    M1llh0use wrote:
    so those of you with power meters, how close to accurate is the strava power estimate figure?
    Comparison 1

    Comparison 2
  • napoleond
    napoleond Posts: 5,992
    FatTed wrote:
    Be interested in more results from long climbs when all the other variables are less important.

    So you only ride up hill? You don't ride to or from the hill, or back down it?... ;)
    Insta: ATEnduranceCoaching
    ABCC Cycling Coach
  • FatTed
    FatTed Posts: 1,205
    Of course not, but you only get to see the power estimates of the TDF riders, and uphill segment power outputs are interesting.
    https://www.strava.com/pros/laurenstendam
  • jgsi
    jgsi Posts: 5,062
    FatTed wrote:
    Of course not, but you only get to see the power estimates of the TDF riders, and uphill segment power outputs are interesting.
    https://www.strava.com/pros/laurenstendam

    That shows a power meter output.
    How to tell.....?
    When he stops pedalling , the watts go to zero... unlike fluffy bunny power.
  • DavidJB
    DavidJB Posts: 2,019
    Yes Mr Ten Dam post's his power. He's a bit of a legend.