Strava shows inaccurate power data

I've noticed issues related to power measuring/output/estimations. 

It appears that for some reason when using a power meter (Garmin Vector/Edge1000) Strava appears to "downgrade" the ride by approximately 15% (both average power and NP), yet when using the Strava iOS app the power ESTIMATE as compared to an actual power meter is about 15-20% MORE than what the power meter is demonstrating. I can see 1-3% differences, but these are huge. I've taken to not relying on Strava for much of anything except for cataloging my rides. 

How in the world can anyone even trust the Strava data? The only thing that appears to be accurate is timing and distance. Even then I've seen minor discrepancies. 

Why is this occurring? Anyone with similar issues? I've got a "ticket" in the queue, but two weeks out and I've yet to hear anything back. 

I've taken to using Garmin Connect since the data is consistent with what the Edge 1000 is reporting. At least I can gauge my performance with some consistency. 

18

Comentários

90 comentários
  • Bill, never did get an answer from Strava. My general feeling is that they parse the data and then apply their own algorithms to it. What I don't understand is why even do that? The data is there in the raw so why not use that? Frustrated to say the least. 

    6
    Ações de comentário Permalink
  • I got same issue. But later on forum some one from Strava tell us ther recalculate data as: Speed, power for a better for us :D 

    They calculating stops and some non sense data. I believe Garmins data on Garmin Connect much more then from Strava. Because Garmin Connect give me data exactly from my devices with take info from my powermeter and rest ANT+ systems on my bike. 

    For example my latest ride https://www.strava.com/activities/563261751 
    Shows me avg speed 30.8Km/h but Garmin shows me 31Km/h 
    both calulate avg speed and avg speed via moving. So different is 0.2 Km/h 

    also power data: Strava shows me 186W NP but Garmin shows me 204W its huge different. 24W its a big different. 

    I dont know why Strava do this. They should take data same as Garmin Connect or TP which use correct Garmin data. 

    Also Strava Correct elevation doesnt work well. Garmin shows me on a different ride climb 720m and Strava shows me only 400m. Really stupid. 300m different??? Seriously???


    I am very sad about this what Strava do with data. 


    4
    Ações de comentário Permalink
  • Elle,

    These sudden jumps are important values. These values show instantaneous power that is critical for analysis

    Alex

    3
    Ações de comentário Permalink
  • Alex, 1000w for 10 seconds would not be filtered out. The threshold for what is considered an outlier is pretty strict. This is not the most common reason power numbers don't align between platforms. 

    3
    Ações de comentário Permalink
  • Same problem here. For example: last ride showed average 238w and Norm as 278w. After loading the ride to Strava it read 238 weighted average power and 168w as average. The difference is huge and I'd love to understand why the big difference.

     

    2
    Ações de comentário Permalink
  • Nathan, seems like Strava is mostly concerned about their interpretation of the data rather than just putting it out there. That's the part that we'll never understand because they don't share their thinking behind the algorithms. 

    I've long since moved on from Strava as a means of keeping track of performance. I still use them to log the rides, but beyond that I trust Garmin and others (Zwift, Sufferfest, Trainer Road, Training Peaks) to keep the data true and accurate. Those sites always agree on the data so Strava is the odd man out on this one. Perhaps one day Strava will change their thinking. 

    2
    Ações de comentário Permalink
  • Don't hold your breath on the membership. I put out a ticket on this in April of last year (2016). No resolution. It just appears that Strava has their own set of algorithms that they and only they understand. As you noted, Training Peaks demonstrates the Garmin data. So does Trainer Road. I use Strava primarily as a catalog to keep track of rides. It's also in Garmin Connect so I'm pretty much covered. Maybe we'll know someday why these rides are so different from platform to platform. 

    2
    Ações de comentário Permalink
  • Strava, why cant you just allow power data from a power meter to be recorded? Why do you have to manipulate it? 

    2
    Ações de comentário Permalink
  • Elle,

    Yes I have turned off the setting to exclude zeros. I stopped premium membership untill this issue is resolved. The point is not about what my activity is the point is that STRAVA must link third party devices and upload directly and only use their algorithm if those sources are not available. Thanks for response.

    Regards
    Alex

    2
    Ações de comentário Permalink
  • Also, here is some additional info regarding weighted average power: 

    Weighted Average Power is similar to - but NOT the same as - Garmin's Normalized Power.
    Both numbers attempt to capture the average effort of a workout by taking into account lactate accumulation and other types of fatigue. Strava uses an algorithm very similar to xPower for computing Weighted Average Power.

    There will be differences between the Weighted Average Power and Normalize Power. How much and which direction will depend on the ride. Typically, for shorter rides, Weighted Average Power will be lower than Normalized Power. This is a result of the different types of smoothing used by the algorithms. Both numbers have been used successfully by professional coaches.

     

     

    As to the question of why Strava can't simply copy values exactly from the Garmin file in regard to average power and normalized power: Well, our engineers have worked hard to process the raw data in the file, clean it up, and use algorithms that are standard across any file created by any compatible Strava device. It's just the way we have always approached activity data. We don't have plans to change or disable these core functions and algorithms in our data processing. 

    Please let me know if there are any other questions. 

    2
    Ações de comentário Permalink
  • I just read through this entire thread going back 2 years... seems like Strava never got around to fixing this. I have a left sided stages power meter and Strava consistently underestimates my average power. I’m new to premium membership but will not renew as this seems like an issue they don’t care about fixing.

    2
    Ações de comentário Permalink
  • Same here. After a ride yesterday where my Garmin recorded an average of 270 watts Strava presented me with 236.
    Frustrated by this, I searched and came across this thread, as Matthew R points out, going back over two years. It seems Strava manipulates our data to make it consistent with people using the least accurate devices - probably a basic phone running the app. This is not acceptable for users wanting to accurately monitor their training, and in most cases paying Strava the premium fee for them to skew our data from what has actually been recorded. .

    2
    Ações de comentário Permalink
  • Alas, the silence from Strava is deafening. I was hoping to use the power tools like power curve, TP estimate, and fitness and  freshness charts as an all in one replacement for WKO etc. But with the inaccurate capture of data I have cancelled my Summit membership after 7 years.

    I am recommending to all my followers on Strava and Insta they do the same. That is about 2000 people #strava if you care.

    2
    Ações de comentário Permalink
  • Did you ever get an answer?  I have the same issue.  My expectation is that Strava would use the actual data.  What did you find?  Thanks, Bill

    1
    Ações de comentário Permalink
  • I think its their own algorithm whoes make this problems.
    I think people will much love when Strava starts using raw data and shows them as Garmin Connect. Avg speed, avg speed in moving, climb data after correction same as garmin connect ( they cant use same map system or they use but they got a bug in code for recalculate climb ), power data...i think its not so big deal only if they listen people...

    1
    Ações de comentário Permalink
  • Sucks that you never got a reply for this issue. I have been getting ripped off by sometimes 50w average power most rides. 

    PM is reading correctly, garmin is logging it correctly, something fishy going on when uploaded to Strava though. 
    :( 

    1
    Ações de comentário Permalink
  • I noticed that issue from the first time I used a Power Meter (which is not the best as it is) however Strava sometimes robbs me for up to 75W. I then synced up Garmin directly to Strava hoping it would go away but the problem persists. The logical explanation I have is Strava don't have agreement with power meter companies and cannot use their data. Even during race one time my Garmin showed a 300W average but Strava downgraded this to 220.

    1
    Ações de comentário Permalink
  • Same issue here.

    On latest ride garmin had ave of 213W. (InPower PM) and strava had 184W

    Could it be PM is discounting zeroes and stops and strava includes them?

    Frustrating

    1
    Ações de comentário Permalink
  • So what is going on with this then?

    Strava, we are paying members. Why haven't you addressed this issue and why haven't you addressed this thread?

    It is on your support site!

    Please can you respond. You offer a great service but there is obviously something fundamentally wrong here.

    1
    Ações de comentário Permalink
  • Elle Anderson;

    You are correct in a sense that Strava shows normalized power vs average. I have turned 0 setting off but discrepancy still there. STRAVA should and must take the data from 3rd party sources( if used) DIRECTLY w/out applying their comprehensive algorithm and must only do so if the sources not present. Faster response is appreciated.

    Regards
    Alex

    1
    Ações de comentário Permalink
  • I took an example activity from Alex BK: https://www.strava.com/activities/1100055569

    (Hope you don't mind Alex!)

    Strava lists the average power as 166w. 

    TrainingPeaks and Garmin Connect list the average power as 210w. 

    Interestingly, Alex reports above that he HAS turned off the setting that excludes zeros, so that means that zeros should be included. 

     

    When we analyzed the power data from the file, there were indeed zeros present (this just confirms that the Garmin device is recording zeros in the data stream). Zero power should be recorded when you are not pedaling but still moving. I don't think zero power will be recorded when you are stopped...but this could be interesting to dig into to confirm that Garmin isn't recording zeros when you are stationary.

    In the above example, Strava is including the zeros to get an average of 166w.

    After a quick calculation, it seems that if the zeros in the above file where excluded, the average power would be about 210. 

    This appears to indicate that Garmin is recording zeros, but still excluding them from the average watt calculation. It could be worth experimenting with Golden Cheetah which is a data analysis software that allows you to switch between the two according to this discussion.

    Happy to look into other examples, so just let us know. 

    1
    Ações de comentário Permalink
  • Another note, Strava does detect outliers in the power data (a sudden jump to 1000 watts for example) and removes these erroneous values. So, it is possible that if Strava detects and removes these outliers, and Garmin does not, you may see a difference in the average values, zeros aside. 

    1
    Ações de comentário Permalink
  • Hi Elle,

    How are we supposed to record data, analyse it on Strava, and use it to better our training and cycling when there are such large discrepancies?

    The values shown on Strava contradict the live data we are are viewing during the activity! 

    How is it possible then for Strava be used as a viable training platform?

    1
    Ações de comentário Permalink
  • Hi Jake, 

    That is correct. Strava looks for outliers in the streams that are imported in the file: distance, power, heart rate, elevation, etc. For example if power data spikes to 1000w for just one second, that value would be removed, and averages recalculated. Similarly, if a large acceleration in speed is detected between points then Strava removes that point and recalculates the distance using the remaining coordinates. 

    Strava does this for all data imported to Strava, regardless of device. 

    You are also correct that these types of errors shouldn't happen often. Garmin usually does a good job recording data accurately. I don't think this outlier detection is the cause of the discrepancies reported here. 

     

    What Garmin does to calculate stats and averages on the device itself is not known. When Garmin creates the file, there is the raw data for each data point but there is also summary data at either the beginning or end of the file. The Garmin device computes a single value for average power when it creates the activity file, before uploading to any platform. For power, for example, there is a single value for average power under the Extensions field. For your ride on August 3rd it appears to be 269w. For your ride on July 4th, it appears to be 189w. Without contacting Garmin, I have no information about how this number is computed. 

    Here's what I did to troubleshoot: 

    1. I downloaded your original Garmin files and uploaded them to a Garmin Connect test account. The average power shows 269w and 189w respectively.

    2. I then exported the TCX file from Garmin Connect and opened it in a text editor. At the end of the file under Extensions I found the average power field and I removed it.

    3. Then I re-uploaded the TCX to Garmin Connect with the average power field removed. Without being able to rely on the Extension field for average watts, Garmin Connect must re-calculated the average power data based on the actual raw data points recorded in the file. The result? 142w and 131w respectively for average power. 

    4. I uploaded both files to Strava next. I got 168w and 148w respectively. 

    Where does this land us? I can't tell you how Garmin computes average power on the device. But I can tell you that once that value is removed from the file, the average power computed by Garmin Connect using the raw data is different and much lower. There is not one definitive answer regarding average power. It's possible that moving time or auto-pause is also impacting these averages (I notice the TCX version on Garmin Connect computes longer moving time for both files: 1:11:28/1:16:02 and 1:16:37/1:25:42 respectively). 

    The above is all the information I can provide at this time. 

    1
    Ações de comentário Permalink
  • Upload the data from the device & leave it alone. How hard can that be?

    The Strave 'formula', proven inaccurate , also renders the TSS & fitness graph questionable at best. Oh well 'Premium' out.

    1
    Ações de comentário Permalink
  • Strava is still doin this I notice for a few months. Even today I did short 'reactive' sprints, just one or two secs to 'jump on wheels' but none of these 800w values show, strava shows max 400w. The overall power/normalised power are close but totally removes all my peaks

    1
    Ações de comentário Permalink
  • What makes it more frustrating is that I, like many, upload accurately gathered data having invested in the best sensors I can justify/afford, including using IFTTT and a smart scale to ensure that my weight is always up to date. My age is also correctly entered. Yet contacts I follow who never bother with all this, and upload their rides from a smartphone with no HRM, cadence or power and leave their weight at the default ‘less than 124lbs’ get vastly inflated estimated power figures, which they honestly believe to be accurate, because Strava couldn’t possibly be wrong.

    1
    Ações de comentário Permalink
  • Sad that this is still an issue on Strava.  Speaking as a Senior Software Engineer, the data is the data.  I've never personally worked on or implemented a system that changes data from the authoritative source.  In this case it's the sensors that generate the data and the cycling computers that collect the data from the sensors.  This is the major reason why I have not signed up for Strava's premium service.  Why would I pay someone to change the data I worked hard to achieve.  Basically, Strava is asserting that the power meter, sensors, and cycling computer that I've paid good money for are all wrong.  

    1
    Ações de comentário Permalink
  • It took me forever to figure this out, but most of us have our head units set to not average in zeros for average power. Garmin connect shows us the power like we have it set up in the head unit. For example, if I do a trainer ride the power in Garmin and Strava is the same. If I ride outside and end up not peddling, I get a difference. For a flat ride where my average power and NP are very close, Strava power is also very close.

    1
    Ações de comentário Permalink
  • I see what you’re saying George Birdwell, Strava replied to me with the following:

    “Thanks for your message, sorry for the confusion. We may be processing data differently than your Wahoo device. Wahoo tends to use summary data to calculate these figures, whereas we use all data that is recorded in the file. There will be differences between the Weighted Average Power and Normalize Power. How much and which direction will depend on the ride. Typically, for shorter rides, Weighted Average Power will be lower than Normalized Power. This is a result of the different types of smoothing used by the algorithms. Both numbers have been used successfully by professional coaches.”

    Trouble is they don’t seem to use ALL the data recorded in the file, peak power will often be reduced by Strava - I had this happen last week on a Wattbike ride where they dropped off 33w from my peak power, due to their “algorithm”

    Personally I would prefer if Strava report back what the power meter reports, and save the made up figures for their estimated power for users without power meters.

    It’s also interesting that despite having power data to work with when I upload a ride on my 2nd bike with no PM Strava estimated power almost always shows a lack of meaningful work in Zones 2-5 which is completely the opposite of the data shown in rides with a PM, instead they show huge volume in zone 1 and 7, this is despite having actual power files to work with from the same user.

    I think that considering we pay a premium for this service it’s not great to see data manipulated so much, when all we are asking for is the accurate truth of how we performed!

    1
    Ações de comentário Permalink

Por favor, entrar para comentar.

Não encontrou o que estava à procura?

Nova publicação