Strava Estimated Power vs Real Power

OK, so I’ve always wondered how accurate the algorithm is on Strava for calculating the estimated power. Well today I finally tested it out for myself!

Now, I’ve always figured that the Strava one works by some kind of weighted algorithm thing, utilising the data that it has from the GPS and your pre programmed data ie; time, elevation change, distance, speed, maybe temperature?, weight (rider and bike) and probably some coefficients based on whether its offroad or not for rolling resistance etc etc.

Now, the biggie it can’t get is wind resistance, so I figured that Strava would maybe be pretty accurate in some conditions and not others. ie a consistent long climb with a consistent average surface, no wind would yeild accurate results because effectively you are reducing the unknown variables wheras if you are say pounding along on the flat with a big head or tailwind it’s gonna be wayyyyyy out.

OK, so to test it out, I went for a ride today in Upper Hutt. I had my Powertap wheel on, and got a Garmin file with that data, and also turned on the Strava app on my iphone in my pocket, so it would record the exact same ride but I’d be able to see what the difference was under different conditions.

I did test sections on 2 climbs, one short and steep and one longer and less steep, both at and below FTP, on the flat, and a max power sprinting test.

This is pretty non scientific but I think decent enough. Nerds out there, feel free to tell me the 10,000 things I’m doing wrong here…Actually don’t.

OK, so here’s the tests and what happened:

1:  Approx. 5 minute climb at around FTP.  (Little head wind, around 5% avg. grade)

Powertap average watts = 356.  Strava estimate = 374. Strava = 5.1% over

2: Approx. 1 minute climb at around FTP. (Little head wind, around 6% avg. grade)

Powertap average watts = 378.  Strava estimate = 279. Strava = 35.5% under

3: On the flat at FTP. (For this one, I tried to negate wind by riding the same section one way, then the other and averaging the two)

Powertap average watts = 382.  Strava estimate = 292. Strava = 30.1% under

4: Max Power (Sprint on the flat). (as above – averaged the two)

Powertap average watts = 980.  Strava estimate = 710. Strava = 38% under

So, if you take the four conditions, they average 24.6% under the real power. 
Pretty rubbish. As predicted, the closest measure was where the conditions allowed the algorithms to fudge better and better to a more accurate number. I predict that in that sort of condition (Test 1) that the estimate would get better and better as time went on to maybe get quite close, but for everything else, it’s not even close. Although these are pretty short tests so I guess the longer the test the more accurate it would get but I think we can safely say that Strava estimates  aren’t ever going to be very useful for performance testing since the variables are just too hard to estimate.
So there you go, in the end, if you want accurate power get a power meter. Bummer.

Looks like this guy did a similar thing, with similar results. FYI


11 thoughts on “Strava Estimated Power vs Real Power

  1. There are just too many variables that would need to be known in order for a algorithm estimate to get close to a power meter. What would be useful is if strava took your powermeter data, course info, local conditions and spat out a CDA estimate

    1. That would be cool. I always thought that they should use the power data recorded for existing segments, and use the rider data to correct the algorithm for that segment. ie, if, say, a 75 kg dude rocks up X climb on Y weight bike etc etc in 5min 10 seconds at such and such average wattage, then they would plug that data into the estimated power for the other riders on that segment and adjust the estimate wattage according to that riders weight. A browse over some segments shows that thats DEFINITELY not the case. You might be waiting a while for that CDA estimate Joel : )

  2. Thanks for doing the test and posting the results! I’ve been riding for about a year, using Strava for roughly six months and finally bought a set of Garmin Vector 2S pedals today….your results confirm my decision. Thanks again!

    1. If a MTBer falls in the forest does it make a sound? Answer: Not if you’re on bloody vodafone because you won’t have any signal. Thanks Stephen!

  3. Something you haven’t considered is that strava calculates the total power output of your bike and you together, but your crank meter only measures your human power. Drivetrain losses and rolling friction are accounted for.

    1. Thanks Dan. Drivetrain losses are only maybe a couple of %, so at around 360W, you might be looking at 3 – 9 watts maybe – something like that. So I’m not convinced that accounts for the difference. Cheers steve

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s