Strava and Power

jgsi

I see that whenever you do a Strava segment it will give you their estmation of watts used to get yourself up there.
Anyone checked the accuracy of Strava's algorithm or whatever it is against eg a powertap?
ta

racingcondor

I havn't checked it but it's not very accurate. You can tell who has a powermeter by the lighting bolt icon (I seem to remember), find yourself a segment or two which include people with them and it's pretty clear that it's 'very estimated'.

jgsi

Cheers for that!
It is still strangely addictive.

jibberjim

racingcondor wrote:

I havn't checked it but it's not very accurate. You can tell who has a powermeter by the lighting bolt icon (I seem to remember), find yourself a segment or two which include people with them and it's pretty clear that it's 'very estimated'.

Most people don't enter accurate weights of themselves and bikes though either, so you can't really tell if the guys who are estimating are rubbish 'cos they took 5kg's off their weight 'cos they forgot all the pies they ate last month, and put in 6kg's for the bike forgetting about the saddle bag full of tools and 2l's of water on it. It also of course cannot know about drafting or the wind direction, and those are really the biggest difference unless the hill is very steep drafting can make quite a difference. (Box Hill for example, the first 4 times all involved drafting but you'd never know that from the leaderboard) And that's on a hill commonly used as a solo test, on other hills more often ridden in groups it's even worse - Staple Lane only has John Storms and Wouter Sybrandy of the top 10 that weren't group rides for sure.

It also cannot know how aerodynamic you are of course, and that will again make a difference on anything that goes at a reasonable speed. With steep hills it can be more accurate, as then the speeds tend to be low enough that it's only weight that particularly influences it, and that can be well known.

Mike67

Yes it can be wildly out. Seems to work it out from a combination of weight, speed, incline and HR.

For instance it's got my 5 min power at 315W (I wish ) which was on a generally downhill road but with some sharp up bits that the accompanying massive tailwind carried me up at a decent speed.
To compare I did a 3 mile section of downhill (-4.6% grade) yesterday but had to pedal hard to maintain an average of 28mph due to a strong headwind and ended up well down the leaderboard. It gave me a power level of just 30W!!!! It must take that to walk a reasonable pace.

I think it's reasonable as a comparator for your own rides but little else really.

RoadMeridaBen

Last year on a 25 mile TT for the first 9 minutes I averaged 522 watts according to strava, and I weigh 66kg's lol

spokegeezer

The power guess in Strava is of course complete rubbish once you realize the graph has you pushing 300Watts when your cadence shows zero!!! Pretty impressive eh? lol

kettrinboy

Ive seen 5 and 10 min power figures of 700 to 800 watts for some users on Strava, if that were true Dave Brailsford would be falling over himself to get them in Team GB for the olympics.

term1te

The figures don't look too far off for 5 - 10% climbs, but seem to fall down a bit out side of that range. I saw one segment where a guy had pushed 2.5kw over a 90 m elevation difference. I guess he stayed in the car for that one.

doyler78

Sometimes ok and sometimes absolute rubbish. Wind is the biggy from what I've seen. If there's a tailwind and you are going for a segment then you are going to get a big power number in comparison to what you actually did and the big problem with Strava is that I bet tailwinds play a pretty big part in a lot of times on the leaderboards so pretty useless when you factor that in IMHO.

vs

doyler78 wrote:

the big problem with Strava is that I bet tailwinds play a pretty big part in a lot of times on the leaderboards so pretty useless when you factor that in IMHO.

Surely the leaderboards are just a bit of fun for those who wish to participate.

The power is not going to be accurate for all the reasons stated by jibberjim.

Strava say so themselves:
"Average power is most accurately measured through a power meter, though if you don't have a power meter we give a rough approximation through our power estimator."

Strava is an amazing development, something that is beyond all our dreams - if you can say otherwise then I guess you missed out on fame and fortune and I look forward to seeing your rival app.

doyler78

vs wrote:

doyler78 wrote:

the big problem with Strava is that I bet tailwinds play a pretty big part in a lot of times on the leaderboards so pretty useless when you factor that in IMHO.

Surely the leaderboards are just a bit of fun for those who wish to participate.

The power is not going to be accurate for all the reasons stated by jibberjim.

Strava say so themselves:
"Average power is most accurately measured through a power meter, though if you don't have a power meter we give a rough approximation through our power estimator."

Strava is an amazing development, something that is beyond all our dreams - if you can say otherwise then I guess you missed out on fame and fortune and I look forward to seeing your rival app.

I mentioned the leaderboards quite simply because that's where I see these dodgy power numbers most often as that's where I most often see other peoples numbers

Well it's quite simple for me is something is so utterly useless then why have it?

You come you come across as bit of a fan boy in that last remark. There is a great deal that is wrong with Strava - it is basic to say the least in terms of analysis, even with premium and quite frankly the level debate when it gets down to well "if you can do better" type arguments is just plain silly. I don't propose to be able build my own wheels however I feel justified in complaining about them if they aren't true

alihisgreat

They should include a wind algorithm? just take data from the Met office and they can work out who's got a tailwind and who hasn't?

vs

ALIHISGREAT wrote:

They should include a wind algorithm? just take data from the Met office and they can work out who's got a tailwind and who hasn't?

That's actually a great idea.

amaferanga

vs wrote:

ALIHISGREAT wrote:

They should include a wind algorithm? just take data from the Met office and they can work out who's got a tailwind and who hasn't?

That's actually a great idea.

Really? Cos the met office always get the weather right don't they

They should really just ditch the power thing. Even with a 'wind algorithm' the figures are unlikely to be anywhere near reality. If you can't measure power directly then there's no point in trying to guess it.

Omar Little

would be good if they could have a symbol on the segment leaderboards identifying who used a gps unit (particularly one with a speed sensor magnet too) and those who used a phone.

I realise it doesn't matter but i was out today recording with my garmin whereas the guy i was with was using his iphone. i beat him to the top of all climbs but when he uploaded his data he ended up getting faster times and the virtual bragging rights with clubmates congratulating him on setting a new KOM time that we've all been locked in battle over for the last few weeks. Bastard!!!

a_n_t

I have a powertap. Strava is usually way over in its estimations.

vs

amaferanga wrote:

vs wrote:

ALIHISGREAT wrote:

They should include a wind algorithm? just take data from the Met office and they can work out who's got a tailwind and who hasn't?

That's actually a great idea.

Really? Cos the met office always get the weather right don't they

Yep, present time data (which is what you would have to use) would be accurate; you wouldn't be using forecasts

GoldenBear

Another problem on leaderboards arises when someone leaves their GPS running whilst driving then not editing it out. One of the hills I use as one of my hill repeats has the fastest going up 20-22mph (depending on drafting as a few chaingangs go up) then this one lad who's gone up at 40mph. There's a few like that near me.
I use it to monitor self-improvement.

doyler78

GoldenBear wrote:

Another problem on leaderboards arises when someone leaves their GPS running whilst driving then not editing it out. One of the hills I use as one of my hill repeats has the fastest going up 20-22mph (depending on drafting as a few chaingangs go up) then this one lad who's gone up at 40mph. There's a few like that near me.
I use it to monitor self-improvement.

Just report them. It's great fun. Their results immediately disappear until the ride file is reviewed.

Alex_Simmons/RST

I can only imagine Strava power numbers exist purely for some light comic relief.

turboslave

The power readings are way out, and I doubt if anyone actually believes them, the mobile apps are also not acurate, I recently completed a TT and another member of our club also rode the event, I didnt get his time after but saw his ride upload, incuding his ride out, the event and ride home his ave was 2mph quicker than my time, I thought fair play he did a good ride, only when the official results were posted, he was 4 sec slower than me, I very much doubt that his ride out and back was done at a greater speed than his race pace, and yes he uses a mobile app.

ded

turboslave wrote:

the mobile apps are also not acurate

I think that probably depends on your phone more than Strava. I did a TT yesterday and if I crop/segment the actual route on Strava the GPS time is 2 secs out from my recorded time (using some humans and some stopwatches) which is pretty close. The Strava app on my Android phone is actually way better at tracking than say Google Tracks even tho' it is using the same GPS hardware...???

I agre on the power stuff tho' - I don't have a powermeter to prove it but my "readings" on Strava go from Cancellera to old lady with shopping on a day to day basis...

t5nel

What do people make of the calculators at analyticcycling.com? For hill climb they tend to give me results about 15% lower than strava. Rest of the time wind has got to be the biggest factor that make strava inconsistent. I am pretty sure it is abbout as good as it can be given the dataset they have.

Alex_Simmons/RST

t5nel wrote:

What do people make of the calculators at analyticcycling.com? For hill climb they tend to give me results about 15% lower than strava. Rest of the time wind has got to be the biggest factor that make strava inconsistent. I am pretty sure it is abbout as good as it can be given the dataset they have.

I reckon you must be about 4'7" tall. It's about as good as I can get with the dataset I have.
What's the point of estimating things so badly?

analyticcycling.com is as good as the quality of the information you supply it with. If your assumptions are rubbish, then the output will be rubbish - GIGO.

If the assumptions are correct, then it is a very good calculator.

willhub

People who pay premium for Strava are wasting their money, all the extras are useless.

Sometimes the power overestimates, sometimes it underestimates, it's stopped working for me but I don't care.

Oh and my Garmin is 6ft under.

jgsi

I am ignoring said power guesstimates... however, its accuracy on android Xperia phone is pretty well spot on as regards speed and pace etc when compared with Endomondo and a polar cs200 - so no complaints there.

t5nel

Alex_Simmons/RST wrote:

t5nel wrote:

What do people make of the calculators at analyticcycling.com? For hill climb they tend to give me results about 15% lower than strava. Rest of the time wind has got to be the biggest factor that make strava inconsistent. I am pretty sure it is about as good as it can be given the dataset they have.

I reckon you must be about 4'7" tall. It's about as good as I can get with the dataset I have.
What's the point of estimating things so badly?

Hmmm actually it is not as good as you can get...a far better 'reckoning' would be to use average data of height across a population - but that wouldn't support your point so well. I do not make money by selling power meters so I will let others make up their own mind about objectivity :P . For those for whom buying a power meter is not going to be an option then I really don't see the harm in estimating as long as people understand the limitations of the estimate (up a long steep hill I would expect that the estimates are at least pretty consistent). People have used estimates'calories burned' estimates for years, when you think about it this is strongly related to power and, as such, subject to the same limitations - yes take it with a pinch of salt but many of us are intelligent enough to find these figures of some use, interest or just a bit of fun!

amaferanga

t5nel wrote:

Alex_Simmons/RST wrote:

Hmmm actually it is not as good as you can get...a far better 'reckoning' would be to use average data of height across a population - but that wouldn't support your point so well. I do not make money by selling power meters so I will let others make up their own mind about objectivity :P . For those for whom buying a power meter is not going to be an option then I really don't see the harm in estimating as long as people understand the limitations of the estimate (up a long steep hill I would expect that the estimates are at least pretty consistent). People have used estimates'calories burned' estimates for years, when you think about it this is strongly related to power and, as such, subject to the same limitations - yes take it with a pinch of salt but many of us are intelligent enough to find these figures of some use, interest or just a bit of fun!

If you don't have a power measuring device then forget about power. It really is that simple. In the example you give of a steep hill you may as well just use average speed or time. On the same bike, with similar weather conditions then that'll be as repeatable measure as you can get. And at least the time you take is actually the time you take. With the power guessing the power guess could be 10%, 20%, 30%,... off the power you actually produce.

On the subject of calorie guesstimators - many of these are way way off. What does that tell you about how badly they estimate power?

t5nel

amaferanga wrote:

If you don't have a power measuring device then forget about power. It really is that simple. In the example you give of a steep hill you may as well just use average speed or time. On the same bike, with similar weather conditions then that'll be as repeatable measure as you can get.

It can still be fun though e.g. if I can calculate power then it lets me calculate how long it might take me to ascend a steeper slope at the same power - not quite so easy to do that with speed (though I accept that you can do this with VAM as well via a different route). Also helping people understand power and specific (power/kg) and how that influences what sort of riding you might be suited for is a benefit. Even if the figures aren't accurate just having them in strava, to some extent, demystifies power.

I totally accept however that if you want to use power as an accurate and useful training tool then you need some consistent device that will provide this (powetap/SRM whatever). So while you can (and perhaps should) forget about power if you don't use a powermeter I, for one, am happy that it is in strava.

Pross

GoldenBear wrote:

Another problem on leaderboards arises when someone leaves their GPS running whilst driving then not editing it out. One of the hills I use as one of my hill repeats has the fastest going up 20-22mph (depending on drafting as a few chaingangs go up) then this one lad who's gone up at 40mph. There's a few like that near me.
I use it to monitor self-improvement.

The problem for me is that a certain Mr Jim Ley and Maryka (plus a few others) used Strava on the Ras when it went over many of my favourite climbs so I can't get anywhere near the leaderboard!

(Nothing to do with me being a crap climber!!)