Why are they different? The Strava Weighted Average seems to be always lower than the Garmin’s NP (Edge 830). I wonder what their respective algorithms are. Anyone has insights to this?
Weighted average is exactly what it says it is.
i.e. if you do a minute at 300w and a minute at 0w, your weighted average is 150w
in reality, that effort “felt” like more than 150w and caused more fatigue, so you’re normalized power may be 200w. the normalized power is aiming to more accurately depict the power expenditure than a pure average
That would just be average, not WEIGHTED, no?
Why are they different? The Strava Weighted Average seems to be always lower than the Garmin’s NP (Edge 830). I wonder what their respective algorithms are. Anyone has insights to this?
It looks like they’re trying to be similar but it’s not NP, it’s their own made up formula so I wouldn’t put much thought into it and just use NP.
In the case where the Garmin already has the NP number right there, why wouldn’t they also display that?
In the case where the Garmin already has the NP number right there, why wouldn’t they also display that?
It’s Strava and it’s “their data  I don’t think we have a font color for snarky as hell.
Stava’s calculation is always a lot lower than NP per Garmin and my powermeter, sometimes by 20-30 watts! I don’t know if it’s some sort of copyright thing for the calculation or if Strava is just doing their own calculation and simply not importing what the NP is. I find the discrepancy bigger when there is a higher VI on my ride. The frustrating thing for me is that my Saris H3 seems to be about 2-3% lower than my P2M cranks, and then having Strava adjust things takes even more off. Oh well…not like we’ll be racing any time soon based on this virus situation, so not all that crucial in the grand scheme.
That would just be average, not WEIGHTED, no?
They are the same in this case:
(300 + 0) / 2 = 150
(300 (.5)) + (0 (.5)) = 150
.
I understand the principle of weighted average and its intended use. I wasn’t talking about his example, which would be the same. I am more interested in general how Strava calculates weighted average, which is decidedly NOT just average as that example tries to illustrate.
The answer to example is the same for both. If you change the example to something like 2 min at 300w and 1 min at 50w, you’d get two different answers. Simple average is 300+50/2=175. Weighted average is ((2300)+(150))/2= 325.
The answer to example is the same for both. If you change the example to something like 2 min at 300w and 1 min at 50w, you’d get two different answers. Simple average is 300+50/2=175. Weighted average is ((2300)+(150))/2= 325.
I think you mean:
(300*(2/3)) + (50*(1/3)) = 216.67
Why are they different? The Strava Weighted Average seems to be always lower than the Garmin’s NP (Edge 830). I wonder what their respective algorithms are. Anyone has insights to this?
‘Normalized Power’ is a registered trademark. Strava may have chosen not to pay for it.
There are online articles that suggest Strava uses xPower, which is similar to normalized power but was developed separately. “Normalized Power uses a 30 second moving average, whereas xPower uses a 25 second exponential average.” – https://science4performance.com/...ava-ride-statistics/
If that’s the case, Normalized Power, whether that be in TrainingPeaks or your Garmin head unit, will rarely match Strava’s weighted average power.
More on xPower
http://perfprostudio.com/...io/scr/BikeScore.htm
Normalized Power, from Dr. Coggan himself:
- starting at the 30 s mark, calculate a rolling 30 s average (of the preceeding time points, obviously).
- raise all the values obtained in step #1 to the 4th power.
- take the average of all of the values obtained in step #2.
- take the 4th root of the value obtained in step #3.
https://forum.slowtwitch.com/...culator…_P3097774/
The answer to example is the same for both. If you change the example to something like 2 min at 300w and 1 min at 50w, you’d get two different answers. Simple average is 300+50/2=175. Weighted average is ((2300)+(150))/2= 325.
I think you mean:
(300*(2/3)) + (50*(1/3)) = 216.67That’s just ‘Average Power’ which is a time-weighted average.
The Strava weighted average and NP are using intensity as a weighting factor rather than simply time such that harder efforts are weighted more heavily in some fashion.
The answer to example is the same for both. If you change the example to something like 2 min at 300w and 1 min at 50w, you’d get two different answers. Simple average is 300+50/2=175. Weighted average is ((2300)+(150))/2= 325.
I think you mean:
(300*(2/3)) + (50*(1/3)) = 216.67
Yes, of course…
That’s just simply average, not weighted on anything.
Strava weighted average matches xPower in golden cheetah for my rides.
Strava is a social tool, not a performance tool. Therefore, they don’t want to pay the royalty fee for using the classic normalized power math formula developed by Coogan and his buddies (Training Peaks). Garmin does pay the royalty fee, or maybe a negotiated arrangement. Regardless, you can use either one and if you stick with it, it does it’s job, whatever that is for you. But the most universally accepted formula is the sum of the squares formula (normalized) used by Training Peaks (and Garmin). If you are using Xert to calculate or estimate your FTP from a relatively hard ride you just did, they use the normalized power. It is the standard.
Stava’s calculation is always a lot lower than NP per Garmin and my powermeter, sometimes by 20-30 watts! I don’t know if it’s some sort of copyright thing for the calculation or if Strava is just doing their own calculation and simply not importing what the NP is. I find the discrepancy bigger when there is a higher VI on my ride. The frustrating thing for me is that my Saris H3 seems to be about 2-3% lower than my P2M cranks, and then having Strava adjust things takes even more off. Oh well…not like we’ll be racing any time soon based on this virus situation, so not all that crucial in the grand scheme.
That’s probably due to drivetrain losses.
The answer to example is the same for both. If you change the example to something like 2 min at 300w and 1 min at 50w, you’d get two different answers. Simple average is 300+50/2=175. Weighted average is ((2300)+(150))/2= 325.
You are not calculating average power in the first case. It’s the average of two values, but not the average power for a session that uses those powers for different amounts of time. There’s a bit of confusion here.
The second calculation as someone else mentioned later should be ((2300)+(150))/3= 217W, but this is just an average, not a weighted average.
Think of it this way. You could state this session equally accurately as either 2min@300W + 1min@50W or 1min@300 + 1min@300W + 1min@50W.
Any calculations should give the same answers regardless how you state the reps. Your calculation for simple average would give different values:
(300+50)/2=150
(300+300+50)/3=217
The first is meaningless, the second one is correct in this case simply because the times for each portion are equal. If the time periods are not equal the result will not be correct for simple average. What you’ve called weighted average is actually a simple average and will give the same result either way. A weighted average must give higher or lower importance to a value depending on what it is, or how it’s generated. So for example if the average is intensity weighted, it would mean a higher value influences the average calculation more than a lower one. This is not the case for a simple average where high or low values have equal impact on the end result. I don’t have the NP calculation to hand but as I recall it IS essentially a weighted average, but the weighting may not be applied the same way as it is by Strava. You can choose how to apply weighting and perform a calculation accordingly.
On my basic (not premium) Strava account, it has never asked for an FTP. So…there’s that.
I think this might be why it is always lower. It probably over-estimates ftp in the algorithm, and thus will show lower than reality “NP” figures.
My guess.