Lower NP than AV during intervals

[giannip]

I did 3 x 10min intervals today and I noticed that the NP for all 3 is between 1 & 3 watts lower than the AP.


I'm pretty much used to the NP being either the same or slightly higher so I was wondering if this is an issue with the software (upgraded to WKO+ 3.0 few days ago), which I doubt, or if there's another explanation, i.e. length of the interval.


A similar workout 1 week ago has the NP lower for 2 of the intervals but higher for the 3rd.


I know it's a relatively short interval, so I don't want to assume there's something wrong and just thought I would ask.




Thanks in advance.

 

[daveryanwyoming]

It's a mathematical anomaly that stems from the 30 second rolling average applied to your power data as part of the NP algorithm. It's most common in short efforts and happens if the first 30 seconds is above the average for the interval.

-Dave

 

[giannip]

Great thanks.

I'll have to slow down then ;-)

 

[jollyrogers]

It's a mathematical anomaly that stems from the 30 second rolling average applied to your power data as part of the NP algorithm. It's most common in short efforts and happens if the first 30 seconds is above the average for the interval.

-Dave

Would a much lower, say recovery zone, power over the 30 seconds immediately prior to the interval also affect interval NP.

 

[daveryanwyoming]

Would a much lower, say recovery zone, power over the 30 seconds immediately prior to the interval also affect interval NP.

I wasn't sure about how WKO+ implemented the NP algorithm, but I just did some playing with data and no, data before or after the interval in question doesn't seem to influence the NP calculation.

From what I can tell they start the sliding 30 second window at the start of the interval so the first data raised to the fourth power in the NP calculation is 30 seconds into the selected interval and they stop the sliding window averaging when it hits the end of the interval. Hand editing data to add 30 seconds of zero power or really high power data points before or after the selected interval doesn't change the NP of the interval. So based on that, the NP for an interval takes the rolling average starting 30 seconds in and the last data point raised to the fourth power is the point where the sliding averaging window hits the end of the interval but AP of course takes data from the complete interval.

 

[jollyrogers]

I wasn't sure about how WKO+ implemented the NP algorithm, but I just did some playing with data and no, data before or after the interval in question doesn't seem to influence the NP calculation.

From what I can tell they start the sliding 30 second window at the start of the interval so the first data raised to the fourth power in the NP calculation is 30 seconds into the selected interval and they stop the sliding window averaging when it hits the end of the interval. Hand editing data to add 30 seconds of zero power or really high power data points before or after the selected interval doesn't change the NP of the interval. So based on that, the NP for an interval takes the rolling average starting 30 seconds in and the last data point raised to the fourth power is the point where the sliding averaging window hits the end of the interval but AP of course takes data from the complete interval.

Thanks - I wasn't sure how WKO+ started computing NP in an interval. My assumption as to why I had seen NP<AP, even on 20 minute intervals was that that NP for the first second of the interval included the previous 29 seconds in the rolling average, which for me is typically L1.

 

[giannip]

Tonight's 10min had the same "issue". I checked the first 30s and the AP it's lower than the total interval AP.

During second 2, the power is higher than the AP. Could just one second in the first 30s cause this ?

The NP was 2 watts lower than AP.

 

[daveryanwyoming]

...Could just one second in the first 30s cause this ?...

Yes, if that one second occurred during the first or last 15 seconds of the interval with a power well above the average for the effort. With a 30 second sliding window that starts at the beginning (doesn't take into account data prior to the start of the interval) and ends when it hits the end of the effort (doesn't take into account data after the end of the interval) values in the first and last 15 seconds are under-valued in terms of NP calculation. Values there much above the average for the interval will raise AP but have less impact on NP.

But as others have said, it really doesn't matter. The value of NP is to estimate effective metabolic load and that really applies to longer efforts that are primarily metabolic in nature. Ten minutes is really on the short end of primarily metabolic efforts and NP for efforts that short really isn't very meaningful. As intervals get longer the impact of the first and final 15 seconds on the overall AP and NP becomes less and less and as it is the differences you're seeing are roughly on the scale of power meter accuracy and not very important. Neither AP nor NP are so accurate that we can track them to the last watt or two.

-Dave

 

[acoggan]

The value of NP is to estimate effective metabolic load and that really applies to longer efforts that are primarily metabolic in nature. Ten minutes is really on the short end of primarily metabolic efforts and NP for efforts that short really isn't very meaningful.

Alas, I tried to take away people's ability to calculate (in WKO+) normalized power for any effort <20 min duration, but the masses would have none of it.

 

[frenchyge]

It's a mathematical anomaly that stems from the 30 second rolling average applied to your power data as part of the NP algorithm. It's most common in short efforts and happens if the first 30 seconds is above the average for the interval.

An easy way to view what Dave's talking about is to go to the graph with 0-second averaging applied and drag over the interval so that it is selected in dark blue from start to finish. Now change the averaging on the graph to 30-seconds and see what happens to the power trace within that dark blue (selected) area.

It appears to me that it's really the low power period *after* the interval which drags to NP downward slightly through the averaging, but you'd have to know the code to be sure. Regardless, NP across the entire set is more indicative than NP within a single interval.

 

[giannip]

Thanks everyone.

Not too bothered about a few watts difference. Just interested to know why.

 

[daveryanwyoming]

...It appears to me that it's really the low power period *after* the interval which drags to NP downward slightly through the averaging, but you'd have to know the code to be sure.l.

That was my thought as well, but I fiddled with editing in very high (e.g. 1000 watts) and zero power into the seconds prior to or following the intervals of interest and neither data edit impacts NP.

It's easy, select an interval and make it visible so you can find it, hit the numeric display (123) button up top in WKO+ to display the data in tabular form. Hand edit different data values before, after, first or last 30 seconds, etc. and see what happens to the selected interval in terms of AP and NP.

It seems clear they've coded the sliding window to start at the beginning of the interval so the first output of that filter is at second 30 and it stops when it hits the end of the interval so last filter output at t-30. So those values in the first and last 15 seconds, especially the first and final second get under valued in terms of feeding into NP.

-Dave