Accuracy of Normalized Power for shorter durations



bradg

New Member
Apr 3, 2013
37
2
0
I was looking at some data from a crit I did not long ago, and checked the 60-second average--540W, although it being a race, there were a few spikes in there (the highest being 1024W). I was pedaling for ~80% of the duration and the VI was 1.26. The normalized power for this block of 60-seconds was 684W. Which sounds extraordinarily high. Anyone have any insight to the accuracy of that reading? I know that while NP is calculated for the shorter durations, its intent is for longer spans, yes?

I suppose I could just try hammering at around 680W for a full minute and uh, see what happens, of course. But that's not exactly on my training plan at the moment!
 
I'm not sure I understand your numbers and your question. What is the 540W number that you refer to as the 60-second average? What is the 684W number that you refer to as the normalized power for this block of 60-seconds? WKO+ doesn't even compute NP for durations shorter than 5mins, so I'm not sure where you're getting the NP number for a 60sec duration.

As a general rule, I ignore NP for efforts shorter than L4 (10+mins) and I do all of my MP benchmark testing at constant power, thus NP=AP.
 
NP for a minute is meaningless and anyone quoting figures based on NP for short durations just comes off desperate to make their power look bigger than it is.
 
You should ignore normalized power for any duration <20 min (and if I have my way, WKO+ 4.0 won't give it to you for such short durations).
 
Quote: Originally Posted by smaryka .
NP for a minute is meaningless and anyone quoting figures based on NP for short durations just comes off desperate to make their power look bigger than it is.


Understood--note well that I'm just asking about this, I'm not trying to brag about any giant power output on my part. Hopefully you're not accusing me of such! Not that I'd even consider 684W an impressive number for someone my size.

RDO: it was the normalized power reading from Training Peaks for that 60-second block. In Peaks, when I selected that 60-second duration, the average power for it was 540W, and the NP was 684. Like I said: it struck me as too much and inaccurate. I suppose part of the problem might be from using TrainingPeaks in the first place. From what you're saying otherwise, I'd just assume test my CP1 on the CS-1000 or a half-mile stretch of open, flat road then. I will assume that this race data isn't helpful for determining any "targets" in a test.

Andy: Thanks, that's what I thought but I figured I'd check regardless. I would also like for WKO+ be Mac compatible :) Still at Wash U? You made it up to Penrose at least once this season, right?
 
bradg said:
Understood--note well that I'm just asking about this, I'm not trying to brag about any giant power output on my part. Hopefully you're not accusing me of such! Not that I'd even consider 684W an impressive number for someone my size. RDO: it was the normalized power reading from Training Peaks for that 60-second block. In Peaks, when I selected that 60-second duration, the average power for it was 540W, and the NP was 684. Like I said: it struck me as too much and inaccurate. I suppose part of the problem might be from using TrainingPeaks in the first place. From what you're saying otherwise, I'd just assume test my CP1 on the CS-1000 or a half-mile stretch of open, flat road then. I will assume that this race data isn't helpful for determining any "targets" in a test. Andy: Thanks, that's what I thought but I figured I'd check regardless. I would also like for WKO+ be Mac compatible :) Still at Wash U? You made it up to Penrose at least once this season, right?
Which Wash U? The one in St. Louis?
 
Quote: Originally Posted by bradg .

Understood--note well that I'm just asking about this, I'm not trying to brag about any giant power output on my part. Hopefully you're not accusing me of such! Not that I'd even consider 684W an impressive number for someone my size.

No, not accusing you but many people don't understand how NP is calculated or what its purpose is; they just see bigger numbers and get wide-eyed. ;)
 
Quote: Originally Posted by bradg .

Understood--note well that I'm just asking about this, I'm not trying to brag about any giant power output on my part. Hopefully you're not accusing me of such! Not that I'd even consider 684W an impressive number for someone my size.

RDO: it was the normalized power reading from Training Peaks for that 60-second block. In Peaks, when I selected that 60-second duration, the average power for it was 540W, and the NP was 684. Like I said: it struck me as too much and inaccurate. I suppose part of the problem might be from using TrainingPeaks in the first place. From what you're saying otherwise, I'd just assume test my CP1 on the CS-1000 or a half-mile stretch of open, flat road then. I will assume that this race data isn't helpful for determining any "targets" in a test.

Andy: Thanks, that's what I thought but I figured I'd check regardless. I would also like for WKO+ be Mac compatible :) Still at Wash U? You made it up to Penrose at least once this season, right?


There are a number of reasons for NP not being a sensible value for inspecting short duration performance, both physiological and mathematical. Software that shows this value for 1-minute makes no sense at all.

In terms of the mathematical, keep in mind the NP algorithm is in part based on a rolling 30-second average power as a first step, IOW it requires at least 30-seconds of data for the algorithm to "kick in", and the way it handles the final 30-seconds of data also impacts the NP number it spits out (not sure the algorithm is clear on precisely how that final 30-seconds should be accounted for). e.g. when I calculate NP, the final data point only gets included in one rolling 30-second average, whereas the 30th last data point is included in 30 individual rolling average values (using a one-second recording rate). The same issues applies to the initial data points, i.e. the first data point only counts to one 30-second average, the second to two, and so on.

As a result, the beginning and end of the power data stream are to some extent underweighted, hence this can artificially impact the NP number, moreso when the overall duration being considered is short.
 
Quote: Originally Posted by Alex Simmons .
As a result, the beginning and end of the power data stream are to some extent underweighted, hence this can artificially impact the NP number, moreso when the overall duration being considered is short.

This is most noticeable when you do a short (5-10 min) interval with a sprint start and finish (e.g., a hill climb race) where the power is otherwise fairly steady and constant. My NP is often a couple of watts below my AP on those kinds of efforts!
 
Quote: Originally Posted by bradg .
I was looking at some data from a crit I did not long ago, and checked the 60-second average--540W, although it being a race, there were a few spikes in there (the highest being 1024W). I was pedaling for ~80% of the duration and the VI was 1.26. The normalized power for this block of 60-seconds was 684W. Which sounds extraordinarily high. Anyone have any insight to the accuracy of that reading? I know that while NP is calculated for the shorter durations, its intent is for longer spans, yes?

I suppose I could just try hammering at around 680W for a full minute and uh, see what happens, of course. But that's not exactly on my training plan at the moment!

NP has no meaning. So accuracy has no meaning.

The whole concept was written up by Andy Coggan as a gimic to get people to use trainingpeaks.com. (At least according to the article by Andy Coggan on the trainingpeaks.com site.)

Andy Coggan cannot even give a clear description of what NP and the other terms measure.
 
Quote: Originally Posted by An old Guy .
NP has no meaning. So accuracy has no meaning.

The whole concept was written up by Andy Coggan as a gimic to get people to use trainingpeaks.com. (At least according to the article by Andy Coggan on the trainingpeaks.com site.)

Andy Coggan cannot even give a clear description of what NP and the other terms measure.

NP is there to help estimate training stress. If you've ever done a ride at a very steady wattage (low variability, no surging) then compared it to a ride with the same average watts but with lots of surging and freewheeling (high variability) then you're know that even though those two rides might have the same average watts, the training stress of the latter is quite a bit harder for most people than the training stress of the former. So NP is an attempt to quantify the stress of surgy riding where AP doesn't paint a good enough picture of how hard it was.

If you're going to slate Coggan and Training Peaks and NP and the rest, at least get your facts straight.
 
  • Like
Reactions: steve
Quote: Originally Posted by smaryka .
NP is there to help estimate training stress. If you've ever done a ride at a very steady wattage (low variability, no surging) then compared it to a ride with the same average watts but with lots of surging and freewheeling (high variability) then you're know that even though those two rides might have the same average watts, the training stress of the latter is quite a bit harder for most people than the training stress of the former. So NP is an attempt to quantify the stress of surgy riding where AP doesn't paint a good enough picture of how hard it was.

If you're going to slate Coggan and Training Peaks and NP and the rest, at least get your facts straight.


don't feed the troll
 
Quote: Originally Posted by smaryka .
NP is there to help estimate training stress.


If you're going to slate Coggan and Training Peaks and NP and the rest, at least get your facts straight.


On trainingpeaks.com Andy Coggan makes the claim it measures glucose utilization. acoggan posted on this site to say that is not correct. Andy Coggan makes the claim it measures fatigue - some combination of glucose utilization, lactate levels, and etc. But omits the functional relationship.

On trainingpeaks.com Andy Coggan wrote that the formula for it is based on a formula for an unrelated undisclosed physical process. It is not based on any model of fatigue, glucose utilization, or lactate level.

acoggan wrote on this site it is not science. Andy Coggan has never produced science to support it. acoggan makse predictions about it and then makes the statement that it does not make predictions.

acoggan wants science based proof that it is invalid. Tough to do without a reasonable defiinition if it. Impossible to do since it is not science.

acoggan has demanded proof of my statements. I have offered to provide proof, but acoggan has never indicated what proof he will accept. I have stated I am not going to attempt to disprove some religious claim, without the profit agreeing on what constitutes proof.

----

I used the pronoun "it" to refer to NP, IP and all those other terms.

---

Are there any facts I have missed?

I appreciate your comments about "training stress." Yes different rides will produce different effects, but without a way to actually measure whatever NP is intended to estimate, there is no way to determine if NP correctly measures it.
 
Quote: Originally Posted by bradg .
I would also like for WKO+ be Mac compatible :) Still at Wash U? You made it up to Penrose at least once this season, right?


1. Yes, there will be a Mac-native version of 4.0.

2. Yes.

3. Yes, when they asked my wife to come out.
 
Quote: Originally Posted by dkrenik .
AOG, you're one funny dude.


It is surprising how gullible people are.

I happened on a paper that said Andy Coggan's formula estimated lactate levels. Praising it because one did not need an invasive blood test. The author has a PhD and is well respected in whatever field this is. Even he has no idea what NP and all the other letters are intended to measure.

Andy pulled one over on a lot of people. Many smarter than you.
 
Quote: Originally Posted by Alex Simmons .

There are a number of reasons for NP not being a sensible value for inspecting short duration performance, both physiological and mathematical. Software that shows this value for 1-minute makes no sense at all.

In terms of the mathematical, keep in mind the NP algorithm is in part based on a rolling 30-second average power as a first step, IOW it requires at least 30-seconds of data for the algorithm to "kick in", and the way it handles the final 30-seconds of data also impacts the NP number it spits out (not sure the algorithm is clear on precisely how that final 30-seconds should be accounted for). e.g. when I calculate NP, the final data point only gets included in one rolling 30-second average, whereas the 30th last data point is included in 30 individual rolling average values (using a one-second recording rate). The same issues applies to the initial data points, i.e. the first data point only counts to one 30-second average, the second to two, and so on.

As a result, the beginning and end of the power data stream are to some extent underweighted, hence this can artificially impact the NP number, moreso when the overall duration being considered is short.


This clears it up nicely, thank you Alex.
As for any comments about Coggan's credibility (which I'll likely never see again, given the handy "block" feature which I urge everyone else to use), I think there was an earnest attempt to quantify efforts and output and like any other tool out there, it's not a cure-all but it sure does help if you know how to use it and how to manage expectations, so to speak. Anyway, the point of this thread was to get an idea of whether or not NP had any relevance in efforts under 20min, and clearly it doesn't. Problem solved, thanks for the contributions and insights.
 
Quote: Originally Posted by bradg .
I was looking at some data from a crit I did not long ago, and checked the 60-second average--540W, although it being a race, there were a few spikes in there (the highest being 1024W). I was pedaling for ~80% of the duration and the VI was 1.26. The normalized power for this block of 60-seconds was 684W. Which sounds extraordinarily high. Anyone have any insight to the accuracy of that reading? I know that while NP is calculated for the shorter durations, its intent is for longer spans, yes?

I suppose I could just try hammering at around 680W for a full minute and uh, see what happens, of course. But that's not exactly on my training plan at the moment!


I guess everyone got the answer wrong. (Since you are ignoring me, you will not get the correct answer.)

The calculation is accurate. (The result might be worthless but that is another issue.) It is accurate because the value depends on 90 seconds of data - 30 seconds before your interval and the 60 seconds of you interval. You have all of that data.

The only "error" occurs during the first 30 seconds of your workout. That is because the previous 30 seconds of data is not available.

---

In terms of TSS something like

TSS(from 0 to A) - TSS(0 to A+delta) = TSS(A to A+delta)

from that you can compute NP for time delta.

At least you could if the concept has any substance.