Average Power higher than NP...?
I just did my JOS FTP test this morning (a couple days early). I think it was a successful test even though it was a little lower than I had hoped, but about what i would have expected. http://www.trainerroad.com/cycling/rides/146653
I dropped the results (from my Garmin Edge 500) into wko+ and am comfortable using 335W for my VO2 number and 280W for my FTP number to start the JOS. At 193lb, this gives me a W/Kg of ~3.16 (down from ~3.45 at my peak last summer). This FTP number is down from 301W when I did IMNYC in August and the 285W I raced IMFL at in early Nov. So the numbers seem about right and I clearly have some work to do getting my W/Kg up if I want to have a good race in hilly IMUSA for someone my size.
So here's the thing I thought was weird and figured I'd throw it out to you all so see if anybody has seen it before or can explain it:
5:00 VO2 - 295W AP, 294W NP with a wko+ calculated VI of 0.99
20:00 FTP - 338W AP, 334W NP with a wko+ calculated VI of 0.99
I have never seen VI numbers below 1.00 before. Said another way, how is it possible that Normalized Power is actually lower than Average Power? Again, I'm confident that the results of the test are good and am not really asking what I should use for training, I'm using 280 FTP and 335 VO2 regardless of any feedback on them... This was more of a theoretical math wizzy type of question about how NP could be below AP...
As a Nerd side note, I use redundant data collection for tests (Garmin and Trainerroad). I think the Garmin data is slightly better than the TR data because on occasion I can see the numbers go to 0 for a second on TR as the ANT+ stick seems to lose contact or something. Again, the numbers are always super close, but almost always the Garmin data is like 1W higher and I'm sure it is because of the few 0 numbers that are averaged in with TR. This is much more pronounced with the HR data as you can see that TR actually misses this data about 1/3 of the time. I should probably get my bike closer to my computer, but for workout purposes, I don't really care because it's close enough.
Comments
I have had this happen to me when I had a big power spike right at the end of the intervalay its normal. As you know, normalized power is just a different method of calculating an average. Also NP and test's less than an hour, you can have this happen more often not so much with an hour test.
See your spike at the end ?
And as to the inconsistency, here's an even stranger one...1-2 watt differences between the Garmin and TrainingPeaks, where TrainingPeaks uses the file uploaded from the Garmin as its data source. Bizarre.
If these inconsistencies were over 4-5 watts I'd sweat them but at this point I know +/- a watt or two isn't going to swing any of my data materially.
Cheers!
Hi John,
This is normal for a short interval at level power, where short is defined as significantly less than an hour.
It's the nature of the equation that AP may be greater than NP for intervals of less than an hour. The shorter the interval, the greater the likelihood that AP>NP. For example, the gap is much greater on your 5m VO2 interval; the values are effectively the same on your 20m interval. As interval length at steady power approaches one hour, NP will approach AP from below. You wouldn't see AP>NP for an interval of 1hr or greater.
Now... WHY this is... without diving too deep into the weedy fine points of the equations... AP is pretty simple arithmatic mean: sum(measurements)/count(measurements). NP takes a number of factors into consideration, including distance above the mean of instantaneous power and duration of sustained effort ... with the goals of (a) supporting calculation of a TSS - a stress score that could be used to optimize the total training load on an athlete - and (b) predicting optimal power output for a 1-hour TT.
Because your body doesn't feel a 20m effort the way it would feel a 1-hour effort, the TSS (stress score), based on the NP equation of a 20m interval is actually not that high - potentially less than the AP.
If you have a lot of variabillity over any interval, meaning peaks and valleys in your power output, then your AP over the time will be low relative to your NP because you're paying the price exponentially on power peaks above your threshold.
This physiological truth, by the way, is why our coaches tend to break our workouts into interval efforts. You can rack up a lot of time just below full power, get a great training effect in return, for an affordable investment in terms of time and training load.
If you want to know more, I'd start with Skiba:
www.physfarm.com/govss.pdf
The telling quotes:
Banister’s system .. that it takes into consideration the observation that higher workloads are more metabolically taxing (exponentially so) than lower workloads of equivalent duration.... Andrew Coggan refined Banister’s concept by developing a system that also incorporated lactate response to workload. This system related the change in lactate concentration with the change in an objective measure of exercise intensity: power output...
Hope this helps - I can go nerdier if you like - I even have the pocket protector and glasses, but I'd have to go get them - they are in the other room.
Best!
Russ
@ john ... I f you have the 2nd edition of Allen & Coggin's Power Meter book, go to page 120, where the formula for NP is explained. Basically, it uses a 30 second rolling average for the power. Meaning that the first 30 seconds of any interval will include data from the previous interaval, which, if it was @ lower power, will lower the NP. The AP, of course, is much more instantaneous, and so for short intervals, you *can* get an NP below the AP if you have been pedaling (as you do on a trainer) at a constant effort level.
I'm guessing this is the reason Training peaks software doesn't even calculate an NP for intervals less than 5 minutes. PowerAgent (PowerTap) doesn't show it for less than 60s, and the Joule doesn't show the NP/IF for the first 30 seconds of any interval.
Al has it right about the 30s averaging and why NP isn't calculated for less than 5 minutes.
It probably is the case that this averaging BEFORE the raising to the 4th power can end up giving you a value less than 1 if you have spikes (and zero/near-zero moments).
For the record, xPower is the analog of NP that you will see elsewhere and it is almost always a little less than the NP. It uses the same formula (fourth root of the average of the fourth power), but it uses an exponentially weighted time-average instead of a rolling (boxcar) average.