My Garmin head-mount has the option to either include or exclude zeros (power when coasting) in the ultimate avg power of a ride. While typically not an issue on the trainer, I’m wondering how TR handles the data for outside rides?
I include zeros for my power data, however, I know plenty of friends on my group rides that don’t include them and get a nice little boost to their average power. I was wondering how TR treats the data? Would it normalize the numbers for riders that choose to exclude zeros in their units to included them to get a more consistent interpretation across riders and across indoor/outdoor efforts?
Assume this is also an issue that has to be accounted for in AI FTP detection which looks at outside ride numbers as well.
Do the friends who are juicing their watts finish any faster? Seems like a bit of a vanity move to me. That said, your question about AI FTP detection makes a lot of sense - I wonder how they treat coasting. My gut says to include the 0s because it gives a much more accurate sense of the physiological demands of the ride
TR simply takes that data captured and provided by the head unit. There is no extra work done with respect to Zeros on/off per that device. Essentially, TR is subject to the data given.
No, it does not to that AFAIK.
I doubt it. Assuming someone nukes zeros, how would TR have any idea that the data is not “correct” and how would it make any worthwhile adjustments?
I see no practical way for them to do that with anything approaching decent data. Some might suggest looking at cadence/coasting and such, but that data is not always clear, and there are cases where pedaling may be happening with no apparent power.
The data is possible to present correctly, so swapping to educated guessing in the absence of that data is a fools errand. Not to mention that I am not sure that TR would have any real way to detect the difference in the final files presented… between ones with or without zeros included.
Bottom line, it is best to include zeros for power data. Excluding zeros for power leads to potentially flawed data and anything built upon that data (garbage in / garbage out).
By default, most Garmin devices have Non-Zero Averaging turned on. This setting throws out all the zeroes when figuring Average and Normalized power. TrainingPeaks does not throw zeroes out because time spent coasting is resting, and rest allows you to hold a higher average power when you are pedaling.
If you include 0s with power, then every time you coast (are not turning the pedals), the power meter will send a 0 (for 0 watts) to the head unit. The 0s while you coast will get averaged into your power data for the ride. The more you coast, the more 0s get sent to your head unit and the lower your average power. This is how it should be. It would be a bit meaninglessness only to average the power from when you were pedaling.
Data collection setting recommendation:
While you’re in this screen of ‘Data Recording’ you may also be interested to consider “Include Zeros” or “Do Not Include Zeros” for your Cadence and Power averages. This is a personal preference for you, but keep in mind that these settings may have a large effect on the average power and cadence reading during your ride.
Specifically states during ride, but no clarity on the file created or transmitted to other apps/services.
OK, I sent the following to support and will reply with any updates:
Does TR take pure data generated from head units for its power calculations?
Or does it investigated within the data file to find things like excluded zero power data?
Looking at this from another direction, does the file generated and shared to TR include Zero data even if the user chooses to exclude them on the head unit?
We are trying to understand the potential impact of data provided with or without zeros included, and what TR does with that (if anything) to adjust for “bad data” that excludes zeros.
Thanks, @redlude97. Yes, this is effectively what I was trying to get at in my original post. My intention wasn’t to debate wether zeros should or should not be included (I’ve come to my own conclusion by including them). As you point out was more how the choice of including/excluding them affect the data provided to TR and if the data is somehow affected by the choice, whether TR has a way to “normalize” the data. Obviously, if you are correct and the FIT file is a “raw” and includes all data regardless of the rider’s preference, my hope would be that TR includes the zeros regardless of rider choice.
So if i understand correctly, you excluded zeros for your ride and Garmin gave you an avg power of 176 watts, but the other program normalized the ride for zeros with the original fit file and it then gave you 139 watts?
If so, that’s what I’d hope the AI is also doing with outside rides. Thanks for the info.
Wonder if Strava does the same… Would be nice to know so I stop comparing my numbers to my riding buddies, as I’m usually the weakest link … I assume Strava would also normalize it to include the zeros, though. Thanks again @redlude97
Seems that this confirms the fit file is all inclusive, and most analysis apps seem to use that full data, even if the head unit excludes zeros. That is good IMO, since we can generally trust and rely on the analysis data for the long term use which matters most.
Then that should mean the excluded zero stuff is just localized to the head unit live and likely summary data. No harm I guess, other than people potentially looking at skewed data and then questioning the outside apps (which is somewhat common here, about even with a mismatch of FTP between locations).
@redlude97… out of curiosity, if you happened to also download the ride to TR…when you look at this ride as TR shows it, does it line up with the intervals ICU avg power of 139 or closer to the Garmin power number?
The discussion on the forums is addressing a really good question!
I spent some time poring over the Garmin forums and other discussions about Zero Averaging on the internet and so far I can’t conclusively confirm that Garmin’s Zero Averaging feature actually impacts the raw data. My intuition tells me that it only effects the readout on the head unit but doesn’t go so far as striking the zeroes from the record.
The best way for us to definitively confirm this would be to get power data from a Garmin head unit with both Zero Averaging on and Off. If the raw data includes zeros with both files we can say conclusively that it won’t have any bearing on the AI FTP detection since that algorithm relies exclusively on the raw data from the head unit.
If you, or possibly the original poster have the ability to send us some data to take a glance at that would be a great chance to get a real definitive answer on this one! Let me know if you’re interested in going down that path.
We might already have the answer from the recent testing above, but if @redlude97 is willing to forward related files to them, we might get a second confirmation of the results. I know you mentioned not using TR now, so I am willing to forward files if they are not up to taking them from you directly.
Seems like just one file would not be of help. Seems like what they desire is a single ride with one Garmin head unit measuring the ride with zeros and a second unit measuring it without zeros so they can compare the data.
I’d be surprised if this issue hadn’t been heavily vetted as regards AT and/or FTP Detection. As was clear from the short 15 min ride @redlude97 posted the numbers have pretty material differences.
The question is whether setting the head unit to ignore zeros for power will also remove zeros in the completed FIT file, or if the file would actually include the full data (including power zeros).
Trying to understand if the ignore zeros setting is effectively local to the head unit only, or if that setting affects the FIT files generated.
… I assume Strava would also normalize it to include the zeros, though. Thanks again