Basically what the title says. It basically leads to breakthroughs being near breakthroughs because Xert for some reason chops off 100 Watts from my bigger efforts even though I know I have actually done those efforts, because I feel them during the moment (I push harder for a second and then go back to what Xert says. But I’ve definitely pushed a little harder for a second or two).
See here for an example:
XO does filter out unusual power spikes when activities are uploaded.
Otherwise EBC is recording the power you produce second to second which you can see plotted on the power chart in the Session Player. If you don’t see a spike there it didn’t happen.
What are you using to dual record like that?
Frequent Near BT’s mean your signature is dialed and the algorithm is working as designed. Consider a Near BT as “signature validation”.
If you want to ensure a BT when MPA is drawdown, switch to Slope mode so you can exceed power, duration, or both. A max effort should be a minimum of 5-7 secs for best results.
If in AUTO mode under ERG control the interval may end before you’ve maxed out.
Well that is the thing. I‘m not dual recording. This is theoretically the same file. Garmin Connect, Zwift AND Strava all agree. Xert for some reason takes out supposed power spikes which are not power spikes but actually happened.
No ERG mode or nothing. Happens indoors and outdoors. So it‘s definitely power smoothing by Xert.
Edit: What I mean by that is that it happens consistently with different power meters and in different situations at different powers. I could show you lots of different examples unfortunately…
And that leads to Xert thinking I only did 1070W or so, even though I actually did well above 1100. that will obviously skew Peak Power numbers…
Spike removal and power smoothing are not the same thing.
PP is the least sensitive of the signature values for training purposes.
For example, it is not unusual for outdoor PP to be 150 watts higher (or more) than indoor PP.
That difference doesn’t affect your ability to train with current signature values.
Consider PP is both actual and theoretical.
Actual PP (highest power sustained for 1 sec) is what your power meter records (+/- 1 to 2% depending on accuracy).
Theoretical is what the algorithm predicts is your max PP based upon a maximal effort under fatigue (those 5-7 sec efforts with MPA drawn down).
As long as your signature is a the right ballpark it can be used as a reference point for training purposes.
To train sprint power you can simply ride a sprint workout in slope mode, so the max efforts are actual max efforts. Most SMART workouts that target sprint power switch to Slope mode for this purpose. For example, I use this session on occasion to prove/disprove my PP estimate. Or a workout like SMART - Bullet with Butterfly Wings.
If you want to take things further, you could analyze the FIT file data with DCR Analyzer.
Well I just don‘t quite understand the system behind what Xert is doing. Because it looks like it manipulates the power somehow. It kind of dampens power “spikes” and seems to even distribute the power over several seconds or something. At least I can see a consistent problem in my Xert files in anything thats not necessarily steady state for me.
I really don’t understand why Xert would do that if all the other analysis tools don’t and actually show me valid data. And it’s not even small adjustments.
For ~1000Watt sprints there is routinely 100 Watts or more missing which I definitely know were there. I have the data and I certainly feel the difference between a 900W and a 1000W+ sprint.
Once I start stringing those kinds of efforts together it becomes really problematic though. Instead of the first 1000W+ sprint already being a BT it just barely isn’t because there are 100W missing. And then obviously stringing several supposed 900W sprints together will be much less taxing than what actually happened:
Several 1000W+ sprints, with each of them actually exceeding my PP.
Now sprints are the most obvious thing but the same happens for longer efforts with frequent power surges. Pretty typical for Zwift races or punchy climbs for example. When I take my valid data and calculate my critical power and W’ it’s definitely close to Xert in predicting efforts in the 1-10 minute range, but it seems like Xert thinks I’m a different athlete type, because it’s working with fundamentally wrong data.
I can see the idea behind trying to smooth power data or maybe removing spikes or whatever. But before that’s done there should be some kind of verification that the data is actually incorrect. I‘m quite surprised, that I seem to be a rare case of this being a problem. I would have thought more people notice their power numbers being fundamentally different on different platforms.
Example of critical power calculation
Xert uses 5 second power smoothing when determining your fitness signature. This process may underrecord a peak power that was performed very briefly but in general helps alleviate data errors caused by many power meters that can overshoot real power on a second-by-second basis, even when they are very accurate on average. Some power meters have their own smoothing built in, in fact, to help avoid these errors.
As we recommend, hold down your effort for 5s or more for best results. You should be able to hold near your PP for 5s. Get your gearing right and find a moment when you don’t have to change gears during the effort.
Note that we often hear the opposite too: “Xert overestimated my Peak Power by giving me a value I didn’t achieve.” This can happen when you sprint close to your current peak under fatigue and in order to fit that effort, Xert may end up increasing your PP value.
If you’re confident PP is underestimated, then update the PP value in the activity manually and save/lock it. This will also help improve the algorithm’s ability to predict changes in PP from training.