I don’t think this assumption works. When I loaded 8 months worth of data with a lot of free rides, my metrics were completely out of wack. Instead of plateauing, XPMC chart was showing wild swings in HEI, PP and TP with every ride, and practically ignoring workouts that felt a lot harder than rides. Trying to flag/edit some of the activities to make it look reasonable felt like trying to squeeze a water balloon. I ended up dumping all of the data and reloading only last couple of months with mostly workouts, and still had to edit or flag my outdoor rides. Now the metrics look good, but to keep doing it manually doesn’t seem to be a good approach, especially with more group rides coming up.
It is not just by feel - in about 6 months or unstructured “difficult” riding with BT on almost every other ride, my real FTP practically didn’t change and may be even slightly declined. (yet according to Exert it went all the way to what I can sustain for about 6-7 mins max) Within last 2-3 months, doing workouts that’s according to Xert were “easy” and so much less strain than rides that they barely registered on XMPC, my FTP went up from 235 to about 255.
So, something is definitely wrong with the way rides with a couple of short, hard efforts, but also a lot of coasting, stopping, etc. influence metrics significantly more than very productive and consistent workouts.
This seems related to your other posts, and see over there that some changes were made to your signature - did that improve things / are you still having issues since then? If you’ve not tried riding hard outside again since, maybe give that a go first.
You shouldn’t have to flag rides (apart from true exceptional data issues) to get an accurate signature, so something is still not right if you still see wild numbers after riding outside (starting with an accurate signature)
You said that power meters indoor and outdoor were similar… does that mean on average, or also at all points in time / accounting for power spikes? Similar meaning 1 to 2% difference, or more?
If anything, outdoor riding should be easier on average or per hour, due to the coasting and stop/start you mention. That said, it wouldn’t be unusual for a hard outdoor effort to result in slightly higher numbers for a range of reasons… but that doesn’t sound like your issue here
Just brainstorming / trying to help as I don’t have similar issues, and haven’t heard of many that do
As I mentioned, I ended up dumping data and reloading just last couple of months from Strava.
I had couple of typical rides last week with some short hard efforts. Resulted in gold and silver BTs and it again bumped up my TP to unsustainable level, so I had to edit activities and undo TP changes.
I just did another powermeter test and the differences are much higher then I remember. On average, it’s close enough, but crank-based data is A LOT more spiky, which actually makes sense - power fluctuation from motor in a rear hub must be a lot smoother than crank strain gauge that is a lot more sensitive and has no inertia.
This might be a contributing factor, but i strongly doubt it is the main reason for erratic metrics from rides.
Do you see wide variations in signature values from one day to the next when you view the Activities table?
Or are you referring to the squiggly lines for PP, TP and HIE when the Progression chart is set to Fitness Signature?
Both. Signature values jump by a lot from one activity to the next AND consequently both XPMC and signature lines VERY squiggly. I can’t show it now as I deleted the data. With mostly workout data now, lines are obviously not straight, but not crazy either. I think the system can’t properly analyze rides with very high variability of power. Something like this:
This wasn’t even a particularly hard ride, but it affected my signature by A LOT more than any workout that I had to push myself very hard to complete.
I’ve found Xert to work very well for high power efforts, even if highly variable… if the power readings are correct, that is.
From your powermeter comparison it looks like the crank based one reads quite a bit higher at higher power (10 to 20w from 5s to 2m) which could have a big impact on MPA drawdown… and they weren’t actually high power efforts compared to your outdoor rides… can you do a test with high power and variable efforts like you do outdoors?
And not sure if you have dual sided or single sided 4iiii, which could be another factor
That looks like a normal spiky outdoor ride profile except MPA seems too low to have been pulled down that far especially if it was not “a particularly hard ride”.
Were you deliberately jumping or surging during those red spikes or just riding along?
MPA too low would explain frequent BTs.
What do others think?
Reference: Breaking Through the Xert Way! – Xert (baronbiosys.com)
Examples: Let’s See Some Recent BT’s - General - Xert Community Forum (xertonline.com)
For comparison purposes here are some of my Fitness Signature progression charts. These will vary considerably between individuals but the patterns should be similar.
This one is me moving from Base in Dec through Build (as of today) with a TED set in April – riding exclusively indoors so far.
Here’s what it looks like when I set filter to Year –
That includes some gaps off the bike and floating between July and November (no progressive TL configured).
I only have a few activities flagged from outdoor rides with unusual power spikes. (I have a Stages left crank meter.)
Here’s mine showing me peaking outdoors in July 2019 (TED was August) then tapering off and riding mostly indoors during 2020 through today.
Notice the big difference in PP outdoors versus indoors. A major factor is the inability to sprint indoors like you can outdoors. My data is also skewed by some health issues but that’s another story. 
Are these lines and patterns similar to yours?
Not sure how it works on other trainers, but at least on Neo, it take 5 sec or so to go from 0 to let’s say 400W, where with strain gauge on crank it’s instant. I don’t think I can produce same kind of variability on a trainer.
It’s a left side only, but if anything it should lower power a bit - my right leg is stronger then left (after I broke my left leg a few years ago)
I live in the area with a lot of small climbs/rollers. It’s hard to find a flat route - you go either up or down 80% of the time, and I like to go hard on climbs.
I don’t have my charts from were I had all the data. But it looked more jagged.
With mostly workouts it looks fine (not as good of a progression as yours - good job!)
Interesting - I rode relatively hard ride today - 50miles/3800ft climbing with quite a few PRs on climbs. And the results in Xert were exactly what I would expect - near BT, and only very small changes in signature. This is night and day from what I’ve seen before. The only difference I can think of is that I forgot to turn on auto-pause. It significantly lowered my ave speed -13.7. (It would be close to 17 with auto-pause on). Unless it’s my whining that made Xert team to adjust something in algorithm, I might’ve found the problem. I will do another ride with autopause on tomorrow and see how it compares.
Hey Simon,
Power meters with greater fluctuation tend to create higher signatures when those errors happen during the breakthrough itself. In addition, longer efforts around TP will tend to skew numbers higher too. This is due to the fact that power above threshold isn’t equally balanced with power below threshold. This fatigue/recovery conundrum is the reason why sustained efforts with small stoppages tend to underestimate the true fitness capability of the athlete.
Autopause isn’t material since the system will fill in the pauses before analysis.
Not necessarily errors. Strain gauges can register rapid increase in forces that last a very short time.
That would probably explain the wild signature swings I have seen. Wouldn’t it make sense then to discount / assign lower weight to results from rides with very high power variability, so they don’t mess up signatures?
Great idea but would have to determine how to establish whether variability is real or not. Could be done but likely a somewhat more of a “project” rather than a simple thing to do.
I don’t think it’s important at all. Assume that it’s all real. Assign weigh inversely proportional to variability of dataset.
I’m not sure what variability measure you want to use though. I want to say Coefficient of Variation, but I don’t remember that much from university statistics. (Have I even passed it? 
)
Though according to this theory you should heavily discount my very random thoughts! 
Sounds like you are talking about riding in ERG mode, where it takes time for power levels and resistance to change? If you ride in slope mode you should be able ramp power more quickly up and down, just like outside… the Neo can capture power surges, like most modern / direct drive trainers (if it were wheel-on you may miss something due to tyre slippage, but not the case here). For sure in slope mode you could alternate between 30s at 2 or 3 x TP and 30s easy quite a few times, or 10s all out and 50s easy (depending on your signature of course) if you do want to test the power meters. Can even push for a BT if you are keen 
Anyway, sounds like your most recent ride outside gave better results which is good. Could be due to the nature of the ride as you say… wonder if it could also be due to @ManofSteele’s signature tweak where HIE was increased to more realistic levels (from the other thread)? But that’s one for the Xert guys to comment on.
That explains it.
I remember those days. Now I am more inclined to be “uh-oh, better save something for later”.
You are probably right about ERG, but still motor power response is order of magnitude slower then strain gauge. I will give it a try anyway (when I find enough motivation for this 
)
Update. I did another ride today. It was actually much easier then yesterday. The group was quite mixed, and for the most part I was noodling along, enjoying a nice, cold sunny day, but I did a few hard efforts, which was easy since overall level was low. Got a gold BT and both PP and TP bumped up by 20W 
Two differences I could think of from yesterday - I had autopause on today (but Armando says it makes no difference), and yesterday’s overall level was higher - not as much difference from baseline to max effort.
What I’m going to try is manually reduce signature changes from activities like this by 3/4. This is very arbitrary, but probably less arbitrary and more consistent then randomly mocking around with HIE.
I wish there was less arbitrary way to deal with it though… 
Yeah something really strange about your data. Your PP is very low for someone with your weight, TP and HIE. This can cause TP to be a lot higher, especially when BTs are at high wattages where PP is contributing a very high proportion of the effort (e.g. PP accounts for roughly 1/3 of the power at your 1 minute power). Your efforts are also very spiky and errors in the data will compound the errors in the signature.
Which power meters are you using? P2Ms are like clockwork (never see issues like this). Some older Quarqs can be more problematic. Left side only PMs can be superb or troublesome, dependng on a variety of factors. Powertaps are great usually - hubs and pedals. Older Garmins … umm… Assiomas are great but abrupt stoppages might let an extra second or two of data creep in which can throw off things during a BT.